trace2 output for bitmap decision path
cover trace2 output for bitmap decision path
A Git repo can be chosen to use the normal bitmap (before MIDX) and MIDX bitmap.
I recently tried to understand this part of the MIDX implementation because I found a bug which has been discovered and repaired in community [1].
I am grateful to Taylor Blau for his help and for introducing me to the testing method according to the git rev-list --test-bitmap <rev>.
In the process of understanding and troubleshooting by using this command, I found when the execution is failed it will output a single line of
"fatal: failed to load bitmap indexes", sometimes will be more informations like if the bitmap file is broken, the outputs maybe contain
"error: Corrupted bitmap index file (wrong header)".), but most of time are single line output I mentioned above. So, this brings a little obstacle for debugging and troubleshooting I think, because "failed to load bitmap indexes" can represent to much informations (many causes like: midx config turn off, bitmap inexistence, etc.)
Therefore, as a git repo can be chosen to use the normal bitmap (before MIDX) or MIDX bitmap, or they can both exist and let git make the decision. I think why not add some extra informations based on TRACE2 to help for showing the bitmap decision path clearer and more plentiful, so when the failure occurs the user can use it to debug around bitmap in a quicker way.
Thanks.
Links:
1. https://public-inbox.org/git/[email protected]/)
Teng Long (3):
pack-bitmap.c: use "ret" in "open_midx_bitmap()" pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
bitmap: add trace outputs during open "bitmap" file
midx.c | 2 ++ pack-bitmap.c | 17 +++++++++++++---- 2 files changed, 15 insertions(+), 4 deletions(-)
Range-diff against v0:
-: ---------- > 1: 3048b4dd29 pack-bitmap.c: use "ret" in "open_midx_bitmap()" -: ---------- > 2: 70500b6343 pack-bitmap.c: add "break" statement in "open_pack_bitmap()" -: ---------- > 3: 9912450fc1 bitmap: add trace outputs during open "bitmap" file
Signature · 1 line
2.35.1.579.g70500b6343.dirty
PATCH 1/3 pack-bitmap.c: use "ret" in "open_midx_bitmap()"
Let's use "ret" value for "return" statement in "open_midx_bitmap()" just as the same way as int "open_pack_bitmap()".
--- pack-bitmap.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-)
pack-bitmap.c +6 -3
@@ -494,15 +494,18 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) { -if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; +if (!open_midx_bitmap_1 (bitmap_git, midx)) { + ret =0 ; +break ; + } } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.1.579.g70500b6343.dirty
Re: [PATCH v1 1/3] pack-bitmap.c: use "ret" in "open_midx_bitmap()"
Quoted 13 lines · Teng Long
pack-bitmap.c +1
-0@@ -494,15 +494,18 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);
pack-bitmap.c static int open_pack_bitmap(struct repository *r, · line 508
Quoted 10 lines
pack-bitmap.c +5
-3for (midx =get_multi_pack_index (r); midx; midx = midx->next) { -if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; +if (!open_midx_bitmap_1 (bitmap_git, midx)) { + ret =0 ; +break ; + } } -return -1 ; +return ret;
This all looks like good clean-up to me, and it indeed preserves the behavior before and after this patch is applied.
But thinking about some of my comments on patch 2/3 here, I think that we don't want to break out of that loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
That _is_ a behavior change with respect to the existing implementation on master, but I think that what's on master is wrong to stop after looking at the first MIDX bitmap. At least, it's wrong in the same sense
of: "we will only load _one_ of these MIDX bitmaps, so if there is more than one to choose from, the caller is mistaken".
I think instead we'd want to do something like this on top:
--- 8< ---
pack-bitmap.c +1 -3
@@ -500,10 +500,8 @@ static int open_midx_bitmap(struct repository *r,assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) { -if (!open_midx_bitmap_1 (bitmap_git, midx)) { +if (!open_midx_bitmap_1 (bitmap_git, midx)) ret =0 ; -break ; - } }return ret; }
--- >8 ---
Thanks,
Taylor
Re: [PATCH v1 1/3] pack-bitmap.c: use "ret" in "open_midx_bitmap()
Quoted 10 lines · Taylor Blau
But thinking about some of my comments on patch 2/3 here, I think that we don't want to break out of that loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
That _is_ a behavior change with respect to the existing implementation on master, but I think that what's on master is wrong to stop after looking at the first MIDX bitmap. At least, it's wrong in the same sense
of: "we will only load _one_ of these MIDX bitmaps, so if there is more than one to choose from, the caller is mistaken".
I'm a little wondering that what's the practial meaning for _ do not _ stop after looking at the first MIDX bitmap?
Although all MIDX bitmap are scanned, only one of them will eventually work , and there seems to be no guarantee that the last MIDX that works is the most appropriate one? (with the same comfusion applies to non-MIDX bitmap's behavour...)
Thanks.
Re: [PATCH v1 1/3] pack-bitmap.c: use "ret" in "open_midx_bitmap()
Quoted 4 lines · Teng Long
Although all MIDX bitmap are scanned, only one of them will eventually work , and there seems to be no guarantee that the last MIDX that works is the most appropriate one? (with the same comfusion applies to non-MIDX bitmap's behavour...)
I think your "which is the most appropriate one?" exactly highlights the problem. Since Git can only handle one .bitmap file at a time, we should tell the user when we found more than one candidate.
In this case, they may be expecting or depending on a particular choice for performance reasons, and so may be rather surprised or upset if Git chose a different available .bitmap.
The one exception to this is when we have both a single- and multi-pack bitmap available, in which case it almost always makes more sense to pick the multi-pack bitmap. So we don't complain in this case (i.e., if open_midx_bitmap() succeeds, we immediately return success from open_bitmap() and proceed to have prepare_bitmap_git() call load_bitmap() on the previous result without even bothering to check for pack bitmaps).
Thanks,
Taylor
PATCH 2/3 pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
In reply to Teng LongThere will be only one operant ".bitmap" file in repo, so let's add "break" statement in "open_pack_bitmap()" when looping all the packs in repo.
--- pack-bitmap.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
pack-bitmap.c +3 -1
@@ -484,8 +484,10 @@ static int open_pack_bitmap(struct repository *r,assert (!bitmap_git->map);for (p =get_all_packs (r); p; p = p->next) { -if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) +if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) { ret =0 ; +break ; + } }return ret;
Signature · 1 line
2.35.1.579.g70500b6343.dirty
Re: [PATCH v1 2/3] pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
Quoted 3 lines · Teng Long
There will be only one operant ".bitmap" file in repo, so let's add "break" statement in "
open_pack_bitmap()" when looping all the packs in repo.
Googling "operant" only gives a psychology term X-<.
: behavior (such as bar pressing by a rat to obtain food) that operates on the environment to produce rewarding and reinforcing effects.
I do not quite get this. We expect that there may be more than one .pack in the object store, and that is why we iterate over the packs get_all_packs() gives us. Then for each of these packs, we try to open pack bitmap.
Ah, are you referring to the current limitation that we can have only one .bitmap file, even when there are many packs in the object store?
I wonder if that limitation is something we need to hardcode here.
IOW, when we ever extend the feature to allow more than one, we may have to extend the "struct bitmap_index" so that we can have bitmap info from more than one on-disk files, and the implementation of open_pack_bitmap_1() (which is "read bitmap for this one pack" as opposed to "read bitmpa for this repository" which is this function) would certainly have to change, but would we need to revert this change to the function? If we do not omit this patch, i.e. even if we know there can be only one pack that may have a corresponding bitmap but if the code does not know it and has to try git_open() only to see it fail, how much wasted effort are we talking about?
Quoted 21 lines
Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)pack-bitmap.c +3
-1@@ -484,8 +484,10 @@ static int open_pack_bitmap(struct repository *r,assert (!bitmap_git->map);for (p =get_all_packs (r); p; p = p->next) { -if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) +if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) { ret =0 ; +break ; + } }return ret;
Re: [PATCH v1 2/3] pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
Quoted 11 lines · Junio C Hamano
Quoted 3 lines · Teng Long
There will be only one operant ".bitmap" file in repo, so let's add "break" statement in "
open_pack_bitmap()" when looping all the packs in repo.Googling "operant" only gives a psychology term X-<.
: behavior (such as bar pressing by a rat to obtain food) that operates on the environment to produce rewarding and reinforcing effects.
I suspect that the author meant "operating", which I would probably have written as "Git will only use information from a single .bitmap file in the repository" or similar.
But...
Quoted 10 lines
I do not quite get this. We expect that there may be more than one .pack in the object store, and that is why we iterate over the packs
get_all_packs()gives us. Then for each of these packs, we try to open pack bitmap.Ah, are you referring to the current limitation that we can have only one .bitmap file, even when there are many packs in the object store?
I wonder if that limitation is something we need to hardcode here.
...you're exactly right: if we see more than one pack bitmap in the repository, we can and should warn there, since we'll only ever *use* information from a single bitmap at a time.
That's the behavior before this patch which is guaranteed by t5310.74. This patch causes that test to break, and I would suggest that is telling us to drop this part of the series.
Thanks,
Taylor
Re: [PATCH v1 2/3] pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
In reply to Teng LongQuoted 14 lines · Teng Long
There will be only one operant ".bitmap" file in repo, so let's add "break" statement in "
open_pack_bitmap()" when looping all the packs in repo.Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)pack-bitmap.c +0
-0
pack-bitmap.c static int open_pack_bitmap(struct repository *r, · line 489
Quoted 7 lines
pack-bitmap.c +2
-1@@ -484,8 +484,10 @@ static int open_pack_bitmap(struct repository *r,assert (!bitmap_git->map);for (p =get_all_packs (r); p; p = p->next) { -if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) +if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) { ret =0 ; +break ;
The lack of a break here is intentional, I think, since having more than one bitmap of the same kind in a repository is an error.
(This is behavior we inherited from the pre-MIDX bitmap days, when having more than one pack bitmap caused Git to signal an error, since it could only use the results from a single bitmap).
You can see in pack-bitmap.c::open_pack_bitmap_1() that we have a condition saying:
if (bitmap_git->pack || bitmap_git->midx) {
/* ignore extra bitmap file; we can only handle one */
warning("...")
close(fd;)
return -1;
}We do want to call that open_pack_bitmap_1() function on every pack we know about to make sure that one and only one of them corresponds to a .bitmap.
Thanks,
Taylor
Re: [PATCH v1 2/3] pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
Quoted 2 lines · Taylor Blau
The lack of a break here is intentional, I think, since having more than one bitmap of the same kind in a repository is an error.
Yeah. I quite sure that it's intentional now.
Thanks for your explanations.
Quoted 14 lines
(This is behavior we inherited from the pre-MIDX bitmap days, when having more than one pack bitmap caused Git to signal an error, since it could only use the results from a single bitmap).
You can see in pack-bitmap.c::open_pack_bitmap_1() that we have a condition saying:
if (bitmap_git->pack || bitmap_git->midx) { /* ignore extra bitmap file; we can only handle one */ warning("...") close(fd;) return -1; }
Quoted 3 lines
We do want to call that
open_pack_bitmap_1()function on every pack we know about to make sure that one and only one of them corresponds to a .bitmap.
Yeah, the condition and "warning" make it clear to me which is if already exists a bitmap of the pack or MIDX is ready, we will give warnings and just let it fail (return -1 means a return of NULL in "prepare_bitmap_git()", and will then die() in usage cases I found).
In addition of above, I had a question that why we need "bitmap_git->midx" in the condition? Because here in "open_pack_bitmap_1()" we intent to open the non-midx-bitmap and it's after we deal with "open_midx_bitmap()" in "open_bitmap()":
static int open_bitmap(struct repository *r,
struct bitmap_index *bitmap_git)
{
assert(!bitmap_git->map);
if (!open_midx_bitmap(r, bitmap_git))
return 0;
return open_pack_bitmap(r, bitmap_git);
}So if I understood correct, maybe we can made condition of "bitmap_git->midx" a little earlier so that we can avoid to open every packfile, maybe it's like:
pack-bitmap.c +5 -4
@@ -483,11 +483,12 @@ static int open_pack_bitmap(struct repository *r,assert (!bitmap_git->map); -for (p =get_all_packs (r); p; p = p->next) { -if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) - ret =0 ; +if (!bitmap_git->midx) { +for (p =get_all_packs (r); p; p = p->next) { +if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) + ret =0 ; + } } -return ret; }
Thanks.
Re: [PATCH v1 2/3] pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
Quoted 9 lines · Teng Long
Yeah, the condition and "warning" make it clear to me which is if already exists a bitmap of the pack or MIDX is ready, we will give warnings and just let it fail (return -1 means a return of NULL in "
prepare_bitmap_git()", and will thendie()in usage cases I found).In addition of above, I had a question that why we need "
bitmap_git->midx" in the condition? Because here in "open_pack_bitmap_1()" we intent to open the non-midx-bitmap and it's after we deal with "open_midx_bitmap()" in "open_bitmap()":
You're right; open_pack_bitmap_1() doesn't need to care about whether or not bitmap_git->midx is or isn't non-NULL, since:
- if we did open a MIDX bitmap (which we will always attempt first before trying to open single-pack bitmaps), then we won't even bother to call open_pack_bitmap() at all.
- if we _do_ end up within open_pack_bitmap_1(), then we _know_ that no MIDX bitmap could be found/opened, so there is no need to check in that case, either.
So I think we realistically could do something like:
--- 8< ---
pack-bitmap.c +1 -1
@@ -387,3 +387,3 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git -if (bitmap_git->pack || bitmap_git->midx) { +if (bitmap_git->pack) {/* ignore extra bitmap file; we can only handle one */
--- >8 ---
...but having the conditional there from the pre-image doesn't hurt, either, and it makes the error clearer in case of an accidental regression where we start looking for single-pack bitmaps after successfully opening a multi-pack one.
Quoted 23 lines
static int open_bitmap(struct repository *r,
structbitmap_index*bitmap_git){ assert(!bitmap_git->map); if (!open_midx_bitmap(r, bitmap_git)) return 0; return open_pack_bitmap(r, bitmap_git); }So if I understood correct, maybe we can made condition of "
bitmap_git->midx" a little earlier so that we can avoid to open every packfile, maybe it's like:pack-bitmap.c +0
-2@@ -483,11 +483,12 @@ static int open_pack_bitmap(struct repository *r,assert (!bitmap_git->map); -for (p =get_all_packs (r); p; p = p->next) { -if (open_pack_bitmap_1 (bitmap_git, p) ==0 )
pack-bitmap.c static int open_pack_bitmap(struct repository *r, · line 493
Quoted 10 lines
pack-bitmap.c +5
-2- ret =0 ; +if (!bitmap_git->midx) { +for (p =get_all_packs (r); p; p = p->next) { +if (open_pack_bitmap_1 (bitmap_git, p) ==0 ) + ret =0 ; + } } -return ret; }
This shouldn't be necessary, since we don't bother calling open_pack_bitmap() at all if open_midx_bitmap() returns success. In other words, based on the way that open_bitmap() (which is the caller for both of these functions) is written, we know that once we're in open_pack_bitmap(), that bitmap_git->midx is definitely NULL, which makes this change a noop.
Thanks,
Taylor
Teng Long
Quoted 26 lines · Taylor Blau
You're right;
open_pack_bitmap_1()doesn't need to care about whether or notbitmap_git->midxis or isn't non-NULL, since:- if we did open a MIDX bitmap (which we will always attempt first before trying to open single-pack bitmaps), then we won't even bother to call
open_pack_bitmap()at all.- if we _do_ end up within
open_pack_bitmap_1(), then we _know_ that no MIDX bitmap could be found/opened, so there is no need to check in that case, either.So I think we realistically could do something like:
--- 8< ---
pack-bitmap.c +1
-2@@ -387,3 +387,3 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git -if (bitmap_git->pack || bitmap_git->midx) { +if (bitmap_git->pack) {/* ignore extra bitmap file; we can only handle one */ --- >8 ---
Quoted 4 lines
...but having the conditional there from the pre-image doesn't hurt, either, and it makes the error clearer in case of an accidental regression where we start looking for single-pack bitmaps after successfully opening a multi-pack one.
I agree with the "accidental regression", it's a protection without any disadvantages so far. So, if we don't remove the "bitmap_git->midx" condition for some robust reason, then I think maybe we can let the warning more detailed if the "accident" happens, like:
pack-bitmap.c +19 -3
@@ -333,7 +333,15 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning ("%signoring extra bitmap file: %s" , + bitmap_git->pack ? +xstrfmt ("A non-MIDX bitmap has been opened: %s, " , + bitmap_git->pack->pack_name) : + bitmap_git->midx ? +xstrfmt ("A MIDX bitmap has been opened : %s, " , +midx_bitmap_filename (bitmap_git->midx)) : +"" , + buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -398,9 +406,17 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitreturn -1 ; } -if (bitmap_git->pack) { +if (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning ("%signoring extra bitmap file: %s" , + bitmap_git->pack ? +xstrfmt ("A non-MIDX bitmap has been opened: %s, " , + bitmap_git->pack->pack_name) : + bitmap_git->midx ? +xstrfmt ("A MIDX bitmap has been opened : %s, " , +midx_bitmap_filename (bitmap_git->midx)) : +"" , + packfile->pack_name);close (fd);return -1 ; }
t/t5310-pack-bitmaps.sh +1 -0
@@ -421,6 +421,7 @@ test_expect_success 'complains about multiple pack bitmaps' ' test_line_count = 2 bitmaps && git rev-list --use-bitmap-index HEAD 2>err && + grep "A non-MIDX bitmap has been opened" err && grep "ignoring extra bitmap file" err )
Does that make sense to you?
And also I look codes in "open_midx_bitmap_1()", it's really nice for me to read and understand, but the "char *idx_name = ..." declaration made me a litte confusing at the beginning, I think maybe it's better to name it as "bitmap_name" for it, I know bitmap is also a kind of index structure for using, but when I saw "idx", it brings me to the ".idx" or "multi-pack-index" like files naturally in git context, of course this is not a problem and maybe it's just me.
Re: Teng Long
Quoted 47 lines · Teng Long
Quoted 26 lines · Taylor Blau
You're right;
open_pack_bitmap_1()doesn't need to care about whether or notbitmap_git->midxis or isn't non-NULL, since:- if we did open a MIDX bitmap (which we will always attempt first before trying to open single-pack bitmaps), then we won't even bother to call
open_pack_bitmap()at all.- if we _do_ end up within
open_pack_bitmap_1(), then we _know_ that no MIDX bitmap could be found/opened, so there is no need to check in that case, either.So I think we realistically could do something like:
--- 8< ---
pack-bitmap.c +1
-2@@ -387,3 +387,3 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git -if (bitmap_git->pack || bitmap_git->midx) { +if (bitmap_git->pack) {/* ignore extra bitmap file; we can only handle one */ --- >8 ---Quoted 4 lines
...but having the conditional there from the pre-image doesn't hurt, either, and it makes the error clearer in case of an accidental regression where we start looking for single-pack bitmaps after successfully opening a multi-pack one.
I agree with the "accidental regression", it's a protection without any disadvantages so far. So, if we don't remove the "
bitmap_git->midx" condition for some robust reason, then I think maybe we can let the warning more detailed if the "accident" happens, like:pack-bitmap.c +0
-1@@ -333,7 +333,15 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf);
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 345
Quoted 10 lines
pack-bitmap.c +9
-0+warning ("%signoring extra bitmap file: %s" , + bitmap_git->pack ? +xstrfmt ("A non-MIDX bitmap has been opened: %s, " , + bitmap_git->pack->pack_name) : + bitmap_git->midx ? +xstrfmt ("A MIDX bitmap has been opened : %s, " , +midx_bitmap_filename (bitmap_git->midx)) : +"" , + buf.buf);close (fd);
Aside from any bitmap-specific issues, let's not compose messages like that via concat, it makes them harder to translate.
In this case the original message doesn't have _(), but looking at the context that seems to be a simple omission.
It's more verbose, but it's better to split this out into 3x warning() calls.
It also has the advantage of not leaking memory from the xstrfmt().
PATCH 3/3 bitmap: add trace outputs during open "bitmap" file
In reply to Teng LongIt's supported for a Git repo to use bitmap in both normal bitmap way or a multi-pack-index bitmap.
Sometimes the debug path is not obvious, for example, when executing:
$git rev-list --test-bitmap HEAD fatal: failed to load bitmap indexes
If we see the output like this, we are not sure about what's happened, because the cause should be :
1. neither normal nor midx bitmap exists
2. only midx bitmap exists but core.multipackIndex="false"
3. core.multipackIndex="true" but midx bitmap file is currupt
4. core.multipackIndex="true" and no midx bitmap exists but normal bitmap file is currupt
....
These are some of the scenarios I briefly tested, but maybe there are others (some scenarios is produced manually like "currupt bitmap file", but it's not represent it's an existed bug.).
Therefore, we added some TRACE2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is working on midx bitmap or normal bitmap, or is it simply because the related configuration is disabled. This may help with logging, user troubleshooting, and development debugging.
Here are some of the outputs of
"$GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD"
based on this commit:
1. core.multipackindex="true" and midx bitmap exists
17:12:50.496112 common-main.c:49 | d0 | main | version | | | | | 2.35.1.579.g70500b6343.dirty
17:12:50.496141 common-main.c:50 | d0 | main | start | | 0.000306 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD
17:12:50.496208 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
17:12:50.496333 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list)
17:12:50.496536 midx.c:395 | d0 | main | data | r0 | 0.000706 | 0.000706 | midx | core.multipackIndex:true
17:12:50.496563 midx.c:185 | d0 | main | data | r0 | 0.000733 | 0.000733 | midx | load/num_packs:1
17:12:50.496571 midx.c:186 | d0 | main | data | r0 | 0.000741 | 0.000741 | midx | load/num_objects:15
17:12:50.496677 pack-revindex.c:315 | d0 | main | data | r0 | 0.000847 | 0.000847 | load_midx_re | source:midx
17:12:50.496685 pack-bitmap.c:512 | d0 | main | data | r0 | 0.000856 | 0.000856 | midx | open bitmap (midx):ok
Bitmap v1 test (4 entries loaded)
Found bitmap for 9a0a2928a280d91f5c8ded46e952add4c4296979. 64 bits / f15de25e checksum
17:12:50.497016 progress.c:268 | d0 | main | region_enter | r0 | 0.001186 | | progress | label:Verifying bitmap entries
Verifying bitmap entries: 100% (15/15), done.
17:12:50.497179 progress.c:339 | d0 | main | data | r0 | 0.001350 | 0.000164 | progress | ..total_objects:15
17:12:50.497188 progress.c:346 | d0 | main | region_leave | r0 | 0.001358 | 0.000172 | progress | label:Verifying bitmap entries
OK!
17:12:50.497204 git.c:718 | d0 | main | exit | | 0.001374 | | | code:0
17:12:50.497213 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001384 | | | code:0
2. core.multipackindex="false" and only midx bitmap exists
17:18:38.423328 common-main.c:49 | d0 | main | version | | | | | 2.35.1.579.g70500b6343.dirty
17:18:38.423357 common-main.c:50 | d0 | main | start | | 0.000263 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD
17:18:38.423443 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
17:18:38.423587 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list)
17:18:38.423809 midx.c:395 | d0 | main | data | r0 | 0.000720 | 0.000720 | midx | core.multipackIndex:false
17:18:38.423933 pack-bitmap.c:512 | d0 | main | data | r0 | 0.000845 | 0.000845 | midx | open bitmap (midx):failed
17:18:38.423947 pack-bitmap.c:493 | d0 | main | data | r0 | 0.000858 | 0.000858 | bitmap | open bitmap (non-midx):failed
17:18:38.423956 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
fatal: failed to load bitmap indexes
17:18:38.423967 usage.c:74 | d0 | main | exit | | 0.000879 | | | code:128
17:18:38.423976 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.000888 | | | code:128
3. core.multipackindex="false" and only normal bitmap exists
17:21:25.519233 common-main.c:49 | d0 | main | version | | | | | 2.35.1.579.g70500b6343.dirty
17:21:25.519261 common-main.c:50 | d0 | main | start | | 0.000284 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD
17:21:25.519335 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
17:21:25.519468 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list)
17:21:25.519673 midx.c:395 | d0 | main | data | r0 | 0.000701 | 0.000701 | midx | core.multipackIndex:false
17:21:25.520051 pack-bitmap.c:512 | d0 | main | data | r0 | 0.001080 | 0.001080 | midx | open bitmap (midx):failed
17:21:25.520070 pack-bitmap.c:493 | d0 | main | data | r0 | 0.001099 | 0.001099 | bitmap | open bitmap (non-midx):ok
Bitmap v1 test (4 entries loaded)
Found bitmap for 9a0a2928a280d91f5c8ded46e952add4c4296979. 64 bits / f15de25e checksum
17:21:25.520395 progress.c:268 | d0 | main | region_enter | r0 | 0.001423 | | progress | label:Verifying bitmap entries
Verifying bitmap entries: 100% (15/15), done.
17:21:25.520526 progress.c:339 | d0 | main | data | r0 | 0.001555 | 0.000132 | progress | ..total_objects:15
17:21:25.520533 progress.c:346 | d0 | main | region_leave | r0 | 0.001562 | 0.000139 | progress | label:Verifying bitmap entries
OK!
17:21:25.520549 git.c:718 | d0 | main | exit | | 0.001578 | | | code:0
17:21:25.520557 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001586 | | | code:0
4. core.multipackindex="false" and both midx and normal bitmaps do not exists
17:22:27.680275 common-main.c:49 | d0 | main | version | | | | | 2.35.1.579.g70500b6343.dirty
17:22:27.680298 common-main.c:50 | d0 | main | start | | 0.000284 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD
17:22:27.680375 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
17:22:27.680511 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list)
17:22:27.680718 midx.c:395 | d0 | main | data | r0 | 0.000709 | 0.000709 | midx | core.multipackIndex:false
17:22:27.681085 pack-bitmap.c:512 | d0 | main | data | r0 | 0.001075 | 0.001075 | midx | open bitmap (midx):failed
17:22:27.681102 pack-bitmap.c:493 | d0 | main | data | r0 | 0.001093 | 0.001093 | bitmap | open bitmap (non-midx):failed
17:22:27.681110 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
fatal: failed to load bitmap indexes
17:22:27.681120 usage.c:74 | d0 | main | exit | | 0.001112 | | | code:128
17:22:27.681129 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001120 | | | code:128--- midx.c | 2 ++ pack-bitmap.c | 4 ++++ 2 files changed, 6 insertions(+)
midx.c +2 -0
@@ -392,6 +392,8 @@ int prepare_multi_pack_index_one(struct repository *r, const char *object_dir, istruct multi_pack_index *m_search;prepare_repo_settings (r); +trace2_data_string ("midx" , r,"core.multipackIndex" , + r->settings.core_multi_pack_index ?"true" :"false" );if (!r->settings.core_multi_pack_index)return 0 ;
pack-bitmap.c +4 -0
@@ -490,6 +490,8 @@ static int open_pack_bitmap(struct repository *r, } } +trace2_data_string ("bitmap" , the_repository,"open bitmap (non-midx)" , + ret ?"failed" :"ok" );return ret; }
@@ -507,6 +509,8 @@ static int open_midx_bitmap(struct repository *r,break ; } } +trace2_data_string ("midx" , the_repository,"open bitmap (midx)" , + ret ?"failed" :"ok" );return ret; }
Signature · 1 line
2.35.1.579.g70500b6343.dirty
Re: [PATCH v1 3/3] bitmap: add trace outputs during open "bitmap" file
Quoted 2 lines · Teng Long
It's supported for a Git repo to use bitmap in both normal bitmap way or a multi-pack-index bitmap.
This step out of 3 patches does match what was sold on the cover letter, and it does not seem to even need the other two.
Quoted 11 lines
Signed-off-by Teng Long <[email protected]>--- midx.c | 2 ++ pack-bitmap.c | 4 ++++ 2 files changed, 6 insertions(+)midx.c +0
-0
midx.c int prepare_multi_pack_index_one(struct repository *r, const char *object_dir, i · line 396
Quoted 5 lines
midx.c +2
-0@@ -392,6 +392,8 @@ int prepare_multi_pack_index_one(struct repository *r, const char *object_dir, istruct multi_pack_index *m_search;prepare_repo_settings (r); +trace2_data_string ("midx" , r,"core.multipackIndex" , + r->settings.core_multi_pack_index ?"true" :"false" );
That's an overly deep contination line.
pack-bitmap.c static int open_pack_bitmap(struct repository *r, · line 494
Quoted 13 lines
midx.c +0
-0if (!r->settings.core_multi_pack_index)return 0 ;pack-bitmap.c +2
-0@@ -490,6 +490,8 @@ static int open_pack_bitmap(struct repository *r, } } +trace2_data_string ("bitmap" , the_repository,"open bitmap (non-midx)" , + ret ?"failed" :"ok" );
Likewise.
pack-bitmap.c static int open_midx_bitmap(struct repository *r, · line 513
Quoted 9 lines
pack-bitmap.c +2
-0return ret; }@@ -507,6 +509,8 @@ static int open_midx_bitmap(struct repository *r,break ; } } +trace2_data_string ("midx" , the_repository,"open bitmap (midx)" , + ret ?"failed" :"ok" );
Likewise.
Quoted 2 lines
pack-bitmap.c +0
-0return ret; }
Quoted 33 lines · Teng Long
A Git repo can be chosen to use the normal bitmap (before MIDX) and MIDX bitmap.
I recently tried to understand this part of the MIDX implementation because I found a bug which has been discovered and repaired in community [1].
I am grateful to Taylor Blau for his help and for introducing me to the testing method according to the
git rev-list --test-bitmap <rev>.In the process of understanding and troubleshooting by using this command, I found when the execution is failed it will output a single line of
"fatal: failed to load bitmap indexes", sometimes will be more informations like if the bitmap file is broken, the outputs maybe contain
"error: Corrupted bitmap index file (wrong header)".), but most of time are single line output I mentioned above. So, this brings a little obstacle for debugging and troubleshooting I think, because "failed to load bitmap indexes" can represent to much informations (many causes like: midx config turn off, bitmap inexistence, etc.)Therefore, as a git repo can be chosen to use the normal bitmap (before MIDX) or MIDX bitmap, or they can both exist and let git make the decision. I think why not add some extra informations based on TRACE2 to help for showing the bitmap decision path clearer and more plentiful, so when the failure occurs the user can use it to debug around bitmap in a quicker way.
Thanks.
Links:
1. https://public-inbox.org/git/[email protected]/)Teng Long (3):
pack-bitmap.c: use "ret" in "open_midx_bitmap()"pack-bitmap.c: add "break" statement in "open_pack_bitmap()"
bitmap: add trace outputs during open "bitmap" file
Quoted 8 lines
midx.c | 2 ++ pack-bitmap.c | 17 +++++++++++++---- 2 files changed, 15 insertions(+), 4 deletions(-)Range-diff against v0:
-: ---------- > 1: 3048b4dd29 pack-bitmap.c: use "ret" in "open_midx_bitmap()" -: ---------- > 2: 70500b6343 pack-bitmap.c: add "break" statement in "open_pack_bitmap()" -: ---------- > 3: 9912450fc1 bitmap: add trace outputs during open "bitmap" file
Was there an on-list v0 (RFC?) or is this a range-diff against nothing? Best not to include it until a v2 then.
Comments:
Sometimes it's better to split up patches, but I think these 3x should really be squashed together. We make incremental progress to nowhere in 1/3 and 2/3, and it all comes together in 3/3. The 1-2/3 are trivial enough that we can squash them in.
We then end up with this, with my comments added:
midx.c | 2 ++
pack-bitmap.c | 17 +++++++++++++----
2 files changed, 15 insertions(+), 4 deletions(-)
diff --git a/midx.c b/midx.c
index 865170bad05..fda96440287 100644
--- a/midx.c
+++ b/midx.c
@@ -392,6 +392,8 @@ int prepare_multi_pack_index_one(struct repository *r, const char *object_dir, i
struct multi_pack_index *m_search;
prepare_repo_settings(r);
+ trace2_data_string("midx", r, "core.multipackIndex",
+ r->settings.core_multi_pack_index ? "true" : "false");Weird indentation here.
Also, if we think it's a good idea to log these shouldn't it be in repo_cfg_bool() in repo-settings.c, why is core.multipackIndex out of all in r->settings special?
if (!r->settings.core_multi_pack_index)
return 0;
diff --git a/pack-bitmap.c b/pack-bitmap.c
index 97909d48da3..cac8d4a978f 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -484,25 +484,34 @@ static int open_pack_bitmap(struct repository *r,
assert(!bitmap_git->map);
for (p = get_all_packs(r); p; p = p->next) {
- if (open_pack_bitmap_1(bitmap_git, p) == 0)
+ if (open_pack_bitmap_1(bitmap_git, p) == 0) {Aside: If we end up changing this line anyway, it's OK to just change it to "if (!open...".
ret = 0;
+ break;
+ }
}
+ trace2_data_string("bitmap", the_repository, "open bitmap (non-midx)",
+ ret ? "failed" : "ok");
return ret;
}
static int open_midx_bitmap(struct repository *r,
struct bitmap_index *bitmap_git)
{
+ int ret = -1;
struct multi_pack_index *midx;
assert(!bitmap_git->map);
for (midx = get_multi_pack_index(r); midx; midx = midx->next) {
- if (!open_midx_bitmap_1(bitmap_git, midx))
- return 0;
+ if (!open_midx_bitmap_1(bitmap_git, midx)) {
+ ret = 0;
+ break;
+ }
}
- return -1;
+ trace2_data_string("midx", the_repository, "open bitmap (midx)",
+ ret ? "failed" : "ok");
+ return ret;
}
static int open_bitmap(struct repository *r,It seems odd not to use trace2 regions for this, and to not add add this data logging open_bitmap(). I came up with this on top of this when testing this:
diff --git a/pack-bitmap.c b/pack-bitmap.c
index cac8d4a978f..ba71a7ea5cd 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -318,11 +318,14 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
free(idx_name);
- if (fd < 0)
+ if (fd < 0) {
+ /* TODO: Log trace2_data_string() here, do we care? */
return -1;
+ }
if (fstat(fd, &st)) {
close(fd);
+ /* TODO: Log trace2_data_string() here, do we care? */
return -1;
}
@@ -330,6 +333,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
struct strbuf buf = STRBUF_INIT;
get_midx_filename(&buf, midx->object_dir);
/* ignore extra bitmap file; we can only handle one */
+ /* NOTE: You'll already get a warning (well, "error") event due to this, and it'll be in your region */
warning("ignoring extra bitmap file: %s", buf.buf);
close(fd);
strbuf_release(&buf);
@@ -344,9 +348,11 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
close(fd);
if (load_bitmap_header(bitmap_git) < 0)
+ /* TODO: Add trace2_data_string() or warning/error here? */
goto cleanup;
if (!hasheq(get_midx_checksum(bitmap_git->midx), bitmap_git->checksum))
+ /* TODO: Add trace2_data_string() or warning/error here? */
goto cleanup;
if (load_midx_revindex(bitmap_git->midx) < 0) {
@@ -479,49 +485,44 @@ static int open_pack_bitmap(struct repository *r,
struct bitmap_index *bitmap_git)
{
struct packed_git *p;
- int ret = -1;
-
- assert(!bitmap_git->map);
for (p = get_all_packs(r); p; p = p->next) {
if (open_pack_bitmap_1(bitmap_git, p) == 0) {
- ret = 0;
- break;
+ return 0;
}
}
-
- trace2_data_string("bitmap", the_repository, "open bitmap (non-midx)",
- ret ? "failed" : "ok");
- return ret;
+ return -1;
}
static int open_midx_bitmap(struct repository *r,
struct bitmap_index *bitmap_git)
{
- int ret = -1;
struct multi_pack_index *midx;
- assert(!bitmap_git->map);
-
for (midx = get_multi_pack_index(r); midx; midx = midx->next) {
if (!open_midx_bitmap_1(bitmap_git, midx)) {
- ret = 0;
- break;
+ return 0;
}
}
- trace2_data_string("midx", the_repository, "open bitmap (midx)",
- ret ? "failed" : "ok");
- return ret;
+ return -1;
}
static int open_bitmap(struct repository *r,
struct bitmap_index *bitmap_git)
{
+ int ret;
+
assert(!bitmap_git->map);
- if (!open_midx_bitmap(r, bitmap_git))
- return 0;
- return open_pack_bitmap(r, bitmap_git);
+ trace2_region_enter("pack-bitmap", "open_bitmap", r);
+ if (!open_midx_bitmap(r, bitmap_git)) {
+ ret = 0;
+ goto done;
+ }
+ ret = open_pack_bitmap(r, bitmap_git);
+done:
+ trace2_region_leave("pack-bitmap", "open_bitmap", r);
+ return ret;
}
struct bitmap_index *prepare_bitmap_git(struct repository *r)I.e. surely you just want to create a region, and if you care enough to log failure shouldn't we log that in open_midx_bitmap_1() if we care, and perhaps error() there instead of silently returning -1?
Teng Long
Quoted 3 lines
Was there an on-list v0 (RFC?) or is this a range-diff against nothing? Best not to include it until a v2 then.
My careless, there's no RFC, actually this patch should begin with v0 and without range-diff.
Quoted 4 lines
Sometimes it's better to split up patches, but I think these 3x should really be squashed together. We make incremental progress to nowhere in 1/3 and 2/3, and it all comes together in 3/3. The 1-2/3 are trivial enough that we can squash them in.
Sure.
Quoted 4 lines
+ trace2_data_string("midx", r, "core.multipackIndex", + r->settings.core_multi_pack_index ? "true" : "false");Weird indentation here.
Will fix.
Quoted 2 lines
I.e. surely you just want to create a region, and if you care enough to log failure shouldn't we log that in
open_midx_bitmap_1()if we care,
Actually, I just want to use "trace2_data_string()" at first because which informations I want to append is not so many, or does "create a region" a preferred principle for using TRACE2 APIs?
Quoted 1 line
and perhaps
error()there instead of silently returning -1?
I think so, after I checked function "error_builtin()" and there is a "trace2_cmd_error_va()" usage in it which is already support to print some debug infos using TRACE2 envs.
Thanks.
Re: Teng Long
Quoted 5 lines · Teng Long
Quoted 2 lines
Was there an on-list v0 (RFC?) or is this a range-diff against nothing? Best not to include it until a v2 then.
My careless, there's no RFC, actually this patch should begin with v0 and without range-diff.
No worries.
Quoted 5 lines
Quoted 4 lines
Sometimes it's better to split up patches, but I think these 3x should really be squashed together. We make incremental progress to nowhere in 1/3 and 2/3, and it all comes together in 3/3. The 1-2/3 are trivial enough that we can squash them in.
Quoted 8 lines
Sure.
Quoted 4 lines
+ trace2_data_string("midx", r, "core.multipackIndex", + r->settings.core_multi_pack_index ? "true" : "false");Weird indentation here.
Will fix.
*Nod*
Quoted 6 lines
Quoted 2 lines
I.e. surely you just want to create a region, and if you care enough to log failure shouldn't we log that in
open_midx_bitmap_1()if we care,Actually, I just want to use "
trace2_data_string()" at first because which informations I want to append is not so many, or does "create a region" a preferred principle for using TRACE2 APIs?
I think a begin/end region gives you everything you want here and more, i.e.:
* We'll get start/end times on this (potentially expensive?) operation automatically.
* You just log a "failed", but as the RFC-on-top shows we can just log the specific reason instead, either via error(), warning() or perhaps even trace2_data_string(). because it's within a region it'll be clear from the data what failed.
B.t.w. I noticed after sending that that this ad-hoc patch fails the tests, it passed the one bitmap test I hacked it up against.
It was just there to make the point/demo that it looked easier to do this one function call above where you were adding this.
Quoted 4 lines
Quoted 1 line
and perhaps
error()there instead of silently returning -1?I think so, after I checked function "
error_builtin()" and there is a "trace2_cmd_error_va()" usage in it which is already support to print some debug infos using TRACE2 envs.
Exactly.
v2 cover trace2 output for bitmap decision path
In reply to Teng LongSorry for the late patch. Main diffs from v1:
1. Add a new commit to rename "idx_name" and this may clear up some misconceptions about the naming
2. Add the warnings more detailed when opening bitmap and also let the texts support translation.
3. using error() instead of silently returning -1 when opening bitmap.
4. Continue looping when first MIDX bitmap is found.
Thanks.
Teng Long (5):
pack-bitmap.c: continue looping when first MIDX bitmap is found pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: make warnings more detailed when opening bitmap
bitmap: add trace2 outputs during open "bitmap" file pack-bitmap.c: using error() instead of silently returning -1
pack-bitmap.c | 73 ++++++++++++++++++++++++++++------------- repo-settings.c | 1 + t/t5310-pack-bitmaps.sh | 3 +- 3 files changed, 54 insertions(+), 23 deletions(-)
Range-diff against v1:
1: 3048b4dd29 ! 1: 1bfd2fb6ab pack-bitmap.c: use "ret" in "open_midx_bitmap()"
@@ Metadata
Author: Teng Long <[email protected]>
## Commit message ##
- pack-bitmap.c: use "ret" in "open_midx_bitmap()"
+ pack-bitmap.c: continue looping when first MIDX bitmap is found
- Let's use "ret" value for "return" statement in "open_midx_bitmap()"
- just as the same way as int "open_pack_bitmap()".
+ In "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when
+ the first one has been found, then will break out by a "return"
+ directly.
+ But actually, it's better to don't stop the loop until we have visited
+ both the MIDX in our repository, as well as any alternates (along with
+ _their_ alternates, recursively).
+
+ The discussion link of community:
+
+ https://public-inbox.org/git/[email protected]/
+
+ Helped-by: Taylor Blau <[email protected]>
## pack-bitmap.c ##
@@ pack-bitmap.c: static int open_pack_bitmap(struct repository *r,
assert(!bitmap_git->map);
for (midx = get_multi_pack_index(r); midx; midx = midx->next) {
-- if (!open_midx_bitmap_1(bitmap_git, midx))
+ if (!open_midx_bitmap_1(bitmap_git, midx))
- return 0;
-+ if (!open_midx_bitmap_1(bitmap_git, midx)) {
+ ret = 0;
-+ break;
-+ }
}
- return -1;
+ return ret;
2: 70500b6343 < -: ---------- pack-bitmap.c: add "break" statement in "open_pack_bitmap()" 3: 9912450fc1 < -: ---------- bitmap: add trace outputs during open "bitmap" file -: ---------- > 2: 1fff3b3ca7 pack-bitmap.c: rename "idx_name" to "bitmap_name" -: ---------- > 3: 5a8f5afccf pack-bitmap.c: make warnings more detailed when opening bitmap -: ---------- > 4: 2016ef2e34 bitmap: add trace2 outputs during open "bitmap" file -: ---------- > 5: 1a169d7b5e pack-bitmap.c: using error() instead of silently returning -1
Signature · 1 line
2.35.1.583.g30faa5f068
PATCH v2 1/5 pack-bitmap.c: continue looping when first MIDX bitmap is found
In "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to don't stop the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -494,15 +494,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.1.583.g30faa5f068
Re: [PATCH v2 1/5] pack-bitmap.c: continue looping when first MIDX bitmap is found
Quoted 5 lines · Teng Long
In "
open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.But actually, it's better to don't stop the loop until we have visited
s/don't stop/continue
Quoted 6 lines
both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The discussion link of community:
In the future, it is often worth summarizing the discussion, optionally linking off to the list archive. In this case, I wouldn't mind a little more detail beyond "it's better to [continue] the loop until ...".
In particular, you say it's better without saying why that is the case. The link to <[email protected]> explains, but this commit message as-is leaves out some details that are important IMHO.
Quoted 3 lines
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
The actual changes look good to me.
Thanks,
Taylor
PATCH v2 2/5 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index , let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -313,10 +313,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -368,14 +368,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.1.583.g30faa5f068
Re: [PATCH v2 2/5] pack-bitmap.c: rename "idx_name" to "bitmap_name"
Quoted 2 lines · Teng Long
although bitmap is essentially can be understood as a kind of index , let's define this name a little more accurate here.
The placement of this comma is a little odd, but otherwise the patch looks good to me.
Thanks,
Taylor
PATCH v2 3/5 pack-bitmap.c: make warnings more detailed when opening bitmap
In reply to Teng LongWhen calling the "open_midx_bitmap_1()" or "open_pack_bitmap_1()", there will be a warning if a normal bitmap or MIDX bitmap already has been opened, then let's make the warning information more detailed. For example, it makes the error clearer in case of an accidental regression where we start looking for single-pack bitmaps after successfully opening a multi-pack one.
At the same time, we made the previous and new warning texts support translation.
Discussion in community:
1. https://public-inbox.org/git/[email protected]/#t
2. https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 6 ++++-- t/t5310-pack-bitmaps.sh | 1 + 2 files changed, 5 insertions(+), 2 deletions(-)
pack-bitmap.c +4 -2
@@ -330,7 +330,8 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning (_ ("a normal or midx bitmap already has been opened" )); +warning (_ ("ignoring extra bitmap file: %s" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +388,8 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning (_ ("a normal or midx bitmap already has been opened " )); +warning (_ ("ignoring extra bitmap file: %s" ), packfile->pack_name);close (fd);return -1 ; }
t/t5310-pack-bitmaps.sh +1 -0
@@ -421,6 +421,7 @@ test_expect_success 'complains about multiple pack bitmaps' ' test_line_count = 2 bitmaps && git rev-list --use-bitmap-index HEAD 2>err && + grep "a bitmap has been opened" err && grep "ignoring extra bitmap file" err ) '
Signature · 1 line
2.35.1.583.g30faa5f068
Re: [PATCH v2 3/5] pack-bitmap.c: make warnings more detailed when opening bitmap
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 334
Quoted 11 lines · Teng Long
pack-bitmap.c +2
-1@@ -330,7 +330,8 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning (_ ("a normal or midx bitmap already has been opened" )); +warning (_ ("ignoring extra bitmap file: %s" ), buf.buf);
I'm glad that the existing warning has been marked for translation. There's no reason that it couldn't have been before, so I'm glad to see it added now.
But I'm not sure that the new warning tells the user anything that the old one didn't. The old warning says "ignoring extra bitmap file: ...", leading us to believe that another one has already been opened. The new warning makes the latter part explicit, but I don't think it adds any new information that wasn't already readily available.
t/t5310-pack-bitmaps.sh test_expect_success 'complains about multiple pack bitmaps' ' · line 424
Quoted 9 lines
t/t5310-pack-bitmaps.sh +1
-0@@ -421,6 +421,7 @@ test_expect_success 'complains about multiple pack bitmaps' ' test_line_count = 2 bitmaps && git rev-list --use-bitmap-index HEAD 2>err && + grep "a bitmap has been opened" err &&
This is out of sync with the warning you added, causing t5310.74 to fail.
Thanks,
Taylor
Re: [PATCH v2 3/5] pack-bitmap.c: make warnings more detailed when opening bitmap
Quoted 3 lines · Taylor Blau
I'm glad that the existing warning has been marked for translation. There's no reason that it couldn't have been before, so I'm glad to see it added now.
Thanks and it's actually suggested by Ævar Arnfjörð Bjarmason.
Quoted 8 lines
But I'm not sure that the new warning tells the user anything that the old one didn't. The old warning says "ignoring extra bitmap file: ...", leading us to believe that another one has already been opened. The new warning makes the latter part explicit, but I don't think it adds any new information that wasn't already readily available.
...
This is out of sync with the warning you added, causing t5310.74 to fail.
Yeah, I agree with you about this.
Will roll back on here and fix the broken test in next patch.
Thanks
PATCH v2 4/5 bitmap: add trace2 outputs during open "bitmap" file
In reply to Teng LongIt's supported for a Git repo to use bitmap in both normal bitmap way or a multi-pack-index bitmap.
Sometimes the debug path is not obvious, for example, when executing:
$git rev-list --test-bitmap HEAD
fatal: failed to load bitmap indexes
If we see the output like this, we are not sure about what's happened, because the cause should be :
1. neither normal nor midx bitmap exists
2. only midx bitmap exists but core.multipackIndex="false"
3. core.multipackIndex="true" but midx bitmap file is currupt
4. core.multipackIndex="true" and no midx bitmap exists but normal bitmap file is currupt
....
These are some of the scenarios I briefly tested, but maybe there are others (some scenarios is produced manually like "currupt bitmap file", but it's not represent it's an existed bug.).
Therefore, we added some TRACE2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is working on midx bitmap or normal bitmap, or is it simply because the related configuration is disabled. This may help with logging, user troubleshooting, and development debugging.
Here are some output examples when executing "$GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD" under different situations:
1. _HAVE_ MIDX bitmap and NORMAL bitmap file, but the "core.multipackIndex" configures as "false":
19:38:43.007840 common-main.c:49 | d0 | main | version | | | | | 2.35.1.582.g8e9092487a 19:38:43.007874 common-main.c:50 | d0 | main | start | | 0.000305 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD 19:38:43.007950 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] 19:38:43.008091 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list) 19:38:43.008284 repo-settings.c:11 | d0 | main | data | r0 | 0.000720 | 0.000720 | config | feature.manyfiles:false 19:38:43.008297 repo-settings.c:11 | d0 | main | data | r0 | 0.000734 | 0.000734 | config | feature.experimental:false 19:38:43.008305 repo-settings.c:11 | d0 | main | data | r0 | 0.000742 | 0.000742 | config | core.commitgraph:true 19:38:43.008313 repo-settings.c:11 | d0 | main | data | r0 | 0.000749 | 0.000749 | config | commitgraph.readchangedpaths:true 19:38:43.008320 repo-settings.c:11 | d0 | main | data | r0 | 0.000756 | 0.000756 | config | gc.writecommitgraph:true 19:38:43.008327 repo-settings.c:11 | d0 | main | data | r0 | 0.000764 | 0.000764 | config | fetch.writecommitgraph:false 19:38:43.008334 repo-settings.c:11 | d0 | main | data | r0 | 0.000770 | 0.000770 | config | pack.usesparse:true 19:38:43.008341 repo-settings.c:11 | d0 | main | data | r0 | 0.000777 | 0.000777 | config | core.multipackindex:false 19:38:43.008348 repo-settings.c:11 | d0 | main | data | r0 | 0.000784 | 0.000784 | config | index.sparse:false 19:38:43.008724 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap 19:38:43.008738 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001173 | 0.000013 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap 19:38:43.008754 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001191 | 0.000031 | pack-bitmap | label:open_bitmap Bitmap v1 test (8 entries loaded) Found bitmap for 0a7df7ae92f8d8ed879c240f8ae9cdd33d18085e. 64 bits / 1801edc6 checksum 19:38:43.009099 progress.c:268 | d0 | main | region_enter | r0 | 0.001535 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (27/27), done. 19:38:43.009294 progress.c:339 | d0 | main | data | r0 | 0.001730 | 0.000195 | progress | ..total_objects:27 19:38:43.009302 progress.c:346 | d0 | main | region_leave | r0 | 0.001739 | 0.000204 | progress | label:Verifying bitmap entries OK! 19:38:43.009321 git.c:718 | d0 | main | exit | | 0.001757 | | | code:0 19:38:43.009329 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001766 | | | code:0
2. _HAVE_ MIDX bitmap and NORMAL bitmap file, but the "core.multipackIndex" configures as "true":
(omit duplicate outputs)
...
20:02:31.288797 repo-settings.c:11 | d0 | main | data | r0 | 0.000811 | 0.000811 | config | core.multipackindex:true 20:02:31.288806 repo-settings.c:11 | d0 | main | data | r0 | 0.000819 | 0.000819 | config | index.sparse:false 20:02:31.288836 midx.c:185 | d0 | main | data | r0 | 0.000849 | 0.000849 | midx | load/num_packs:1 20:02:31.288843 midx.c:186 | d0 | main | data | r0 | 0.000857 | 0.000857 | midx | load/num_objects:27 20:02:31.289217 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001229 | | pack-bitmap | label:open_bitmap 20:02:31.289230 pack-bitmap.c:318 | d0 | main | data | r0 | 0.001242 | 0.000013 | midx | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/multi-pack-index-b6b04fbe053bd500d9ca13354466d3249dc275ac.bitmap 20:02:31.289252 pack-revindex.c:315 | d0 | main | data | r0 | 0.001265 | 0.000036 | load_midx_re | ..source:midx 20:02:31.289261 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001274 | 0.000045 | pack-bitmap | label:open_bitmap Bitmap v1 test (8 entries loaded) Found bitmap for 0a7df7ae92f8d8ed879c240f8ae9cdd33d18085e. 64 bits / 1801edc6 checksum 20:02:31.289594 progress.c:268 | d0 | main | region_enter | r0 | 0.001607 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (27/27), done. 20:02:31.289810 progress.c:339 | d0 | main | data | r0 | 0.001823 | 0.000216 | progress | ..total_objects:27 20:02:31.289824 progress.c:346 | d0 | main | region_leave | r0 | 0.001837 | 0.000230 | progress | label:Verifying bitmap entries OK! 20:02:31.289843 git.c:718 | d0 | main | exit | | 0.001856 | | | code:0 20:02:31.289860 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001873 | | | code:0
3. _HAVE_ MIDX bitmap and a corrupt NORMAL bitmap file, but the "core.multipackIndex" configures as "false" :
(omit duplicate outputs)
...
20:14:06.539305 repo-settings.c:11 | d0 | main | data | r0 | 0.000794 | 0.000794 | config | core.multipackindex:false 20:14:06.539310 repo-settings.c:11 | d0 | main | data | r0 | 0.000799 | 0.000799 | config | index.sparse:false 20:14:06.539658 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001146 | | pack-bitmap | label:open_bitmap 20:14:06.539671 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001160 | 0.000014 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap 20:14:06.539686 usage.c:79 | d0 | main | error | | | | | ..Corrupted bitmap index file (wrong header) error: Corrupted bitmap index file (wrong header) 20:14:06.539696 pack-bitmap.c:426 | d0 | main | data | r0 | 0.001185 | 0.000039 | bitmap | ..load bitmap header:failed 20:14:06.539709 usage.c:79 | d0 | main | error | | | | | ..bitmap header is invalid error: bitmap header is invalid 20:14:06.539719 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001208 | 0.000062 | pack-bitmap | label:open_bitmap 20:14:06.539726 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes fatal: failed to load bitmap indexes 20:14:06.539735 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128 20:14:06.539744 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001233 | | | code:128
--- pack-bitmap.c | 40 +++++++++++++++++++++++++++++++++------- repo-settings.c | 1 + t/t5310-pack-bitmaps.sh | 4 ++-- 3 files changed, 36 insertions(+), 9 deletions(-)
pack-bitmap.c +33 -7
@@ -312,9 +312,12 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx) { +int fd;struct stat st;char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"try to open bitmap" , + bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);
@@ -322,6 +325,8 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,return -1 ;if (fstat (fd, &st)) { +trace2_data_string ("midx" , the_repository,"stat bitmap file" , +"failed" );close (fd);return -1 ; }
@@ -344,12 +349,19 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, MAP_PRIVATE, fd, 0);close (fd); -if (load_bitmap_header (bitmap_git) <0 ) +if (load_bitmap_header (bitmap_git) <0 ) { +trace2_data_string ("midx" , the_repository,"load bitmap header" , +"failed" ); goto cleanup; + } -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +trace2_data_string ("midx" , the_repository,"verify checksum" , +"mismatch" ); goto cleanup; + } +if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" )); goto cleanup;
@@ -375,6 +387,8 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitreturn -1 ; bitmap_name =pack_bitmap_filename (packfile); +trace2_data_string ("bitmap" , the_repository,"try to open bitmap" , + bitmap_name); fd =git_open (bitmap_name);free (bitmap_name);
@@ -382,6 +396,8 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitreturn -1 ;if (fstat (fd, &st)) { +trace2_data_string ("bitmap" , the_repository,"stat bitmap file" , +"failed" );close (fd);return -1 ; }
@@ -395,6 +411,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git }if (!is_pack_valid (packfile)) { +trace2_data_string ("bitmap" , the_repository,"packfile" ,"invalid" );close (fd);return -1 ; }
@@ -406,6 +423,8 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitclose (fd);if (load_bitmap_header (bitmap_git) <0 ) { +trace2_data_string ("bitmap" , the_repository, +"load bitmap header" ,"failed" );munmap (bitmap_git->map, bitmap_git->map_size); bitmap_git->map =NULL ; bitmap_git->map_size =0 ;
@@ -511,11 +530,18 @@ static int open_midx_bitmap(struct repository *r,static int open_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { -assert (!bitmap_git->map); +int ret =-1 ; -if (!open_midx_bitmap (r, bitmap_git)) -return 0 ; -return open_pack_bitmap (r, bitmap_git); +assert (!bitmap_git->map); +trace2_region_enter ("pack-bitmap" ,"open_bitmap" , r); +if (!open_midx_bitmap (r, bitmap_git)) { + ret =0 ; + goto done; + } + ret =open_pack_bitmap (r, bitmap_git); +done: +trace2_region_leave ("pack-bitmap" ,"open_bitmap" , r); +return ret; }struct bitmap_index *prepare_bitmap_git (struct repository *r)
repo-settings.c +1 -0
@@ -8,6 +8,7 @@ static void repo_cfg_bool(struct repository *r, const char *key, int *dest, {if (repo_config_get_bool (r, key, dest)) *dest = def; +trace2_data_string ("config" , r, key, *dest ?"true" :"false" ); }void prepare_repo_settings (struct repository *r)
t/t5310-pack-bitmaps.sh +2 -2
@@ -421,8 +421,8 @@ test_expect_success 'complains about multiple pack bitmaps' ' test_line_count = 2 bitmaps && git rev-list --use-bitmap-index HEAD 2>err && - grep "a bitmap has been opened" err && - grep "ignoring extra bitmap file" err + grep "warning: a normal or midx bitmap already has been opened" err && + grep "warning: ignoring extra bitmap file" err ) '
Signature · 1 line
2.35.1.583.g30faa5f068
Re: [PATCH v2 4/5] bitmap: add trace2 outputs during open "bitmap" file
On Thu, Apr 21 2022, Teng Long wrote:
Thanks for following up..
Quoted 2 lines
19:38:43.007840 common-main.c:49 | d0 | main | version | | | | | 2.35.1.582.g8e9092487a 19:38:43.007874 common-main.c:50 | d0 | main | start | | 0.000305 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD
It's really helpful to have these full examples in the commit message. Thanks.
Quoted 5 lines
pack-bitmap.c +0
-0@@ -312,9 +312,12 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx)
pack-bitmap.c char *pack_bitmap_filename(struct packed_git *p) · line 322
Quoted 10 lines
pack-bitmap.c +4
-1{ +int fd;struct stat st;char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"try to open bitmap" , + bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);
Hrm, so re my comment on 5/5 are you trying to use the "try to open" as a timer to see what our start time is?
I think probably not, in which case this whole variable flip-around is something we won't need.
But if we do actually need it perhaps a sub-region for the timing?
Quoted 4 lines
pack-bitmap.c +0
-0@@ -511,11 +530,18 @@ static int open_midx_bitmap(struct repository *r,static int open_bitmap (struct repository *r,struct bitmap_index *bitmap_git) {
pack-bitmap.c static int open_midx_bitmap(struct repository *r, · line 538
Quoted 10 lines
pack-bitmap.c +5
-4-assert (!bitmap_git->map); +int ret =-1 ; -if (!open_midx_bitmap (r, bitmap_git)) -return 0 ; -return open_pack_bitmap (r, bitmap_git); +assert (!bitmap_git->map); +trace2_region_enter ("pack-bitmap" ,"open_bitmap" , r); +if (!open_midx_bitmap (r, bitmap_git)) { + ret =0 ;
Nit: I think these "goto" patterns are best if your "int ret = N" picks an "N" which is the default that you'll "goto", i.e. if you pick "ret = 0" you'll just need "goto done" here....
pack-bitmap.c static int open_midx_bitmap(struct repository *r, · line 541
Quoted 3 lines
pack-bitmap.c +3
-0+ goto done; + } + ret =open_pack_bitmap (r, bitmap_git);
...which we may then override here.
Just saves you the assignment and the {}, but it tends to add up in longer functions.
pack-bitmap.c static int open_midx_bitmap(struct repository *r, · line 545
Quoted 4 lines
pack-bitmap.c +3
-0+done: +trace2_region_leave ("pack-bitmap" ,"open_bitmap" , r); +return ret; }
Looks good, aside from the 5/5 comments that much of the data string logging looks like it becomes redundant in the end due to error() giving us the same thing.
Quoted 18 lines
pack-bitmap.c +0
-0struct bitmap_index *prepare_bitmap_git (struct repository *r)repo-settings.c +1
-0@@ -8,6 +8,7 @@ static void repo_cfg_bool(struct repository *r, const char *key, int *dest, {if (repo_config_get_bool (r, key, dest)) *dest = def; +trace2_data_string ("config" , r, key, *dest ?"true" :"false" ); }void prepare_repo_settings (struct repository *r)t/t5310-pack-bitmaps.sh +0
-0
t/t5310-pack-bitmaps.sh test_expect_success 'complains about multiple pack bitmaps' ' · line 427
Quoted 9 lines
t/t5310-pack-bitmaps.sh +2
-2@@ -421,8 +421,8 @@ test_expect_success 'complains about multiple pack bitmaps' ' test_line_count = 2 bitmaps && git rev-list --use-bitmap-index HEAD 2>err && - grep "a bitmap has been opened" err && - grep "ignoring extra bitmap file" err + grep "warning: a normal or midx bitmap already has been opened" err && + grep "warning: ignoring extra bitmap file" err ) '
I haven't tested but part of this test change looks like it'll break under bisect, i.e. you changed 1/2 of these strings in 3/5. Did you try with "git rebase -i -x 'make test T=t*bitmap*" or similar?
Re: [PATCH v2 4/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 2 lines · Ævar Arnfjörð Bjarmason
It's really helpful to have these full examples in the commit message. Thanks.
From the previous contribution process, I think it is necessary to provide sufficient information in patch, which can help to save reviewer's time and find problems faster.
Quoted 7 lines
Hrm, so re my comment on 5/5 are you trying to use the "try to open" as a timer to see what our start time is?
I think probably not, in which case this whole variable flip-around is something we won't need.
But if we do actually need it perhaps a sub-region for the timing?
Yes, I looked back on it and found it's redundant and unnecessary now. Will delete the two related references in the code in next patch.
Quoted 4 lines
Nit: I think these "goto" patterns are best if your "int ret = N" picks an "N" which is the default that you'll "goto", i.e. if you pick "ret = 0" you'll just need "goto done" here....
Quoted 8 lines
pack-bitmap.c static int open_midx_bitmap(struct repository *r, · line 541
Quoted 3 lines
pack-bitmap.c +3
-0+ goto done; + } + ret =open_pack_bitmap (r, bitmap_git);...which we may then override here.
Just saves you the assignment and the {}, but it tends to add up in longer functions
Make sense.
Quoted 8 lines
pack-bitmap.c static int open_midx_bitmap(struct repository *r, · line 545
Quoted 4 lines
pack-bitmap.c +3
-0+done: +trace2_region_leave ("pack-bitmap" ,"open_bitmap" , r); +return ret; }Looks good, aside from the 5/5 comments that much of the data string logging looks like it becomes redundant in the end due to
error()giving us the same thing.
Oops, I will read the comments and reply them in 5/5.
Quoted 3 lines
I haven't tested but part of this test change looks like it'll break under bisect, i.e. you changed 1/2 of these strings in 3/5. Did you try with "git rebase -i -x 'make test T=t*bitmap*" or similar?
Yes, they should be in the same commit in 3/5, but now I the new warning text seems like is much verbose which suggested by Taylor Blau, and I will roll back the new warning message and fix the broken test under bisect.
Thanks.
Re: [PATCH v2 4/5] bitmap: add trace2 outputs during open "bitmap" file
In reply to Ævar Arnfjörð BjarmasonQuoted 7 lines
Hrm, so re my comment on 5/5 are you trying to use the "try to open" as a timer to see what our start time is?
I think probably not, in which case this whole variable flip-around is something we won't need.
But if we do actually need it perhaps a sub-region for the timing?
By the way, I think it will be better replace "try to open" with the bitmap name so we can also know about what kind of bitmap we are opening by filename.
Before:
19:38:43.008724 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap
19:38:43.008738 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001173 | 0.000013 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap
19:38:43.008754 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001191 | 0.000031 | pack-bitmap | label:open_bitmapAfter:
19:38:43.008724 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap
19:38:43.008738 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001173 | 0.000013 | bitmap | ../home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap
19:38:43.008754 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001191 | 0.000031 | pack-bitmap | label:open_bitmapThanks.
Re: [PATCH v2 4/5] bitmap: add trace2 outputs during open "bitmap" file
In reply to Teng LongQuoted 116 lines · Teng Long
It's supported for a Git repo to use bitmap in both normal bitmap way or a multi-pack-index bitmap.
Sometimes the debug path is not obvious, for example, when executing:
$git rev-list --test-bitmap HEAD
fatal: failed to load bitmap indexesIf we see the output like this, we are not sure about what's happened, because the cause should be :
1. neither normal nor midx bitmap exists
2. only midx bitmap exists but core.multipackIndex="false"
3. core.multipackIndex="true" but midx bitmap file is currupt
4. core.multipackIndex="true" and no midx bitmap exists but normal bitmap file is currupt
....These are some of the scenarios I briefly tested, but maybe there are others (some scenarios is produced manually like "currupt bitmap file", but it's not represent it's an existed bug.).
Therefore, we added some TRACE2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is working on midx bitmap or normal bitmap, or is it simply because the related configuration is disabled. This may help with logging, user troubleshooting, and development debugging.
Here are some output examples when executing "$GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD" under different situations:
1. _HAVE_ MIDX bitmap and NORMAL bitmap file, but the "core.multipackIndex" configures as "false":
19:38:43.007840 common-main.c:49 | d0 | main | version | | | | | 2.35.1.582.g8e9092487a 19:38:43.007874 common-main.c:50 | d0 | main | start | | 0.000305 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD 19:38:43.007950 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] 19:38:43.008091 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list) 19:38:43.008284 repo-settings.c:11 | d0 | main | data | r0 | 0.000720 | 0.000720 | config | feature.manyfiles:false 19:38:43.008297 repo-settings.c:11 | d0 | main | data | r0 | 0.000734 | 0.000734 | config | feature.experimental:false 19:38:43.008305 repo-settings.c:11 | d0 | main | data | r0 | 0.000742 | 0.000742 | config | core.commitgraph:true 19:38:43.008313 repo-settings.c:11 | d0 | main | data | r0 | 0.000749 | 0.000749 | config | commitgraph.readchangedpaths:true 19:38:43.008320 repo-settings.c:11 | d0 | main | data | r0 | 0.000756 | 0.000756 | config | gc.writecommitgraph:true 19:38:43.008327 repo-settings.c:11 | d0 | main | data | r0 | 0.000764 | 0.000764 | config | fetch.writecommitgraph:false 19:38:43.008334 repo-settings.c:11 | d0 | main | data | r0 | 0.000770 | 0.000770 | config | pack.usesparse:true 19:38:43.008341 repo-settings.c:11 | d0 | main | data | r0 | 0.000777 | 0.000777 | config | core.multipackindex:false 19:38:43.008348 repo-settings.c:11 | d0 | main | data | r0 | 0.000784 | 0.000784 | config | index.sparse:false 19:38:43.008724 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap 19:38:43.008738 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001173 | 0.000013 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap 19:38:43.008754 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001191 | 0.000031 | pack-bitmap | label:open_bitmap Bitmap v1 test (8 entries loaded) Found bitmap for 0a7df7ae92f8d8ed879c240f8ae9cdd33d18085e. 64 bits / 1801edc6 checksum 19:38:43.009099 progress.c:268 | d0 | main | region_enter | r0 | 0.001535 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (27/27), done. 19:38:43.009294 progress.c:339 | d0 | main | data | r0 | 0.001730 | 0.000195 | progress | ..total_objects:27 19:38:43.009302 progress.c:346 | d0 | main | region_leave | r0 | 0.001739 | 0.000204 | progress | label:Verifying bitmap entries OK! 19:38:43.009321 git.c:718 | d0 | main | exit | | 0.001757 | | | code:0 19:38:43.009329 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001766 | | | code:02. _HAVE_ MIDX bitmap and NORMAL bitmap file, but the "core.multipackIndex" configures as "true":
(omit duplicate outputs) ... 20:02:31.288797 repo-settings.c:11 | d0 | main | data | r0 | 0.000811 | 0.000811 | config | core.multipackindex:true 20:02:31.288806 repo-settings.c:11 | d0 | main | data | r0 | 0.000819 | 0.000819 | config | index.sparse:false 20:02:31.288836 midx.c:185 | d0 | main | data | r0 | 0.000849 | 0.000849 | midx | load/num_packs:1 20:02:31.288843 midx.c:186 | d0 | main | data | r0 | 0.000857 | 0.000857 | midx | load/num_objects:27 20:02:31.289217 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001229 | | pack-bitmap | label:open_bitmap 20:02:31.289230 pack-bitmap.c:318 | d0 | main | data | r0 | 0.001242 | 0.000013 | midx | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/multi-pack-index-b6b04fbe053bd500d9ca13354466d3249dc275ac.bitmap 20:02:31.289252 pack-revindex.c:315 | d0 | main | data | r0 | 0.001265 | 0.000036 | load_midx_re | ..source:midx 20:02:31.289261 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001274 | 0.000045 | pack-bitmap | label:open_bitmap Bitmap v1 test (8 entries loaded) Found bitmap for 0a7df7ae92f8d8ed879c240f8ae9cdd33d18085e. 64 bits / 1801edc6 checksum 20:02:31.289594 progress.c:268 | d0 | main | region_enter | r0 | 0.001607 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (27/27), done. 20:02:31.289810 progress.c:339 | d0 | main | data | r0 | 0.001823 | 0.000216 | progress | ..total_objects:27 20:02:31.289824 progress.c:346 | d0 | main | region_leave | r0 | 0.001837 | 0.000230 | progress | label:Verifying bitmap entries OK! 20:02:31.289843 git.c:718 | d0 | main | exit | | 0.001856 | | | code:0 20:02:31.289860 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001873 | | | code:03. _HAVE_ MIDX bitmap and a corrupt NORMAL bitmap file, but the "core.multipackIndex" configures as "false" :
(omit duplicate outputs) ... 20:14:06.539305 repo-settings.c:11 | d0 | main | data | r0 | 0.000794 | 0.000794 | config | core.multipackindex:false 20:14:06.539310 repo-settings.c:11 | d0 | main | data | r0 | 0.000799 | 0.000799 | config | index.sparse:false 20:14:06.539658 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001146 | | pack-bitmap | label:open_bitmap 20:14:06.539671 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001160 | 0.000014 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap 20:14:06.539686 usage.c:79 | d0 | main | error | | | | | ..Corrupted bitmap index file (wrong header) error: Corrupted bitmap index file (wrong header) 20:14:06.539696 pack-bitmap.c:426 | d0 | main | data | r0 | 0.001185 | 0.000039 | bitmap | ..load bitmap header:failed 20:14:06.539709 usage.c:79 | d0 | main | error | | | | | ..bitmap header is invalid error: bitmap header is invalid 20:14:06.539719 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001208 | 0.000062 | pack-bitmap | label:open_bitmap 20:14:06.539726 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes fatal: failed to load bitmap indexes 20:14:06.539735 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128 20:14:06.539744 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001233 | | | code:128Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 40 +++++++++++++++++++++++++++++++++------- repo-settings.c | 1 + t/t5310-pack-bitmaps.sh | 4 ++-- 3 files changed, 36 insertions(+), 9 deletions(-)pack-bitmap.c +0
-0
pack-bitmap.c char *pack_bitmap_filename(struct packed_git *p) · line 320
Quoted 10 lines
pack-bitmap.c +4
-1@@ -312,9 +312,12 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx) { +int fd;struct stat st;char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"try to open bitmap" , + bitmap_name); + fd =git_open (bitmap_name);
This might just be a style thing, but rather than logging the pathname in a separate data_string message, you can use the _printf version of region_enter and region_leave to also print the name of the path -- like I did in read-cache.c for the "do_read_index" calls.
... | region_enter | ... | index | label:do_read_index .git/index
...
... | region_leave | ... | index | label:do_read_index .git/index
As AEvar suggests in another message in this thread, I'm not sure if you need the region timing here for reading the bitmap, but all of the error and any other data messages will be bounded between the region_enter and region_leave events and that might (or might not) be helpful.
Also, I agree with AEvar's statements about using error() and getting the trace2 error messages for free and not needing some of the trace2_data_string() messages that you have later in this file.
I wonder if it would be worth adding the pathname of the invalid file to those new error messages. Granted you'll have it in the trace2 log, but then you'll also get it on stderr if that would be helpful.
Jeff
Re: [PATCH v2 4/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 9 lines
This might just be a style thing, but rather than logging the pathname in a separate
data_stringmessage, you can use the_printfversion ofregion_enterandregion_leaveto also print the name of the path -- like I did inread-cache.cfor the "do_read_index" calls.... |
region_enter| ... | index | label:do_read_index .git/index
...
... |region_leave| ... | index | label:do_read_index .git/index
Appreciate for the input about the _printf version, we can choose to let the region_enter and region_leave to print the pathname by moving the related "midx_bitmap_filename()" and "pack_bitmap_filename" at front, but it's not enough because both midx and normal bitmap support multiple opening, so it's likely we keep on the current way using "trace2_data_string()" in "open_pack_bitmap_1()" and "open_midx_bitmap_1()" is a simpler solution.
I'm not sure If I totally get the meaning about your suggestion, so correct me if I understand you wrong.
Quoted 5 lines
As AEvar suggests in another message in this thread, I'm not sure if you need the region timing here for reading the bitmap, but all of the error and any other data messages will be bounded between the
region_enterandregion_leaveevents and that might (or might not) be helpful.
I think it's needed in my opinion, the bounded between the region is helpful, especially when we want to know the detailed debug info like we do in "open_midx_bitmap_1()".
Quoted 8 lines
Also, I agree with AEvar's statements about using
error()and getting the trace2 error messages for free and not needing some of thetrace2_data_string()messages that you have later in this file.I wonder if it would be worth adding the pathname of the invalid file to those new error messages. Granted you'll have it in the trace2 log, but then you'll also get it on stderr if that would be helpful.
I think I will remove the redundant "trace2_data_string()" code when it will return by "error()".
Thanks.
Re: [PATCH v2 4/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 18 lines · Teng Long
Quoted 9 lines
This might just be a style thing, but rather than logging the pathname in a separate
data_stringmessage, you can use the_printfversion ofregion_enterandregion_leaveto also print the name of the path -- like I did inread-cache.cfor the "do_read_index" calls.... |
region_enter| ... | index | label:do_read_index .git/index
...
... |region_leave| ... | index | label:do_read_index .git/indexAppreciate for the input about the
_printfversion, we can choose to let theregion_enterandregion_leaveto print the pathname by moving the related "midx_bitmap_filename()" and "pack_bitmap_filename" at front, but it's not enough because both midx and normal bitmap support multiple opening, so it's likely we keep on the current way using "trace2_data_string()" in "open_pack_bitmap_1()" and "open_midx_bitmap_1()" is a simpler solution.
Quoted 2 lines
I'm not sure If I totally get the meaning about your suggestion, so correct me if I understand you wrong.
That's fine. I was assuming that you were only operating on a single file at a time and opening, using, closing it in a nicely bracketed fashion. The region_enter and _leave is good for capturing that. But if you might be handling both files concurrently, then your messages would be better since we don't want mix or improperly nest the regions.
(I must admit that I haven't studied the bitmap code, so I defer to your judgement here.)
Quoted 27 lines
Quoted 6 lines
As AEvar suggests in another message in this thread, I'm not sure if you need the region timing here for reading the bitmap, but all of the error and any other data messages will be bounded between the
region_enterandregion_leaveevents and that might (or might not) be helpful.I think it's needed in my opinion, the bounded between the region is helpful, especially when we want to know the detailed debug info like we do in "
open_midx_bitmap_1()".Quoted 8 lines
Also, I agree with AEvar's statements about using
error()and getting the trace2 error messages for free and not needing some of thetrace2_data_string()messages that you have later in this file.I wonder if it would be worth adding the pathname of the invalid file to those new error messages. Granted you'll have it in the trace2 log, but then you'll also get it on stderr if that would be helpful.
I think I will remove the redundant "
trace2_data_string()" code when it will return by "error()".Thanks.
PATCH v2 5/5 pack-bitmap.c: using error() instead of silently returning -1
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
pack-bitmap.c +5 -5
@@ -328,7 +328,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,trace2_data_string ("midx" , the_repository,"stat bitmap file" ,"failed" );close (fd); -return -1 ; +return error ("cannot stat bitmap file" ); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -374,7 +374,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error ("cannot open midx bitmap file" ); }static int open_pack_bitmap_1 (struct bitmap_index *bitmap_git,struct packed_git *packfile)
@@ -399,7 +399,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gittrace2_data_string ("bitmap" , the_repository,"stat bitmap file" ,"failed" );close (fd); -return -1 ; +return error ("cannot stat bitmap file" ); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -413,7 +413,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {trace2_data_string ("bitmap" , the_repository,"packfile" ,"invalid" );close (fd); -return -1 ; +return error ("packfile is invalid" ); } bitmap_git->pack = packfile;
@@ -430,7 +430,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_git->map_size =0 ; bitmap_git->map_pos =0 ; bitmap_git->pack =NULL ; -return -1 ; +return error ("bitmap header is invalid" ); }return 0 ;
Signature · 1 line
2.35.1.583.g30faa5f068
Re: [PATCH v2 5/5] pack-bitmap.c: using error() instead of silently returning -1
Quoted 20 lines · Teng Long
In "
open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to returnerror()instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)pack-bitmap.c +0
-0
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 332
Quoted 6 lines
pack-bitmap.c +1
-1@@ -328,7 +328,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,trace2_data_string ("midx" , the_repository,"stat bitmap file" ,"failed" );close (fd); -return -1 ; +return error ("cannot stat bitmap file" ); }
First, I wondered if we were missing _(), but looking at other string in the file they're not using that already, looks like these all should, but we can fix this all up some other time in some i18n commit. It's fine to keep this for now.
But more importantly: I think this should be your 4/5. I.e. just make these an error() and you won't need to add e.g. this trace2_data_string() for a failed stat.
You will be inside your trace2 region, so any failure to stat etc. will be obvious from the structure of the data and the "error" event, no reason to have an additional trace2_data_string().
Aside from that & as a general matter: Unless you have some use-case for trace2 data in this detail I'd think that it would be better just to skip logging it (except if we get it for free, such as with error()).
I.e. is this particular callsite really different from other places we fail to stat() or open() something?
It's all a moot point with the region + error, but just something to keep in mind.
Re: [PATCH v2 5/5] pack-bitmap.c: using error() instead of silently returning -1
Quoted 4 lines · Ævar Arnfjörð Bjarmason
First, I wondered if we were missing
_(), but looking at other string in the file they're not using that already, looks like these all should, but we can fix this all up some other time in some i18n commit. It's fine to keep this for now.
Yes, I agree with you.
I also think I have a willingness to make another patchset to solve the _() missing problems recently.
Quoted 3 lines
But more importantly: I think this should be your 4/5. I.e. just make these an
error()and you won't need to add e.g. thistrace2_data_string()for a failed stat.
Make sense. Will adjust the order in next path.
Quoted 3 lines
You will be inside your trace2 region, so any failure to stat etc. will be obvious from the structure of the data and the "error" event, no reason to have an additional
trace2_data_string().
Yeah, I forgot about the "error()" already load the trace2 functions in. Will remove the redundant trace2_data_string() where it's obviously will return error().
Quoted 9 lines
Aside from that & as a general matter: Unless you have some use-case for trace2 data in this detail I'd think that it would be better just to skip logging it (except if we get it for free, such as with
error()).I.e. is this particular callsite really different from other places we fail to
stat()oropen()something?It's all a moot point with the region + error, but just something to keep in mind.
Make sense.
And I think it's a good case in "open_midx_bitmap_1()" to add related trace2_data_string() because there only a general error info in "cleanup:".
Thanks.
v3 cover [PATCH v3 0/5]
In reply to Teng LongSorry for the late reply (in emergency project).
I reread the replies in patch v2 and fixed/optimized in patch v3 in respective, I hope I haven't missed anything.
Thanks.
Teng Long (5):
pack-bitmap.c: continue looping when first MIDX bitmap is found pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: make warnings support i18N when opening bitmap pack-bitmap.c: using error() instead of silently returning -1
bitmap: add trace2 outputs during open "bitmap" file
pack-bitmap.c | 58 +++++++++++++++++++++++++++++++------------------ repo-settings.c | 1 + 2 files changed, 38 insertions(+), 21 deletions(-)
Range-diff against v2:
1: 1bfd2fb6ab ! 1: 589e3f4075 pack-bitmap.c: continue looping when first MIDX bitmap is found
@@ Commit message
the first one has been found, then will break out by a "return" directly.
- But actually, it's better to don't stop the loop until we have visited
+ But actually, it's better to continue the loop until we have visited
both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
+ The reason for this is, there may exist more than one MIDX file in
+ a repo. The "multi_pack_index" struct is actually designed as a singly
+ linked list, and if a MIDX file has been already opened successfully,
+ then the other MIDX files will be skipped and left with a warning
+ "ignoring extra bitmap file." to the output.
+
The discussion link of community:
https://public-inbox.org/git/[email protected]/
2: 1fff3b3ca7 ! 2: b6b30047fc pack-bitmap.c: rename "idx_name" to "bitmap_name"
@@ Commit message
There may bring some confusion in this "idx_name" naming, which
might lead us to think of ".idx "or" multi-pack-index" files, - although bitmap is essentially can be understood as a kind of index
- , let's define this name a little more accurate here.
+ although bitmap is essentially can be understood as a kind of index,
+ let's define this name a little more accurate here.
3: 5a8f5afccf ! 3: d8dfe53dd4 pack-bitmap.c: make warnings more detailed when opening bitmap
@@ Metadata
Author: Teng Long <[email protected]>
## Commit message ##
- pack-bitmap.c: make warnings more detailed when opening bitmap
+ pack-bitmap.c: make warnings support i18N when opening bitmap
When calling the "open_midx_bitmap_1()" or "open_pack_bitmap_1()", there will be a warning if a normal bitmap or MIDX bitmap already has been
- opened, then let's make the warning information more detailed. For
- example, it makes the error clearer in case of an accidental
- regression where we start looking for single-pack bitmaps after
- successfully opening a multi-pack one.
-
- At the same time, we made the previous and new warning texts support
- translation.
+ opened, then let's make the warning text supporting for translation.
Discussion in community:
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
get_midx_filename(&buf, midx->object_dir);
/* ignore extra bitmap file; we can only handle one */
- warning("ignoring extra bitmap file: %s", buf.buf);
-+ warning(_("a normal or midx bitmap already has been opened"));
+ warning(_("ignoring extra bitmap file: %s"), buf.buf);
close(fd);
strbuf_release(&buf);
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, st
if (bitmap_git->pack || bitmap_git->midx) {
/* ignore extra bitmap file; we can only handle one */
- warning("ignoring extra bitmap file: %s", packfile->pack_name);
-+ warning(_("a normal or midx bitmap already has been opened "));
+ warning(_("ignoring extra bitmap file: %s"), packfile->pack_name);
close(fd);
return -1;
}
-
- ## t/t5310-pack-bitmaps.sh ##
-@@ t/t5310-pack-bitmaps.sh: test_expect_success 'complains about multiple pack bitmaps' '
- test_line_count = 2 bitmaps &&
-
- git rev-list --use-bitmap-index HEAD 2>err &&
-+ grep "a bitmap has been opened" err &&
- grep "ignoring extra bitmap file" err
- )
- '
5: 1a169d7b5e ! 4: 72da3b5844 pack-bitmap.c: using error() instead of silently returning -1
@@ Commit message
## pack-bitmap.c ##
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
- trace2_data_string("midx", the_repository, "stat bitmap file",
- "failed");
+
+ if (fstat(fd, &st)) {
close(fd);
- return -1;
-+ return error("cannot stat bitmap file");
++ return error(_("cannot stat bitmap file"));
}
if (bitmap_git->pack || bitmap_git->midx) {
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git *packfile)
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
- trace2_data_string("bitmap", the_repository, "stat bitmap file",
- "failed");
+
+ if (fstat(fd, &st)) {
close(fd);
- return -1;
-+ return error("cannot stat bitmap file");
++ return error(_("cannot stat bitmap file"));
}
if (bitmap_git->pack || bitmap_git->midx) {
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
+
if (!is_pack_valid(packfile)) {
- trace2_data_string("bitmap", the_repository, "packfile", "invalid");
close(fd);
- return -1;
-+ return error("packfile is invalid");
++ return error(_("packfile is invalid"));
}
bitmap_git->pack = packfile;
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, st
bitmap_git->map_pos = 0;
bitmap_git->pack = NULL;
- return -1;
-+ return error("bitmap header is invalid");
++ return error(_("bitmap header is invalid"));
}
return 0;4: 2016ef2e34 ! 5: e118758d1d bitmap: add trace2 outputs during open "bitmap" file
@@ Metadata
## Commit message ##
bitmap: add trace2 outputs during open "bitmap" file
- It's supported for a Git repo to use bitmap in both normal bitmap way
- or a multi-pack-index bitmap.
-
- Sometimes the debug path is not obvious, for example, when executing:
+ It's supported for a repo to use bitmap in both "NORMAL" bitmap way
+ or a MIDX (multi-pack-index) bitmap. Either of two bitmap kinds can
+ exist in the repository, or both can be stored but let the config
+ controls which kind of bitmap is used (like "core.multipackIndex",
+ etc.). Because of this, sometimes the bitmap debug path is not
+ obvious enough, for example, when executing:
$git rev-list --test-bitmap HEAD
fatal: failed to load bitmap indexes
- If we see the output like this, we are not sure about what's happened,
- because the cause should be :
+ If we see the output like this, It's not sure for us to know
+ what's happened concretely, because the cause should be :
- 1. neither normal nor midx bitmap exists
- 2. only midx bitmap exists but core.multipackIndex="false"
- 3. core.multipackIndex="true" but midx bitmap file is currupt
- 4. core.multipackIndex="true" and no midx bitmap exists but
- normal bitmap file is currupt
+ 1. Neither normal nor MIDX bitmap exists.
+ 2. Only MIDX bitmap exists but core.multipackIndex="false".
+ 3. Config core.multipackIndex set to "true" but MIDX bitmap is
+ corrupted.
+ 4. Config core.multipackIndex set to "true" and no MIDX bitmap
+ exists but normal bitmap file is corrupted.
....
These are some of the scenarios I briefly tested, but maybe there are
- others (some scenarios is produced manually like "currupt bitmap file",
+ others (some scenarios is produced manually like "corrupted bitmap file",
but it's not represent it's an existed bug.).
Therefore, we added some TRACE2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is
- working on midx bitmap or normal bitmap, or is it simply because the
- related configuration is disabled. This may help with logging, user
- troubleshooting, and development debugging.
-
- Here are some output examples when executing
- "$GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD" under different
- situations:
-
- 1. _HAVE_ MIDX bitmap and NORMAL bitmap file, but the
- "core.multipackIndex" configures as "false":
-
- 19:38:43.007840 common-main.c:49 | d0 | main | version | | | | | 2.35.1.582.g8e9092487a
- 19:38:43.007874 common-main.c:50 | d0 | main | start | | 0.000305 | | | /opt/git/master/bin/git rev-list --test-bitmap HEAD
- 19:38:43.007950 compat/linux/procinfo.c:170 | d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
- 19:38:43.008091 git.c:460 | d0 | main | cmd_name | | | | | rev-list (rev-list)
- 19:38:43.008284 repo-settings.c:11 | d0 | main | data | r0 | 0.000720 | 0.000720 | config | feature.manyfiles:false
- 19:38:43.008297 repo-settings.c:11 | d0 | main | data | r0 | 0.000734 | 0.000734 | config | feature.experimental:false
- 19:38:43.008305 repo-settings.c:11 | d0 | main | data | r0 | 0.000742 | 0.000742 | config | core.commitgraph:true
- 19:38:43.008313 repo-settings.c:11 | d0 | main | data | r0 | 0.000749 | 0.000749 | config | commitgraph.readchangedpaths:true
- 19:38:43.008320 repo-settings.c:11 | d0 | main | data | r0 | 0.000756 | 0.000756 | config | gc.writecommitgraph:true
- 19:38:43.008327 repo-settings.c:11 | d0 | main | data | r0 | 0.000764 | 0.000764 | config | fetch.writecommitgraph:false
- 19:38:43.008334 repo-settings.c:11 | d0 | main | data | r0 | 0.000770 | 0.000770 | config | pack.usesparse:true
- 19:38:43.008341 repo-settings.c:11 | d0 | main | data | r0 | 0.000777 | 0.000777 | config | core.multipackindex:false
- 19:38:43.008348 repo-settings.c:11 | d0 | main | data | r0 | 0.000784 | 0.000784 | config | index.sparse:false
- 19:38:43.008724 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap
- 19:38:43.008738 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001173 | 0.000013 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap
- 19:38:43.008754 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001191 | 0.000031 | pack-bitmap | label:open_bitmap
- Bitmap v1 test (8 entries loaded)
- Found bitmap for 0a7df7ae92f8d8ed879c240f8ae9cdd33d18085e. 64 bits / 1801edc6 checksum
- 19:38:43.009099 progress.c:268 | d0 | main | region_enter | r0 | 0.001535 | | progress | label:Verifying bitmap entries
- Verifying bitmap entries: 100% (27/27), done.
- 19:38:43.009294 progress.c:339 | d0 | main | data | r0 | 0.001730 | 0.000195 | progress | ..total_objects:27
- 19:38:43.009302 progress.c:346 | d0 | main | region_leave | r0 | 0.001739 | 0.000204 | progress | label:Verifying bitmap entries
- OK!
- 19:38:43.009321 git.c:718 | d0 | main | exit | | 0.001757 | | | code:0
- 19:38:43.009329 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001766 | | | code:0
-
- 2. _HAVE_ MIDX bitmap and NORMAL bitmap file, but the
- "core.multipackIndex" configures as "true":
-
- (omit duplicate outputs)
- ...
- 20:02:31.288797 repo-settings.c:11 | d0 | main | data | r0 | 0.000811 | 0.000811 | config | core.multipackindex:true
- 20:02:31.288806 repo-settings.c:11 | d0 | main | data | r0 | 0.000819 | 0.000819 | config | index.sparse:false
- 20:02:31.288836 midx.c:185 | d0 | main | data | r0 | 0.000849 | 0.000849 | midx | load/num_packs:1
- 20:02:31.288843 midx.c:186 | d0 | main | data | r0 | 0.000857 | 0.000857 | midx | load/num_objects:27
- 20:02:31.289217 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001229 | | pack-bitmap | label:open_bitmap
- 20:02:31.289230 pack-bitmap.c:318 | d0 | main | data | r0 | 0.001242 | 0.000013 | midx | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/multi-pack-index-b6b04fbe053bd500d9ca13354466d3249dc275ac.bitmap
- 20:02:31.289252 pack-revindex.c:315 | d0 | main | data | r0 | 0.001265 | 0.000036 | load_midx_re | ..source:midx
- 20:02:31.289261 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001274 | 0.000045 | pack-bitmap | label:open_bitmap
- Bitmap v1 test (8 entries loaded)
- Found bitmap for 0a7df7ae92f8d8ed879c240f8ae9cdd33d18085e. 64 bits / 1801edc6 checksum
- 20:02:31.289594 progress.c:268 | d0 | main | region_enter | r0 | 0.001607 | | progress | label:Verifying bitmap entries
- Verifying bitmap entries: 100% (27/27), done.
- 20:02:31.289810 progress.c:339 | d0 | main | data | r0 | 0.001823 | 0.000216 | progress | ..total_objects:27
- 20:02:31.289824 progress.c:346 | d0 | main | region_leave | r0 | 0.001837 | 0.000230 | progress | label:Verifying bitmap entries
- OK!
- 20:02:31.289843 git.c:718 | d0 | main | exit | | 0.001856 | | | code:0
- 20:02:31.289860 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001873 | | | code:0
-
- 3. _HAVE_ MIDX bitmap and a corrupt NORMAL bitmap file, but the
- "core.multipackIndex" configures as "false" :
-
- (omit duplicate outputs)
- ...
- 20:14:06.539305 repo-settings.c:11 | d0 | main | data | r0 | 0.000794 | 0.000794 | config | core.multipackindex:false
- 20:14:06.539310 repo-settings.c:11 | d0 | main | data | r0 | 0.000799 | 0.000799 | config | index.sparse:false
- 20:14:06.539658 pack-bitmap.c:536 | d0 | main | region_enter | r0 | 0.001146 | | pack-bitmap | label:open_bitmap
- 20:14:06.539671 pack-bitmap.c:390 | d0 | main | data | r0 | 0.001160 | 0.000014 | bitmap | ..try to open bitmap:/home/tenglong.tl/test/dyrone/.git/objects/pack/pack-2d1b3f749fb859b874710e33263d0847ef009e03.bitmap
- 20:14:06.539686 usage.c:79 | d0 | main | error | | | | | ..Corrupted bitmap index file (wrong header)
- error: Corrupted bitmap index file (wrong header)
- 20:14:06.539696 pack-bitmap.c:426 | d0 | main | data | r0 | 0.001185 | 0.000039 | bitmap | ..load bitmap header:failed
- 20:14:06.539709 usage.c:79 | d0 | main | error | | | | | ..bitmap header is invalid
- error: bitmap header is invalid
- 20:14:06.539719 pack-bitmap.c:543 | d0 | main | region_leave | r0 | 0.001208 | 0.000062 | pack-bitmap | label:open_bitmap
- 20:14:06.539726 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
- fatal: failed to load bitmap indexes
- 20:14:06.539735 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128
- 20:14:06.539744 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001233 | | | code:128
+ working on MIDX or NORMAL bitmap at present, or the related config is
+ enabled or not. This may help with logging, user troubleshooting, and
+ development debugging.
+
+ Here are some brief output examples on different scenarios when
+ executing:
+
+ $GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD
+
+ Scenario 1: core.multipackIndex [false], midx bitmap exists [Y],
+ normal bitmap exists [N]
+
+ 19:21:56.580349 repo-settings.c:11 | d0 | main | data | r1 | 0.000827 | 0.000827 | config | core.multipackindex:false
+ 19:21:56.580356 repo-settings.c:11 | d0 | main | data | r1 | 0.000834 | 0.000834 | config | index.sparse:false
+ 19:21:56.580706 pack-bitmap.c:525 | d0 | main | region_enter | r1 | 0.001183 | | pack-bitmap | label:open_bitmap
+ 19:21:56.580719 pack-bitmap.c:386 | d0 | main | data | r1 | 0.001196 | 0.000013 | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
+ 19:21:56.580729 pack-bitmap.c:530 | d0 | main | region_leave | r1 | 0.001207 | 0.000024 | pack-bitmap | label:open_bitmap
+ 19:21:56.580737 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
+ fatal: failed to load bitmap indexes
+ 19:21:56.580746 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128
+ 19:21:56.580754 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001232 | | | code:128
+
+ Scenario 2: core.multipackIndex [false], midx bitmap exists [Y],
+ normal bitmap exists [Y]
+
+ 19:23:44.692384 repo-settings.c:11 | d0 | main | data | r0 | 0.000765 | 0.000765 | config | core.multipackindex:false
+ 19:23:44.692755 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001135 | | pack-bitmap | label:open_bitmap
+ 19:23:44.692768 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001149 | 0.000014 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
+ 19:23:44.692790 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001171 | 0.000036 | pack-bitmap | label:open_bitmap
+ Bitmap v1 test (1 entries loaded)
+ Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
+ 19:23:44.693119 progress.c:268 | d0 | main | region_enter | r0 | 0.001500 | | progress | label:Verifying bitmap entries
+ Verifying bitmap entries: 100% (3/3), done.
+ 19:23:44.693208 progress.c:339 | d0 | main | data | r0 | 0.001589 | 0.000089 | progress | ..total_objects:3
+ 19:23:44.693216 progress.c:346 | d0 | main | region_leave | r0 | 0.001597 | 0.000097 | progress | label:Verifying bitmap entries
+ OK!
+ 19:23:44.693234 git.c:718 | d0 | main | exit | | 0.001615 | | | code:0
+ 19:23:44.693244 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001625 | | | code:0
+
+ Scenario 3: core.multipackIndex [true], midx bitmap exists [Y],
+ normal bitmap exists [Y]
+
+ 19:26:03.625055 repo-settings.c:11 | d0 | main | data | r0 | 0.000760 | 0.000760 | config | core.multipackindex:true
+ 19:26:03.625090 midx.c:185 | d0 | main | data | r0 | 0.000795 | 0.000795 | midx | load/num_packs:1
+ 19:26:03.625097 midx.c:186 | d0 | main | data | r0 | 0.000803 | 0.000803 | midx | load/num_objects:3
+ 19:26:03.625455 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap
+ 19:26:03.625470 pack-bitmap.c:318 | d0 | main | data | r0 | 0.001175 | 0.000015 | midx | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/multi-pack-index-fe8e96790bd34926423bdf3efd762dbbea9f3213.bitmap
+ 19:26:03.625489 pack-revindex.c:315 | d0 | main | data | r0 | 0.001194 | 0.000034 | load_midx_re | ..source:midx
+ 19:26:03.625496 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001202 | 0.000042 | pack-bitmap | label:open_bitmap
+ Bitmap v1 test (1 entries loaded)
+ Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
+ 19:26:03.625818 progress.c:268 | d0 | main | region_enter | r0 | 0.001523 | | progress | label:Verifying bitmap entries
+ Verifying bitmap entries: 100% (3/3), done.
+ 19:26:03.625916 progress.c:339 | d0 | main | data | r0 | 0.001622 | 0.000099 | progress | ..total_objects:3
+ 19:26:03.625925 progress.c:346 | d0 | main | region_leave | r0 | 0.001630 | 0.000107 | progress | label:Verifying bitmap entries
+ OK!
+ 19:26:03.625943 git.c:718 | d0 | main | exit | | 0.001648 | | | code:0
+ 19:26:03.625952 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001658 | | | code:0
+
+ Situation 4: core.multipackIndex [false], midx bitmap exists [N],
+ normal bitmap exists [Y].
+
+ 19:27:15.383037 repo-settings.c:11 | d0 | main | data | r0 | 0.000746 | 0.000746 | config | core.multipackindex:true
+ 19:27:15.383397 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001105 | | pack-bitmap | label:open_bitmap
+ 19:27:15.383408 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001116 | 0.000011 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
+ 19:27:15.383419 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001128 | 0.000023 | pack-bitmap | label:open_bitmap
+ Bitmap v1 test (1 entries loaded)
+ Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
+ 19:27:15.383730 progress.c:268 | d0 | main | region_enter | r0 | 0.001439 | | progress | label:Verifying bitmap entries
+ Verifying bitmap entries: 100% (3/3), done.
+ 19:27:15.383822 progress.c:339 | d0 | main | data | r0 | 0.001531 | 0.000092 | progress | ..total_objects:3
+ 19:27:15.383830 progress.c:346 | d0 | main | region_leave | r0 | 0.001539 | 0.000100 | progress | label:Verifying bitmap entries
+ OK!
+ 19:27:15.383848 git.c:718 | d0 | main | exit | | 0.001557 | | | code:0
+ 19:27:15.383867 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001576 | | | code:0
+
+ Scenario 5: core.multipackIndex [true], midx bitmap exists [Y] but corrupted,
+ normal bitmap exists [Y]
+
+ 19:29:25.888233 repo-settings.c:11 | d0 | main | data | r0 | 0.000794 | 0.000794 | config | core.multipackindex:true
+ 19:29:25.888591 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001152 | | pack-bitmap | label:open_bitmap
+ 19:29:25.888603 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001163 | 0.000011 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
+ 19:29:25.888622 usage.c:79 | d0 | main | error | | | | | ..Corrupted bitmap index file (wrong header)
+ error: Corrupted bitmap index file (wrong header)
+ 19:29:25.888638 usage.c:79 | d0 | main | error | | | | | ..bitmap header is invalid
+ error: bitmap header is invalid
+ 19:29:25.888650 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001211 | 0.000059 | pack-bitmap | label:open_bitmap
+ 19:29:25.888659 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
+ fatal: failed to load bitmap indexes
+ 19:29:25.888670 usage.c:74 | d0 | main | exit | | 0.001231 | | | code:128
+ 19:29:25.888680 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001241 | | | code:128
@@ pack-bitmap.c: char *pack_bitmap_filename(struct packed_git *p)
struct stat st;
char *bitmap_name = midx_bitmap_filename(midx);
- int fd = git_open(bitmap_name);
-+ trace2_data_string("midx", the_repository, "try to open bitmap",
-+ bitmap_name);
++ trace2_data_string("midx", the_repository, "path", bitmap_name);
+ fd = git_open(bitmap_name);
free(bitmap_name);
-@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
- return -1;
-
- if (fstat(fd, &st)) {
-+ trace2_data_string("midx", the_repository, "stat bitmap file",
-+ "failed");
- close(fd);
- return -1;
- }
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
MAP_PRIVATE, fd, 0);
close(fd);
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, st
return -1;
bitmap_name = pack_bitmap_filename(packfile);
-+ trace2_data_string("bitmap", the_repository, "try to open bitmap",
-+ bitmap_name);
++ trace2_data_string("bitmap", the_repository, "path", bitmap_name);
fd = git_open(bitmap_name); free(bitmap_name);
-@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
- return -1;
-
- if (fstat(fd, &st)) {
-+ trace2_data_string("bitmap", the_repository, "stat bitmap file",
-+ "failed");
- close(fd);
- return -1;
- }
-@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
- }
-
- if (!is_pack_valid(packfile)) {
-+ trace2_data_string("bitmap", the_repository, "packfile", "invalid");
- close(fd);
- return -1;
- }
-@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
- close(fd);
-
- if (load_bitmap_header(bitmap_git) < 0) {
-+ trace2_data_string("bitmap", the_repository,
-+ "load bitmap header", "failed");
- munmap(bitmap_git->map, bitmap_git->map_size);
- bitmap_git->map = NULL;
- bitmap_git->map_size = 0;
@@ pack-bitmap.c: static int open_midx_bitmap(struct repository *r,
static int open_bitmap(struct repository *r,
struct bitmap_index *bitmap_git)
{
- assert(!bitmap_git->map);
-+ int ret = -1;
++ int ret = 0;
-- if (!open_midx_bitmap(r, bitmap_git))
-- return 0;
-- return open_pack_bitmap(r, bitmap_git);
+ assert(!bitmap_git->map);
+ trace2_region_enter("pack-bitmap", "open_bitmap", r);
-+ if (!open_midx_bitmap(r, bitmap_git)) {
-+ ret = 0;
+ if (!open_midx_bitmap(r, bitmap_git))
+- return 0;
+- return open_pack_bitmap(r, bitmap_git);
+ goto done;
-+ }
+ ret = open_pack_bitmap(r, bitmap_git);
+done:
+ trace2_region_leave("pack-bitmap", "open_bitmap", r);
@@ repo-settings.c: static void repo_cfg_bool(struct repository *r, const char *key
}
void prepare_repo_settings(struct repository *r)
-
- ## t/t5310-pack-bitmaps.sh ##
-@@ t/t5310-pack-bitmaps.sh: test_expect_success 'complains about multiple pack bitmaps' '
- test_line_count = 2 bitmaps &&
-
- git rev-list --use-bitmap-index HEAD 2>err &&
-- grep "a bitmap has been opened" err &&
-- grep "ignoring extra bitmap file" err
-+ grep "warning: a normal or midx bitmap already has been opened" err &&
-+ grep "warning: ignoring extra bitmap file" err
- )
- '
-
Signature · 1 line
2.35.1.582.g320e881567
PATCH v3 1/5 pack-bitmap.c: continue looping when first MIDX bitmap is found
In "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to continue the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The reason for this is, there may exist more than one MIDX file in a repo. The "multi_pack_index" struct is actually designed as a singly linked list, and if a MIDX file has been already opened successfully, then the other MIDX files will be skipped and left with a warning "ignoring extra bitmap file." to the output.
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -494,15 +494,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.1.582.g320e881567
PATCH v3 2/5 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index, let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -313,10 +313,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -368,14 +368,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.1.582.g320e881567
PATCH v3 3/5 pack-bitmap.c: make warnings support i18N when opening bitmap
In reply to Teng LongWhen calling the "open_midx_bitmap_1()" or "open_pack_bitmap_1()", there will be a warning if a normal bitmap or MIDX bitmap already has been opened, then let's make the warning text supporting for translation.
Discussion in community:
1. https://public-inbox.org/git/[email protected]/#t
2. https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
pack-bitmap.c +2 -2
@@ -330,7 +330,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning (_ ("ignoring extra bitmap file: %s" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +387,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: %s" ), packfile->pack_name);close (fd);return -1 ; }
Signature · 1 line
2.35.1.582.g320e881567
PATCH v3 4/5 pack-bitmap.c: using error() instead of silently returning -1
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
pack-bitmap.c +5 -5
@@ -323,7 +323,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (fstat (fd, &st)) {close (fd); -return -1 ; +return error (_ ("cannot stat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -361,7 +361,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error ("cannot open midx bitmap file" ); }static int open_pack_bitmap_1 (struct bitmap_index *bitmap_git,struct packed_git *packfile)
@@ -382,7 +382,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error (_ ("cannot stat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -394,7 +394,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {close (fd); -return -1 ; +return error (_ ("packfile is invalid" )); } bitmap_git->pack = packfile;
@@ -409,7 +409,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_git->map_size =0 ; bitmap_git->map_pos =0 ; bitmap_git->pack =NULL ; -return -1 ; +return error (_ ("bitmap header is invalid" )); }return 0 ;
Signature · 1 line
2.35.1.582.g320e881567
Re: [PATCH v3 4/5] pack-bitmap.c: using error() instead of silently returning -1
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 326
Quoted 6 lines · Teng Long
pack-bitmap.c +1
-1@@ -323,7 +323,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (fstat (fd, &st)) {close (fd); -return -1 ; +return error (_ ("cannot stat bitmap file" ));
Since we are handling an error from fstat here, the errno variable contains useful information that we should include in the error via error_errno().
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 364
Quoted 6 lines
pack-bitmap.c +1
-1@@ -361,7 +361,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error ("cannot open midx bitmap file" );
The other error strings are marked for translation, but this one is not. Was that intentional, or just a typo / oversight?
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 385
Quoted 7 lines
pack-bitmap.c +1
-1static int open_pack_bitmap_1 (struct bitmap_index *bitmap_git,struct packed_git *packfile)@@ -382,7 +382,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error (_ ("cannot stat bitmap file" ));
Same note here about using error_errno() instead of just error().
Thanks,
Taylor
Re: [PATCH v3 4/5] pack-bitmap.c: using error() instead of silently returning -1
Quoted 3 lines · Taylor Blau
Since we are handling an error from fstat here, the errno variable contains useful information that we should include in the error via
error_errno().
Appreciate for reminding this.
Will fix.
Quoted 2 lines
The other error strings are marked for translation, but this one is not. Was that intentional, or just a typo / oversight?
It's an oversight.
Will fix.
Quoted 1 line
Same note here about using
error_errno()instead of justerror().
As same as the first one, will fix.
Thanks.
PATCH v3 5/5 bitmap: add trace2 outputs during open "bitmap" file
In reply to Teng LongIt's supported for a repo to use bitmap in both "NORMAL" bitmap way or a MIDX (multi-pack-index) bitmap. Either of two bitmap kinds can exist in the repository, or both can be stored but let the config controls which kind of bitmap is used (like "core.multipackIndex", etc.). Because of this, sometimes the bitmap debug path is not obvious enough, for example, when executing:
$git rev-list --test-bitmap HEAD
fatal: failed to load bitmap indexes
If we see the output like this, It's not sure for us to know what's happened concretely, because the cause should be :
1. Neither normal nor MIDX bitmap exists.
2. Only MIDX bitmap exists but core.multipackIndex="false".
3. Config core.multipackIndex set to "true" but MIDX bitmap is corrupted.
4. Config core.multipackIndex set to "true" and no MIDX bitmap exists but normal bitmap file is corrupted.
....
These are some of the scenarios I briefly tested, but maybe there are others (some scenarios is produced manually like "corrupted bitmap file", but it's not represent it's an existed bug.).
Therefore, we added some TRACE2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is working on MIDX or NORMAL bitmap at present, or the related config is enabled or not. This may help with logging, user troubleshooting, and development debugging.
Here are some brief output examples on different scenarios when executing:
$GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD
Scenario 1: core.multipackIndex [false], midx bitmap exists [Y], normal bitmap exists [N]
19:21:56.580349 repo-settings.c:11 | d0 | main | data | r1 | 0.000827 | 0.000827 | config | core.multipackindex:false
19:21:56.580356 repo-settings.c:11 | d0 | main | data | r1 | 0.000834 | 0.000834 | config | index.sparse:false
19:21:56.580706 pack-bitmap.c:525 | d0 | main | region_enter | r1 | 0.001183 | | pack-bitmap | label:open_bitmap
19:21:56.580719 pack-bitmap.c:386 | d0 | main | data | r1 | 0.001196 | 0.000013 | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
19:21:56.580729 pack-bitmap.c:530 | d0 | main | region_leave | r1 | 0.001207 | 0.000024 | pack-bitmap | label:open_bitmap
19:21:56.580737 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
fatal: failed to load bitmap indexes
19:21:56.580746 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128
19:21:56.580754 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001232 | | | code:128Scenario 2: core.multipackIndex [false], midx bitmap exists [Y], normal bitmap exists [Y]
19:23:44.692384 repo-settings.c:11 | d0 | main | data | r0 | 0.000765 | 0.000765 | config | core.multipackindex:false 19:23:44.692755 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001135 | | pack-bitmap | label:open_bitmap 19:23:44.692768 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001149 | 0.000014 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap 19:23:44.692790 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001171 | 0.000036 | pack-bitmap | label:open_bitmap Bitmap v1 test (1 entries loaded) Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum 19:23:44.693119 progress.c:268 | d0 | main | region_enter | r0 | 0.001500 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (3/3), done. 19:23:44.693208 progress.c:339 | d0 | main | data | r0 | 0.001589 | 0.000089 | progress | ..total_objects:3 19:23:44.693216 progress.c:346 | d0 | main | region_leave | r0 | 0.001597 | 0.000097 | progress | label:Verifying bitmap entries OK! 19:23:44.693234 git.c:718 | d0 | main | exit | | 0.001615 | | | code:0 19:23:44.693244 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001625 | | | code:0
Scenario 3: core.multipackIndex [true], midx bitmap exists [Y], normal bitmap exists [Y]
19:26:03.625055 repo-settings.c:11 | d0 | main | data | r0 | 0.000760 | 0.000760 | config | core.multipackindex:true 19:26:03.625090 midx.c:185 | d0 | main | data | r0 | 0.000795 | 0.000795 | midx | load/num_packs:1 19:26:03.625097 midx.c:186 | d0 | main | data | r0 | 0.000803 | 0.000803 | midx | load/num_objects:3 19:26:03.625455 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap 19:26:03.625470 pack-bitmap.c:318 | d0 | main | data | r0 | 0.001175 | 0.000015 | midx | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/multi-pack-index-fe8e96790bd34926423bdf3efd762dbbea9f3213.bitmap 19:26:03.625489 pack-revindex.c:315 | d0 | main | data | r0 | 0.001194 | 0.000034 | load_midx_re | ..source:midx 19:26:03.625496 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001202 | 0.000042 | pack-bitmap | label:open_bitmap Bitmap v1 test (1 entries loaded) Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum 19:26:03.625818 progress.c:268 | d0 | main | region_enter | r0 | 0.001523 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (3/3), done. 19:26:03.625916 progress.c:339 | d0 | main | data | r0 | 0.001622 | 0.000099 | progress | ..total_objects:3 19:26:03.625925 progress.c:346 | d0 | main | region_leave | r0 | 0.001630 | 0.000107 | progress | label:Verifying bitmap entries OK! 19:26:03.625943 git.c:718 | d0 | main | exit | | 0.001648 | | | code:0 19:26:03.625952 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001658 | | | code:0
Situation 4: core.multipackIndex [false], midx bitmap exists [N], normal bitmap exists [Y].
19:27:15.383037 repo-settings.c:11 | d0 | main | data | r0 | 0.000746 | 0.000746 | config | core.multipackindex:true 19:27:15.383397 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001105 | | pack-bitmap | label:open_bitmap 19:27:15.383408 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001116 | 0.000011 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap 19:27:15.383419 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001128 | 0.000023 | pack-bitmap | label:open_bitmap Bitmap v1 test (1 entries loaded) Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum 19:27:15.383730 progress.c:268 | d0 | main | region_enter | r0 | 0.001439 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (3/3), done. 19:27:15.383822 progress.c:339 | d0 | main | data | r0 | 0.001531 | 0.000092 | progress | ..total_objects:3 19:27:15.383830 progress.c:346 | d0 | main | region_leave | r0 | 0.001539 | 0.000100 | progress | label:Verifying bitmap entries OK! 19:27:15.383848 git.c:718 | d0 | main | exit | | 0.001557 | | | code:0 19:27:15.383867 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001576 | | | code:0
Scenario 5: core.multipackIndex [true], midx bitmap exists [Y] but corrupted, normal bitmap exists [Y]
19:29:25.888233 repo-settings.c:11 | d0 | main | data | r0 | 0.000794 | 0.000794 | config | core.multipackindex:true 19:29:25.888591 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001152 | | pack-bitmap | label:open_bitmap 19:29:25.888603 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001163 | 0.000011 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap 19:29:25.888622 usage.c:79 | d0 | main | error | | | | | ..Corrupted bitmap index file (wrong header)
error: Corrupted bitmap index file (wrong header)
19:29:25.888638 usage.c:79 | d0 | main | error | | | | | ..bitmap header is invalid
error: bitmap header is invalid
19:29:25.888650 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001211 | 0.000059 | pack-bitmap | label:open_bitmap
19:29:25.888659 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
fatal: failed to load bitmap indexes
19:29:25.888670 usage.c:74 | d0 | main | exit | | 0.001231 | | | code:128
19:29:25.888680 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001241 | | | code:128
--- pack-bitmap.c | 27 +++++++++++++++++++++------ repo-settings.c | 1 + 2 files changed, 22 insertions(+), 6 deletions(-)
pack-bitmap.c +21 -6
@@ -312,9 +312,11 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx) { +int fd;struct stat st;char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"path" , bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);
@@ -343,12 +345,19 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, MAP_PRIVATE, fd, 0);close (fd); -if (load_bitmap_header (bitmap_git) <0 ) +if (load_bitmap_header (bitmap_git) <0 ) { +trace2_data_string ("midx" , the_repository,"load bitmap header" , +"failed" ); goto cleanup; + } -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +trace2_data_string ("midx" , the_repository,"verify checksum" , +"mismatch" ); goto cleanup; + } +if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" )); goto cleanup;
@@ -374,6 +383,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitreturn -1 ; bitmap_name =pack_bitmap_filename (packfile); +trace2_data_string ("bitmap" , the_repository,"path" , bitmap_name); fd =git_open (bitmap_name);free (bitmap_name);
@@ -509,11 +519,16 @@ static int open_midx_bitmap(struct repository *r,static int open_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { -assert (!bitmap_git->map); +int ret =0 ; +assert (!bitmap_git->map); +trace2_region_enter ("pack-bitmap" ,"open_bitmap" , r);if (!open_midx_bitmap (r, bitmap_git)) -return 0 ; -return open_pack_bitmap (r, bitmap_git); + goto done; + ret =open_pack_bitmap (r, bitmap_git); +done: +trace2_region_leave ("pack-bitmap" ,"open_bitmap" , r); +return ret; }struct bitmap_index *prepare_bitmap_git (struct repository *r)
repo-settings.c +1 -0
@@ -8,6 +8,7 @@ static void repo_cfg_bool(struct repository *r, const char *key, int *dest, {if (repo_config_get_bool (r, key, dest)) *dest = def; +trace2_data_string ("config" , r, key, *dest ?"true" :"false" ); }void prepare_repo_settings (struct repository *r)
Signature · 1 line
2.35.1.582.g320e881567
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 6 lines · Teng Long
pack-bitmap.c +0
-0@@ -312,9 +312,11 @@ char *pack_bitmap_filename(struct packed_git *p) static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
pack-bitmap.c char *pack_bitmap_filename(struct packed_git *p) · line 321
Quoted 10 lines
pack-bitmap.c +3
-1struct multi_pack_index *midx) { +int fd;struct stat st;char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"path" , bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);
The patch adds new statements at a wrong place. The block of declarations and the first statement in the block were separated by a blank line, but they no longer are.
These things tend to show up in merges quite clearly. They do not cause more unnecessary conflicts but can make resolution more error prone.
Thanks.
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 7 lines · Junio C Hamano
The patch adds new statements at a wrong place. The block of declarations and the first statement in the block were separated by a blank line, but they no longer are.
These things tend to show up in merges quite clearly. They do not cause more unnecessary conflicts but can make resolution more error prone.
You are absolutely right.
Will fix.
Thanks.
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
In reply to Teng LongQuoted 1 line · Teng Long
It's supported for a repo to use bitmap in both "NORMAL" bitmap way
I'm nitpicking, but I usually say "single-pack bitmap" or "multi-pack (MIDX) bitmap" when distinguishing between the two.
Quoted 8 lines
or a MIDX (multi-pack-index) bitmap. Either of two bitmap kinds can exist in the repository, or both can be stored but let the config controls which kind of bitmap is used (like "core.multipackIndex", etc.). Because of this, sometimes the bitmap debug path is not obvious enough, for example, when executing:
$git rev-list --test-bitmap HEAD
fatal: failed to load bitmap indexes
Odd spacing, and there should be a single space character separating "$" from "git" (like "$ git").
While I'm thinking about it: is the error message here up-to-date with the changes made by the previous patch?
Quoted 11 lines
If we see the output like this, It's not sure for us to know what's happened concretely, because the cause should be :
1. Neither normal nor MIDX bitmap exists.
2. Only MIDX bitmap exists but core.multipackIndex="false".
3. Config core.multipackIndex set to "true" but MIDX bitmap is corrupted.
4. Config core.multipackIndex set to "true" and no MIDX bitmap exists but normal bitmap file is corrupted.
....
Quoted 3 lines
These are some of the scenarios I briefly tested, but maybe there are others (some scenarios is produced manually like "corrupted bitmap file", but it's not represent it's an existed bug.).
This could probably be trimmed down for brevity, but I don't feel strongly about it. If you wanted to make your commit message a tad shorter, perhaps something like:
When a user sees output like this, it's unclear which kind(s) of .bitmap exist, and which were read. For example, it's possible a MIDX bitmap exists, but was not read (e.g., because core.multiPackIndex=false), among many other scenarios.
would suffice.
Quoted 1 line
Therefore, we added some TRACE2 code so that when we read the bitmap
s/TRACE2/trace2
Quoted 2 lines
we can be more clear about the decision path, such as whether it is working on MIDX or NORMAL bitmap at present, or the related config is
s/NORMAL/pack
Quoted 7 lines
enabled or not. This may help with logging, user troubleshooting, and development debugging.
Here are some brief output examples on different scenarios when executing:
$GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD
s/$GIT/$ GIT
Quoted 2 lines
Scenario 1: core.multipackIndex [false], midx bitmap exists [Y], normal bitmap exists [N]
The output here is quite wide, and I wonder if this whole section could be shortened. For example, scenario 2 is arguably more interesting than scenario 1 (I think readers would reasonably infer what happens in scenario 1 by reading what happens in scenario 2).
Quoted 13 lines
19:21:56.580349 repo-settings.c:11 | d0 | main | data | r1 | 0.000827 | 0.000827 | config | core.multipackindex:false 19:21:56.580356 repo-settings.c:11 | d0 | main | data | r1 | 0.000834 | 0.000834 | config | index.sparse:false 19:21:56.580706 pack-bitmap.c:525 | d0 | main | region_enter | r1 | 0.001183 | | pack-bitmap | label:open_bitmap 19:21:56.580719 pack-bitmap.c:386 | d0 | main | data | r1 | 0.001196 | 0.000013 | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap 19:21:56.580729 pack-bitmap.c:530 | d0 | main | region_leave | r1 | 0.001207 | 0.000024 | pack-bitmap | label:open_bitmap 19:21:56.580737 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes fatal: failed to load bitmap indexes 19:21:56.580746 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128 19:21:56.580754 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001232 | | | code:128Scenario 2: core.multipackIndex [false], midx bitmap exists [Y], normal bitmap exists [Y]
Quoted 13 lines
19:23:44.692384 repo-settings.c:11 | d0 | main | data | r0 | 0.000765 | 0.000765 | config | core.multipackindex:false 19:23:44.692755 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001135 | | pack-bitmap | label:open_bitmap 19:23:44.692768 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001149 | 0.000014 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap 19:23:44.692790 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001171 | 0.000036 | pack-bitmap | label:open_bitmap Bitmap v1 test (1 entries loaded) Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum 19:23:44.693119 progress.c:268 | d0 | main | region_enter | r0 | 0.001500 | | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (3/3), done. 19:23:44.693208 progress.c:339 | d0 | main | data | r0 | 0.001589 | 0.000089 | progress | ..total_objects:3 19:23:44.693216 progress.c:346 | d0 | main | region_leave | r0 | 0.001597 | 0.000097 | progress | label:Verifying bitmap entries OK! 19:23:44.693234 git.c:718 | d0 | main | exit | | 0.001615 | | | code:0 19:23:44.693244 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001625 | | | code:0
And scenario 2 could be cleaned up by just showing a few of the columns from the trace2 output. Perhaps along the lines of:
Quoted 5 lines
| data | r0 | 0.000765 | 0.000765 | config | core.multipackindex:false | region_enter | r0 | 0.001135 | | pack-bitmap | label:open_bitmap | data | r0 | 0.001149 | 0.000014 | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap | region_leave | r0 | 0.001171 | 0.000036 | pack-bitmap | label:open_bitmap | region_enter | r0 | 0.001500 | | progress | label:Verifying bitmap entries
Reading the below scenarios, I think just showing this example is more than sufficient for illustrating your point.
Quoted 24 lines
Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 27 +++++++++++++++++++++------ repo-settings.c | 1 + 2 files changed, 22 insertions(+), 6 deletions(-)pack-bitmap.c +3
-1@@ -312,9 +312,11 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx) { +int fd;struct stat st;char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"path" , bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 350
Quoted 7 lines
pack-bitmap.c +3
-1@@ -343,12 +345,19 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, MAP_PRIVATE, fd, 0);close (fd); -if (load_bitmap_header (bitmap_git) <0 ) +if (load_bitmap_header (bitmap_git) <0 ) { +trace2_data_string ("midx" , the_repository,"load bitmap header" , +"failed" );
I wonder, why don't we show these errors to the user? Should we introduce a "ret" variable and set it to (for e.g.)
ret = error(_("failed to load bitmap header"));or something?
repo-settings.c static void repo_cfg_bool(struct repository *r, const char *key, int *dest, · line 11
Quoted 10 lines
pack-bitmap.c +0
-0struct bitmap_index *prepare_bitmap_git (struct repository *r)repo-settings.c +1
-0@@ -8,6 +8,7 @@ static void repo_cfg_bool(struct repository *r, const char *key, int *dest, {if (repo_config_get_bool (r, key, dest)) *dest = def; +trace2_data_string ("config" , r, key, *dest ?"true" :"false" );
I'm not sure we want to dump the whole repository configuration into trace2 output. Is there a more convenient place to log any important value(s) _after_ they have been read?
Thanks,
Taylor
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 2 lines · Taylor Blau
I'm nitpicking, but I usually say "single-pack bitmap" or "multi-pack (MIDX) bitmap" when distinguishing between the two.
Ha, it's OK, I'm hungry. Your description is a better way I think.
Will optimize the relevant places.
Quoted 2 lines
Odd spacing, and there should be a single space character separating "$" from "git" (like "$ git").
Yeah, will Fix.
Quoted 2 lines
While I'm thinking about it: is the error message here up-to-date with the changes made by the previous patch?
It's still the same after the remake and test (actually it's just a case not all the cases have the same error message. now by the previous path if "core.multipackindex=false" and single-pack doesn't exist, the error message will look like this).
Quoted 10 lines
This could probably be trimmed down for brevity, but I don't feel strongly about it. If you wanted to make your commit message a tad shorter, perhaps something like:
When a user sees output like this, it's unclear which kind(s) of .bitmap exist, and which were read. For example, it's possible a MIDX bitmap exists, but was not read (e.g., because core.multiPackIndex=false), among many other scenarios.
would suffice.
I think yours is better, I will use it instead the listed way.
Quoted 1 line
s/TRACE2/trace2
Will fix.
Quoted 1 line
s/NORMAL/pack
Will fix as s/NORMAL/single-pack.
Quoted 1 line
s/$GIT/$ GIT
Will fix.
Quoted 10 lines
The output here is quite wide, and I wonder if this whole section could be shortened. For example, scenario 2 is arguably more interesting than scenario 1 (I think readers would reasonably infer what happens in scenario 1 by reading what happens in scenario 2).
...
And scenario 2 could be cleaned up by just showing a few of the columns from the trace2 output. Perhaps along the lines of:
...
Reading the below scenarios, I think just showing this example is more than sufficient for illustrating your point.
That makes sense. I'll keep two scenarios and omit some of the output to keep it shorter in next patch.
Quoted 6 lines
I wonder, why don't we show these errors to the user? Should we introduce a "ret" variable and set it to (for e.g.)
ret = error(_("failed to load bitmap header"));or something?
Just this "load header error", I think it's unnecessary to show more errors to uses because in "load_bitmap_header()", the function already does, but in my patch, it's unnecessary to add "trace2_data_string()" too, so here I will remove it (directly "goto cleanup" maybe) in next patch.
Quoted 3 lines
I'm not sure we want to dump the whole repository configuration into trace2 output. Is there a more convenient place to log any important value(s) _after_ they have been read?
I think maybe I can add a argument like "int trace" in next patch, so the caller can decide whether to print to trace or not.
Thanks for your help, Taylor.
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
In reply to Teng LongOn 6/12/22 3:44 AM, Teng Long wrote:
[...]
repo-settings.c static void repo_cfg_bool(struct repository *r, const char *key, int *dest, · line 12
Quoted 12 lines
struct
bitmap_index*prepare_bitmap_git(struct repository *r)repo-settings.c +1
-0@@ -8,6 +8,7 @@ static void repo_cfg_bool(struct repository *r, const char *key, int *dest, {if (repo_config_get_bool (r, key, dest)) *dest = def; +trace2_data_string ("config" , r, key, *dest ?"true" :"false" ); }
We should not be doing. This would dump every repo-related boolean value on every command.
I already have a GIT_TRACE2_CONFIG_PARAMS and trace2.configparams that will dump "interesting" config values to the trace2 log. Just set one of them to a list of regex's. Look at the comment above trace2_cmd_list_config() in trace2.h for details.
We also have GIT_TRACE2_ENV_VARS and trace2.envvars that will dump the values of "interesting" env vars.
You can use these in your testing to log the config and env var values that you are interested in.
Jeff
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 2 lines · Jeff Hostetler
We should not be doing. This would dump every repo-related boolean value on every command.
Yes. in v4 patch I use a new arg in repo_cfg_bool to decide whether to trace or not, but after the above you mentioned, I think it's still not a correct solution in my v4.
Quoted 4 lines
I already have a
GIT_TRACE2_CONFIG_PARAMSand trace2.configparams that will dump "interesting" config values to the trace2 log. Just set one of them to a list of regex's. Look at the comment abovetrace2_cmd_list_config()intrace2.hfor details.
That's fresh for me, thanks for telling me that.
I remove the changes of "repo_config_bool" from my v4 patch then I try to know about "GIT_TRACE2_CONFIG_PARAMS" and "trace2.configparams". Thereby, when I:
Execute "GIT_TRACE2_PERF=1 GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex git rev-list --test-bitmap HEAD"
15:21:38.812782 git.c:461 | d0 | main | def_param | | | | | core.multipackindex:false 15:21:38.812797 git.c:461 | d0 | main | def_param | | | | | core.multipackindex:false
I checked my configs, I found if there exists multiple level configs. it'll print multiple times. Like If I config all the global, system and local on "core.multipackIndex=false" , the output will be:
15:41:50.614108 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:41:50.614123 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:41:50.614136 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false
And if I modified the local scope of core.multipackIndex to "true", the output will be:
15:45:39.200172 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:45:39.200186 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:45:39.200200 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:true
I'm not sure it's an intentional design or here should be only print the final value that takes effect or should print all the values if config multiple scopes on the same config.
Hence, I made a temporary patch below to try to add some identifying information to know why we output these lines, like:
trace2/tr2_tgt_event.c +2 -0
@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);
commit a089800b9dbc93a5dff9a46da7e66111e6e4343e (HEAD -> master, dyrone/master, dyrone/HEAD)
Author: Teng Long <[email protected]> Date: Thu Jun 23 17:24:15 2022 +0800
tr2: append scope info when same configs exist in multiple scopes
trace2/tr2_tgt_event.c +4 -0
@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);jw_object_begin (&jw,0 );event_fmt_prepare (event_name, file, line,NULL , &jw);
+ jw_object_string(&jw, "scope", scope_name);
jw_object_string(&jw, "param", param);
jw_object_string(&jw, "value", value);
jw_end(&jw);trace2/tr2_tgt_normal.c +3 -1
@@ -298,8 +298,10 @@ static void fn_param_fl(const char *file, int line, const char *param, const char *value) {struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope); -strbuf_addf (&buf_payload,"def_param %s=%s" , param, value); +strbuf_addf (&buf_payload,"def_param scope:%s %s=%s" , scope_name, param, value);normal_io_write_fl (file, line, &buf_payload);strbuf_release (&buf_payload); }
trace2/tr2_tgt_perf.c +2 -2
@@ -441,10 +441,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);strbuf_addf (&buf_payload,"%s:%s" , param, value); - perf_io_write_fl(file, line, event_name, NULL, NULL, NULL, NULL, ----
The partial tr2 log looks like:
PERF:
17:00:35.933094 git.c:462 | d0 | main | def_param | | | | system | core.multipackindex:false 17:00:35.933110 git.c:462 | d0 | main | def_param | | | | global | core.multipackindex:false 17:00:35.933128 git.c:462 | d0 | main | def_param | | | | local | core.multipackindex:true
NORMAL:
17:14:32.905359 git.c:462 def_param scope:system core.multipackindex=false 17:14:32.905370 git.c:462 def_param scope:global core.multipackindex=false 17:14:32.905383 git.c:462 def_param scope:local core.multipackindex=true
EVENT:
{"event":"def_param","sid":"20220623T092115.703660Z-H82fddc29-P0001812a","thread":"main","time":"2022-06-23T09:21:15.703920Z","file":"git.c","line":462,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param","sid":"20220623T092115.703660Z-H82fddc29-P0001812a","thread":"main","time":"2022-06-23T09:21:15.703936Z","file":"git.c","line":462,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param","sid":"20220623T092115.703660Z-H82fddc29-P0001812a","thread":"main","time":"2022-06-23T09:21:15.703952Z","file":"git.c","line":462,"scope":"local","param":"core.multipackindex","value":"true"}
Quoted 5 lines
We also have
GIT_TRACE2_ENV_VARSand trace2.envvars that will dump the values of "interesting" env vars.You can use these in your testing to log the config and env var values that you are interested in.
Wow, cool as GIT_TRACE2_CONFIG_PARAMS and trace2.configparams you mentioned before.
Thanks.
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 2 lines · Teng Long
On Wed, 22 Jun 2022 08:51:18 -0400, Jeff Hostetler wrote:
[...]
Quoted 24 lines
I remove the changes of "
repo_config_bool" from my v4 patch then I try to know about "GIT_TRACE2_CONFIG_PARAMS" and "trace2.configparams". Thereby, when I:Execute "GIT_TRACE2_PERF=1 GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex git rev-list --test-bitmap HEAD"
15:21:38.812782 git.c:461 | d0 | main | def_param | | | | | core.multipackindex:false 15:21:38.812797 git.c:461 | d0 | main | def_param | | | | | core.multipackindex:falseI checked my configs, I found if there exists multiple level configs. it'll print multiple times. Like If I config all the global, system and local on "core.multipackIndex=false" , the output will be:
15:41:50.614108 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:41:50.614123 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:41:50.614136 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:falseAnd if I modified the local scope of core.multipackIndex to "true", the output will be:
15:45:39.200172 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:45:39.200186 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:false 15:45:39.200200 git.c:462 | d0 | main | def_param | | | | | core.multipackindex:true
Quoted 6 lines
I'm not sure it's an intentional design or here should be only print the final value that takes effect or should print all the values if config multiple scopes on the same config.
Hence, I made a temporary patch below to try to add some identifying information to know why we output these lines, like:
I had intended it to only print the final "effective" value that the command would actually use. I don't remember if I ever experimented with config values that are set at multiple levels, but testing it now shows that you're right it does print a "def_param" message for each level.
I would suggest we fix it to only print the final value since we really just want to know whether the command will or will not use a feature. The inheritance is somewhat irrelevant.
But the change to do that may be quite a mess, so printing the scope name as you have here is fine. And may help with support when debugging users having confusing problems, so I'm fine with it.
Thanks,
Jeff
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap"
Quoted 11 lines · Jeff Hostetler
I had intended it to only print the final "effective" value that the command would actually use. I don't remember if I ever experimented with config values that are set at multiple levels, but testing it now shows that you're right it does print a "
def_param" message for each level.I would suggest we fix it to only print the final value since we really just want to know whether the command will or will not use a feature. The inheritance is somewhat irrelevant.
Quoted 4 lines
But the change to do that may be quite a mess, so printing the scope name as you have here is fine. And may help with support when debugging users having confusing problems, so I'm fine with it.
Yes, as I replied both way is OK for me, but sorry for only replied the append-scope way. I will post a new patch v5 and include the related changes to fix this.
Thanks.
v3 cover trace2 output for bitmap decision path
In reply to Teng LongMain diff from v4 to v3:
* Optimize the commit message of [5/5].
* Fix the missing blank line in [5/5].
* Only print the config we want to instead of the whole configs .
* use "error_errno()" instead in [4/5].
Thanks for Junio C Hamano and Taylor Blau for their help in v3.
Teng Long (5):
pack-bitmap.c: continue looping when first MIDX bitmap is found pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: make warnings support i18N when opening bitmap pack-bitmap.c: using error() instead of silently returning -1
bitmap: add trace2 outputs during open "bitmap" file
pack-bitmap.c | 52 ++++++++++++++++++++++++++++++------------------- repo-settings.c | 22 +++++++++++---------- 2 files changed, 44 insertions(+), 30 deletions(-)
Range-diff against v2:
-: ---------- > 1: 589e3f4075 pack-bitmap.c: continue looping when first MIDX bitmap is found
-: ---------- > 2: b6b30047fc pack-bitmap.c: rename "idx_name" to "bitmap_name"
-: ---------- > 3: d8dfe53dd4 pack-bitmap.c: make warnings support i18N when opening bitmap
1: 72da3b5844 ! 4: 917551f2b5 pack-bitmap.c: using error() instead of silently returning -1
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
if (fstat(fd, &st)) {
close(fd);
- return -1;
-+ return error(_("cannot stat bitmap file"));
++ return error_errno(_("cannot stat bitmap file"));
}
if (bitmap_git->pack || bitmap_git->midx) {
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
bitmap_git->map = NULL;
bitmap_git->midx = NULL;
- return -1;
-+ return error("cannot open midx bitmap file");
++ return error(_("cannot open midx bitmap file"));
}
static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git *packfile)
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, st
if (fstat(fd, &st)) {
close(fd);
- return -1;
-+ return error(_("cannot stat bitmap file"));
++ return error_errno(_("cannot stat bitmap file"));
}
if (bitmap_git->pack || bitmap_git->midx) {2: e118758d1d ! 5: 8735ae9979 bitmap: add trace2 outputs during open "bitmap" file
@@ Metadata
## Commit message ##
bitmap: add trace2 outputs during open "bitmap" file
- It's supported for a repo to use bitmap in both "NORMAL" bitmap way
- or a MIDX (multi-pack-index) bitmap. Either of two bitmap kinds can
+ It's supported for a repo to use bitmap in both single-pack bitmap
+ way or a multi-pack(MIDX) bitmap. Either of two bitmap kinds can
exist in the repository, or both can be stored but let the config controls which kind of bitmap is used (like "core.multipackIndex", etc.). Because of this, sometimes the bitmap debug path is not obvious enough, for example, when executing:
- $git rev-list --test-bitmap HEAD
+ $ git rev-list --test-bitmap HEAD
fatal: failed to load bitmap indexes
- If we see the output like this, It's not sure for us to know
- what's happened concretely, because the cause should be :
+ When the output look like this, it's unclear which kind(s) of
+ .bitmap exist, and which were read. For example, it's possible a MIDX
+ bitmap exists, but was not read (e.g., because
+ core.multiPackIndex=false), among many other scenarios.
- 1. Neither normal nor MIDX bitmap exists.
- 2. Only MIDX bitmap exists but core.multipackIndex="false".
- 3. Config core.multipackIndex set to "true" but MIDX bitmap is
- corrupted.
- 4. Config core.multipackIndex set to "true" and no MIDX bitmap
- exists but normal bitmap file is corrupted.
- ....
-
- These are some of the scenarios I briefly tested, but maybe there are
- others (some scenarios is produced manually like "corrupted bitmap file",
- but it's not represent it's an existed bug.).
-
- Therefore, we added some TRACE2 code so that when we read the bitmap
+ Therefore, we added some trace2 code so that when we read the bitmap
we can be more clear about the decision path, such as whether it is
- working on MIDX or NORMAL bitmap at present, or the related config is
- enabled or not. This may help with logging, user troubleshooting, and
+ working on MIDX or single-pack bitmap at present, or the related config
+ is enabled or not. This may help with logging, user troubleshooting, and
development debugging.
- Here are some brief output examples on different scenarios when
+ Here are some brief output (omitted some unrelated or repetitive rows
+ and columns, using "..." instead of) examples on two scenarios when
executing:
- $GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD
-
- Scenario 1: core.multipackIndex [false], midx bitmap exists [Y],
- normal bitmap exists [N]
-
- 19:21:56.580349 repo-settings.c:11 | d0 | main | data | r1 | 0.000827 | 0.000827 | config | core.multipackindex:false
- 19:21:56.580356 repo-settings.c:11 | d0 | main | data | r1 | 0.000834 | 0.000834 | config | index.sparse:false
- 19:21:56.580706 pack-bitmap.c:525 | d0 | main | region_enter | r1 | 0.001183 | | pack-bitmap | label:open_bitmap
- 19:21:56.580719 pack-bitmap.c:386 | d0 | main | data | r1 | 0.001196 | 0.000013 | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
- 19:21:56.580729 pack-bitmap.c:530 | d0 | main | region_leave | r1 | 0.001207 | 0.000024 | pack-bitmap | label:open_bitmap
- 19:21:56.580737 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
- fatal: failed to load bitmap indexes
- 19:21:56.580746 usage.c:74 | d0 | main | exit | | 0.001224 | | | code:128
- 19:21:56.580754 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001232 | | | code:128
-
- Scenario 2: core.multipackIndex [false], midx bitmap exists [Y],
- normal bitmap exists [Y]
-
- 19:23:44.692384 repo-settings.c:11 | d0 | main | data | r0 | 0.000765 | 0.000765 | config | core.multipackindex:false
- 19:23:44.692755 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001135 | | pack-bitmap | label:open_bitmap
- 19:23:44.692768 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001149 | 0.000014 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
- 19:23:44.692790 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001171 | 0.000036 | pack-bitmap | label:open_bitmap
- Bitmap v1 test (1 entries loaded)
- Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
- 19:23:44.693119 progress.c:268 | d0 | main | region_enter | r0 | 0.001500 | | progress | label:Verifying bitmap entries
- Verifying bitmap entries: 100% (3/3), done.
- 19:23:44.693208 progress.c:339 | d0 | main | data | r0 | 0.001589 | 0.000089 | progress | ..total_objects:3
- 19:23:44.693216 progress.c:346 | d0 | main | region_leave | r0 | 0.001597 | 0.000097 | progress | label:Verifying bitmap entries
- OK!
- 19:23:44.693234 git.c:718 | d0 | main | exit | | 0.001615 | | | code:0
- 19:23:44.693244 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001625 | | | code:0
-
- Scenario 3: core.multipackIndex [true], midx bitmap exists [Y],
- normal bitmap exists [Y]
-
- 19:26:03.625055 repo-settings.c:11 | d0 | main | data | r0 | 0.000760 | 0.000760 | config | core.multipackindex:true
- 19:26:03.625090 midx.c:185 | d0 | main | data | r0 | 0.000795 | 0.000795 | midx | load/num_packs:1
- 19:26:03.625097 midx.c:186 | d0 | main | data | r0 | 0.000803 | 0.000803 | midx | load/num_objects:3
- 19:26:03.625455 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001160 | | pack-bitmap | label:open_bitmap
- 19:26:03.625470 pack-bitmap.c:318 | d0 | main | data | r0 | 0.001175 | 0.000015 | midx | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/multi-pack-index-fe8e96790bd34926423bdf3efd762dbbea9f3213.bitmap
- 19:26:03.625489 pack-revindex.c:315 | d0 | main | data | r0 | 0.001194 | 0.000034 | load_midx_re | ..source:midx
- 19:26:03.625496 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001202 | 0.000042 | pack-bitmap | label:open_bitmap
- Bitmap v1 test (1 entries loaded)
- Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
- 19:26:03.625818 progress.c:268 | d0 | main | region_enter | r0 | 0.001523 | | progress | label:Verifying bitmap entries
- Verifying bitmap entries: 100% (3/3), done.
- 19:26:03.625916 progress.c:339 | d0 | main | data | r0 | 0.001622 | 0.000099 | progress | ..total_objects:3
- 19:26:03.625925 progress.c:346 | d0 | main | region_leave | r0 | 0.001630 | 0.000107 | progress | label:Verifying bitmap entries
- OK!
- 19:26:03.625943 git.c:718 | d0 | main | exit | | 0.001648 | | | code:0
- 19:26:03.625952 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001658 | | | code:0
-
- Situation 4: core.multipackIndex [false], midx bitmap exists [N],
- normal bitmap exists [Y].
-
- 19:27:15.383037 repo-settings.c:11 | d0 | main | data | r0 | 0.000746 | 0.000746 | config | core.multipackindex:true
- 19:27:15.383397 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001105 | | pack-bitmap | label:open_bitmap
- 19:27:15.383408 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001116 | 0.000011 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
- 19:27:15.383419 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001128 | 0.000023 | pack-bitmap | label:open_bitmap
- Bitmap v1 test (1 entries loaded)
- Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
- 19:27:15.383730 progress.c:268 | d0 | main | region_enter | r0 | 0.001439 | | progress | label:Verifying bitmap entries
- Verifying bitmap entries: 100% (3/3), done.
- 19:27:15.383822 progress.c:339 | d0 | main | data | r0 | 0.001531 | 0.000092 | progress | ..total_objects:3
- 19:27:15.383830 progress.c:346 | d0 | main | region_leave | r0 | 0.001539 | 0.000100 | progress | label:Verifying bitmap entries
- OK!
- 19:27:15.383848 git.c:718 | d0 | main | exit | | 0.001557 | | | code:0
- 19:27:15.383867 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001576 | | | code:0
-
- Scenario 5: core.multipackIndex [true], midx bitmap exists [Y] but corrupted,
- normal bitmap exists [Y]
-
- 19:29:25.888233 repo-settings.c:11 | d0 | main | data | r0 | 0.000794 | 0.000794 | config | core.multipackindex:true
- 19:29:25.888591 pack-bitmap.c:525 | d0 | main | region_enter | r0 | 0.001152 | | pack-bitmap | label:open_bitmap
- 19:29:25.888603 pack-bitmap.c:386 | d0 | main | data | r0 | 0.001163 | 0.000011 | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
- 19:29:25.888622 usage.c:79 | d0 | main | error | | | | | ..Corrupted bitmap index file (wrong header)
- error: Corrupted bitmap index file (wrong header)
- 19:29:25.888638 usage.c:79 | d0 | main | error | | | | | ..bitmap header is invalid
- error: bitmap header is invalid
- 19:29:25.888650 pack-bitmap.c:530 | d0 | main | region_leave | r0 | 0.001211 | 0.000059 | pack-bitmap | label:open_bitmap
- 19:29:25.888659 usage.c:60 | d0 | main | error | | | | | failed to load bitmap indexes
- fatal: failed to load bitmap indexes
- 19:29:25.888670 usage.c:74 | d0 | main | exit | | 0.001231 | | | code:128
- 19:29:25.888680 trace2/tr2_tgt_perf.c:215 | d0 | main | atexit | | 0.001241 | | | code:128
+ $ GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD
+
+ Scenario 1:
+ core.multipackIndex [false], MIDX bitmap exists [Y], single-pack bitmap exists [N]
+
+ ...
+ ... | main | data | r1 | ... | config | core.multipackindex:false
+ ... | d0 | main | region_enter | r1 | ... | pack-bitmap | label:open_bitmap
+ ... | d0 | main | data | r1 | ... | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
+ ... | main | region_leave | r1 | ... | pack-bitmap | label:open_bitmap
+ ... | main | error | | ... | | failed to load bitmap indexes
+ fatal: failed to load bitmap indexes
+ ... | d0 | main | exit | | ... | | code:128
+ ...
+
+ Scenario 2:
+ core.multipackIndex [false], MIDX bitmap exists [Y], single-pack bitmap exists [Y]
+
+ ... | d0 | main | region_enter | r0 | ... | pack-bitmap | label:open_bitmap
+ ... | d0 | main | data | r0 | ... | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap
+ ... | main | region_leave | r0 | ... | pack-bitmap | label:open_bitmap
+ Bitmap v1 test (1 entries loaded)
+ Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum
+ ... | main | region_enter | r0 | ... | progress | label:Verifying bitmap entries
+ Verifying bitmap entries: 100% (3/3), done.
+ ... | main | data | r0 | ... | progress | ..total_objects:3
+ ... | main | region_leave | r0 | ... | progress | label:Verifying bitmap entries
+ OK!
+ ... | d0 | main | exit | | ... | | code:0
+ ...
@@ pack-bitmap.c: char *pack_bitmap_filename(struct packed_git *p)
{
+ int fd;
struct stat st;
++
char *bitmap_name = midx_bitmap_filename(midx);
- int fd = git_open(bitmap_name);
+ trace2_data_string("midx", the_repository, "path", bitmap_name);
@@ pack-bitmap.c: char *pack_bitmap_filename(struct packed_git *p)
free(bitmap_name);
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
- MAP_PRIVATE, fd, 0);
- close(fd);
-
-- if (load_bitmap_header(bitmap_git) < 0)
-+ if (load_bitmap_header(bitmap_git) < 0) {
-+ trace2_data_string("midx", the_repository, "load bitmap header",
-+ "failed");
+ if (load_bitmap_header(bitmap_git) < 0)
goto cleanup;
-+ }
- if (!hasheq(get_midx_checksum(bitmap_git->midx), bitmap_git->checksum))
+ if (!hasheq(get_midx_checksum(bitmap_git->midx), bitmap_git->checksum)) {
-+ trace2_data_string("midx", the_repository, "verify checksum",
-+ "mismatch");
++ error(_("midx and bitmap checksum don't match"));
goto cleanup;
-
+ }
-+
+
if (load_midx_revindex(bitmap_git->midx) < 0) {
warning(_("multi-pack bitmap is missing required reverse index"));
- goto cleanup;
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
return -1;
@@ pack-bitmap.c: static int open_midx_bitmap(struct repository *r,
struct bitmap_index *prepare_bitmap_git(struct repository *r)
## repo-settings.c ##
-@@ repo-settings.c: static void repo_cfg_bool(struct repository *r, const char *key, int *dest, +@@
+ #include "midx.h"
+
+ static void repo_cfg_bool(struct repository *r, const char *key, int *dest,
+- int def)
++ int def, int trace)
{
if (repo_config_get_bool(r, key, dest))
*dest = def;
-+ trace2_data_string("config", r, key, *dest ? "true" : "false");
++ if (trace)
++ trace2_data_string("config", r, key, *dest ? "true" : "false");
}
void prepare_repo_settings(struct repository *r)
+@@ repo-settings.c: void prepare_repo_settings(struct repository *r)
+ r->settings.fetch_negotiation_algorithm = FETCH_NEGOTIATION_CONSECUTIVE;
+
+ /* Booleans config or default, cascades to other settings */
+- repo_cfg_bool(r, "feature.manyfiles", &manyfiles, 0);
+- repo_cfg_bool(r, "feature.experimental", &experimental, 0);
++ repo_cfg_bool(r, "feature.manyfiles", &manyfiles, 0, 0);
++ repo_cfg_bool(r, "feature.experimental", &experimental, 0, 0);
+
+ /* Defaults modified by feature.* */
+ if (experimental) {
+@@ repo-settings.c: void prepare_repo_settings(struct repository *r)
+ }
+
+ /* Boolean config or default, does not cascade (simple) */
+- repo_cfg_bool(r, "core.commitgraph", &r->settings.core_commit_graph, 1);
+- repo_cfg_bool(r, "commitgraph.readchangedpaths", &r->settings.commit_graph_read_changed_paths, 1);
+- repo_cfg_bool(r, "gc.writecommitgraph", &r->settings.gc_write_commit_graph, 1);
+- repo_cfg_bool(r, "fetch.writecommitgraph", &r->settings.fetch_write_commit_graph, 0);
+- repo_cfg_bool(r, "pack.usesparse", &r->settings.pack_use_sparse, 1);
+- repo_cfg_bool(r, "core.multipackindex", &r->settings.core_multi_pack_index, 1);
+- repo_cfg_bool(r, "index.sparse", &r->settings.sparse_index, 0);
++ repo_cfg_bool(r, "core.commitgraph", &r->settings.core_commit_graph, 1, 0);
++ repo_cfg_bool(r, "commitgraph.readchangedpaths", &r->settings.commit_graph_read_changed_paths, 1, 0);
++ repo_cfg_bool(r, "gc.writecommitgraph", &r->settings.gc_write_commit_graph, 1, 0);
++ repo_cfg_bool(r, "fetch.writecommitgraph", &r->settings.fetch_write_commit_graph, 0, 0);
++ repo_cfg_bool(r, "pack.usesparse", &r->settings.pack_use_sparse, 1, 0);
++ repo_cfg_bool(r, "core.multipackindex", &r->settings.core_multi_pack_index, 1, 1);
++ repo_cfg_bool(r, "index.sparse", &r->settings.sparse_index, 0, 0);
+
+ /*
+ * The GIT_TEST_MULTI_PACK_INDEX variable is special in that
Signature · 1 line
2.35.1.582.g270d558070.dirty
PATCH v3 1/5 pack-bitmap.c: continue looping when first MIDX bitmap is found
In "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to continue the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The reason for this is, there may exist more than one MIDX file in a repo. The "multi_pack_index" struct is actually designed as a singly linked list, and if a MIDX file has been already opened successfully, then the other MIDX files will be skipped and left with a warning "ignoring extra bitmap file." to the output.
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -494,15 +494,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.1.582.g270d558070.dirty
PATCH v3 2/5 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index, let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -313,10 +313,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -368,14 +368,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.1.582.g270d558070.dirty
PATCH v3 3/5 pack-bitmap.c: make warnings support i18N when opening bitmap
In reply to Teng LongWhen calling the "open_midx_bitmap_1()" or "open_pack_bitmap_1()", there will be a warning if a normal bitmap or MIDX bitmap already has been opened, then let's make the warning text supporting for translation.
Discussion in community:
1. https://public-inbox.org/git/[email protected]/#t
2. https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
pack-bitmap.c +2 -2
@@ -330,7 +330,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning (_ ("ignoring extra bitmap file: %s" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +387,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: %s" ), packfile->pack_name);close (fd);return -1 ; }
Signature · 1 line
2.35.1.582.g270d558070.dirty
PATCH v3 4/5 pack-bitmap.c: using error() instead of silently returning -1
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
pack-bitmap.c +5 -5
@@ -323,7 +323,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot stat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -361,7 +361,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error (_ ("cannot open midx bitmap file" )); }static int open_pack_bitmap_1 (struct bitmap_index *bitmap_git,struct packed_git *packfile)
@@ -382,7 +382,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot stat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -394,7 +394,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {close (fd); -return -1 ; +return error (_ ("packfile is invalid" )); } bitmap_git->pack = packfile;
@@ -409,7 +409,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_git->map_size =0 ; bitmap_git->map_pos =0 ; bitmap_git->pack =NULL ; -return -1 ; +return error (_ ("bitmap header is invalid" )); }return 0 ;
Signature · 1 line
2.35.1.582.g270d558070.dirty
PATCH v3 5/5 bitmap: add trace2 outputs during open "bitmap" file
In reply to Teng LongIt's supported for a repo to use bitmap in both single-pack bitmap way or a multi-pack(MIDX) bitmap. Either of two bitmap kinds can exist in the repository, or both can be stored but let the config controls which kind of bitmap is used (like "core.multipackIndex", etc.). Because of this, sometimes the bitmap debug path is not obvious enough, for example, when executing:
$ git rev-list --test-bitmap HEAD fatal: failed to load bitmap indexes
When the output look like this, it's unclear which kind(s) of .bitmap exist, and which were read. For example, it's possible a MIDX bitmap exists, but was not read (e.g., because core.multiPackIndex=false), among many other scenarios.
Therefore, we added some trace2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is working on MIDX or single-pack bitmap at present, or the related config is enabled or not. This may help with logging, user troubleshooting, and development debugging.
Here are some brief output (omitted some unrelated or repetitive rows and columns, using "..." instead of) examples on two scenarios when executing:
$ GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEAD
Scenario 1:
core.multipackIndex [false], MIDX bitmap exists [Y], single-pack bitmap exists [N]
...
... | main | data | r1 | ... | config | core.multipackindex:false ... | d0 | main | region_enter | r1 | ... | pack-bitmap | label:open_bitmap ... | d0 | main | data | r1 | ... | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap ... | main | region_leave | r1 | ... | pack-bitmap | label:open_bitmap ... | main | error | | ... | | failed to load bitmap indexes
fatal: failed to load bitmap indexes
... | d0 | main | exit | | ... | | code:128
...
Scenario 2:
core.multipackIndex [false], MIDX bitmap exists [Y], single-pack bitmap exists [Y]
... | d0 | main | region_enter | r0 | ... | pack-bitmap | label:open_bitmap ... | d0 | main | data | r0 | ... | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap ... | main | region_leave | r0 | ... | pack-bitmap | label:open_bitmap Bitmap v1 test (1 entries loaded) Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum ... | main | region_enter | r0 | ... | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (3/3), done. ... | main | data | r0 | ... | progress | ..total_objects:3 ... | main | region_leave | r0 | ... | progress | label:Verifying bitmap entries
OK!
... | d0 | main | exit | | ... | | code:0
...
--- pack-bitmap.c | 21 ++++++++++++++++----- repo-settings.c | 22 ++++++++++++---------- 2 files changed, 28 insertions(+), 15 deletions(-)
pack-bitmap.c +16 -5
@@ -312,9 +312,12 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx) { +int fd;struct stat st; +char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"path" , bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);
@@ -346,8 +349,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (load_bitmap_header (bitmap_git) <0 ) goto cleanup; -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +error (_ ("midx and bitmap checksum don't match" )); goto cleanup; + }if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" ));
@@ -374,6 +379,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitreturn -1 ; bitmap_name =pack_bitmap_filename (packfile); +trace2_data_string ("bitmap" , the_repository,"path" , bitmap_name); fd =git_open (bitmap_name);free (bitmap_name);
@@ -509,11 +515,16 @@ static int open_midx_bitmap(struct repository *r,static int open_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { -assert (!bitmap_git->map); +int ret =0 ; +assert (!bitmap_git->map); +trace2_region_enter ("pack-bitmap" ,"open_bitmap" , r);if (!open_midx_bitmap (r, bitmap_git)) -return 0 ; -return open_pack_bitmap (r, bitmap_git); + goto done; + ret =open_pack_bitmap (r, bitmap_git); +done: +trace2_region_leave ("pack-bitmap" ,"open_bitmap" , r); +return ret; }struct bitmap_index *prepare_bitmap_git (struct repository *r)
repo-settings.c +12 -10
@@ -4,10 +4,12 @@#include "midx.h" static void repo_cfg_bool (struct repository *r,const char *key,int *dest, -int def) +int def,int trace) {if (repo_config_get_bool (r, key, dest)) *dest = def; +if (trace) +trace2_data_string ("config" , r, key, *dest ?"true" :"false" ); }void prepare_repo_settings (struct repository *r)
@@ -29,8 +31,8 @@ void prepare_repo_settings(struct repository *r) r->settings.fetch_negotiation_algorithm =FETCH_NEGOTIATION_CONSECUTIVE ;/* Booleans config or default, cascades to other settings */ -repo_cfg_bool (r,"feature.manyfiles" , &manyfiles,0 ); -repo_cfg_bool (r,"feature.experimental" , &experimental,0 ); +repo_cfg_bool (r,"feature.manyfiles" , &manyfiles,0 ,0 ); +repo_cfg_bool (r,"feature.experimental" , &experimental,0 ,0 );/* Defaults modified by feature.* */ if (experimental) {
@@ -42,13 +44,13 @@ void prepare_repo_settings(struct repository *r) }/* Boolean config or default, does not cascade (simple) */ -repo_cfg_bool (r,"core.commitgraph" , &r->settings.core_commit_graph,1 ); -repo_cfg_bool (r,"commitgraph.readchangedpaths" , &r->settings.commit_graph_read_changed_paths,1 ); -repo_cfg_bool (r,"gc.writecommitgraph" , &r->settings.gc_write_commit_graph,1 ); -repo_cfg_bool (r,"fetch.writecommitgraph" , &r->settings.fetch_write_commit_graph,0 ); -repo_cfg_bool (r,"pack.usesparse" , &r->settings.pack_use_sparse,1 ); -repo_cfg_bool (r,"core.multipackindex" , &r->settings.core_multi_pack_index,1 ); -repo_cfg_bool (r,"index.sparse" , &r->settings.sparse_index,0 ); +repo_cfg_bool (r,"core.commitgraph" , &r->settings.core_commit_graph,1 ,0 ); +repo_cfg_bool (r,"commitgraph.readchangedpaths" , &r->settings.commit_graph_read_changed_paths,1 ,0 ); +repo_cfg_bool (r,"gc.writecommitgraph" , &r->settings.gc_write_commit_graph,1 ,0 ); +repo_cfg_bool (r,"fetch.writecommitgraph" , &r->settings.fetch_write_commit_graph,0 ,0 ); +repo_cfg_bool (r,"pack.usesparse" , &r->settings.pack_use_sparse,1 ,0 ); +repo_cfg_bool (r,"core.multipackindex" , &r->settings.core_multi_pack_index,1 ,1 ); +repo_cfg_bool (r,"index.sparse" , &r->settings.sparse_index,0 ,0 );/* * The GIT_TEST_MULTI_PACK_INDEX variable is special in that
Signature · 1 line
2.35.1.582.g270d558070.dirty
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 127 lines · Teng Long
It's supported for a repo to use bitmap in both single-pack bitmap way or a multi-pack(MIDX) bitmap. Either of two bitmap kinds can exist in the repository, or both can be stored but let the config controls which kind of bitmap is used (like "core.multipackIndex", etc.). Because of this, sometimes the bitmap debug path is not obvious enough, for example, when executing:
$ git rev-list --test-bitmap HEAD fatal: failed to load bitmap indexesWhen the output look like this, it's unclear which kind(s) of .bitmap exist, and which were read. For example, it's possible a MIDX bitmap exists, but was not read (e.g., because core.multiPackIndex=false), among many other scenarios.
Therefore, we added some trace2 code so that when we read the bitmap we can be more clear about the decision path, such as whether it is working on MIDX or single-pack bitmap at present, or the related config is enabled or not. This may help with logging, user troubleshooting, and development debugging.
Here are some brief output (omitted some unrelated or repetitive rows and columns, using "..." instead of) examples on two scenarios when executing:
$ GIT_TRACE2_PERF=1 git rev-list --test-bitmap HEADScenario 1:
core.multipackIndex [false], MIDX bitmap exists [Y], single-pack bitmap exists [N]...
... | main | data | r1 | ... | config | core.multipackindex:false ... | d0 | main | region_enter | r1 | ... | pack-bitmap | label:open_bitmap ... | d0 | main | data | r1 | ... | bitmap | ..path:.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap ... | main | region_leave | r1 | ... | pack-bitmap | label:open_bitmap ... | main | error | | ... | | failed to load bitmap indexesfatal: failed to load bitmap indexes
... | d0 | main | exit | | ... | | code:128
...Scenario 2:
core.multipackIndex [false], MIDX bitmap exists [Y], single-pack bitmap exists [Y]... | d0 | main | region_enter | r0 | ... | pack-bitmap | label:open_bitmap ... | d0 | main | data | r0 | ... | bitmap | ..path:/home/tenglong.tl/test/dyrone_bitmap/.git/objects/pack/pack-e9eb18e6a423057f4424a762069e13804a75d01e.bitmap ... | main | region_leave | r0 | ... | pack-bitmap | label:open_bitmap Bitmap v1 test (1 entries loaded) Found bitmap for d864fefa87415d6cd289c72aa9ffd45b4a8ffd84. 64 bits / 11030517 checksum ... | main | region_enter | r0 | ... | progress | label:Verifying bitmap entries Verifying bitmap entries: 100% (3/3), done. ... | main | data | r0 | ... | progress | ..total_objects:3 ... | main | region_leave | r0 | ... | progress | label:Verifying bitmap entriesOK!
... | d0 | main | exit | | ... | | code:0
...Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 21 ++++++++++++++++----- repo-settings.c | 22 ++++++++++++---------- 2 files changed, 28 insertions(+), 15 deletions(-)pack-bitmap.c +16
-5@@ -312,9 +312,12 @@ char *pack_bitmap_filename(struct packed_git *p)static int open_midx_bitmap_1 (struct bitmap_index *bitmap_git,struct multi_pack_index *midx) { +int fd;struct stat st; +char *bitmap_name =midx_bitmap_filename (midx); -int fd =git_open (bitmap_name); +trace2_data_string ("midx" , the_repository,"path" , bitmap_name); + fd =git_open (bitmap_name);free (bitmap_name);@@ -346,8 +349,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (load_bitmap_header (bitmap_git) <0 ) goto cleanup; -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +error (_ ("midx and bitmap checksum don't match" )); goto cleanup; + }if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" ));@@ -374,6 +379,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitreturn -1 ; bitmap_name =pack_bitmap_filename (packfile); +trace2_data_string ("bitmap" , the_repository,"path" , bitmap_name); fd =git_open (bitmap_name);free (bitmap_name);@@ -509,11 +515,16 @@ static int open_midx_bitmap(struct repository *r,static int open_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { -assert (!bitmap_git->map); +int ret =0 ; +assert (!bitmap_git->map); +trace2_region_enter ("pack-bitmap" ,"open_bitmap" , r);if (!open_midx_bitmap (r, bitmap_git)) -return 0 ; -return open_pack_bitmap (r, bitmap_git); + goto done; + ret =open_pack_bitmap (r, bitmap_git); +done: +trace2_region_leave ("pack-bitmap" ,"open_bitmap" , r); +return ret; }struct bitmap_index *prepare_bitmap_git (struct repository *r)repo-settings.c +0
-0@@ -4,10 +4,12 @@#include "midx.h"
repo-settings.c line 14
Quoted 10 lines
repo-settings.c +3
-1static void repo_cfg_bool (struct repository *r,const char *key,int *dest, -int def) +int def,int trace) {if (repo_config_get_bool (r, key, dest)) *dest = def; +if (trace) +trace2_data_string ("config" , r, key, *dest ?"true" :"false" ); }
(I just sent a response to your V2 before I saw your V3, so I'll my response here so that it doesn't get lost.)
We should not be doing this. This would dump every repo-related boolean value on every command. I see that in V3 that you have a "trace" flag to control this. But again, this seems wrong here.
I already have a GIT_TRACE2_CONFIG_PARAMS and trace2.configparams that will dump "interesting" config values to the trace2 log. Just set one of them to a list of regex's. Look at the comment above trace2_cmd_list_config() in trace2.h for details.
We also have GIT_TRACE2_ENV_VARS and trace2.envvars that will dump the values of "interesting" env vars.
You can use these in your testing to log the config and env var values that you are interested in.
Quoted 20 lines
repo-settings.c +2
-6void prepare_repo_settings (struct repository *r)@@ -29,8 +31,8 @@ void prepare_repo_settings(struct repository *r) r->settings.fetch_negotiation_algorithm =FETCH_NEGOTIATION_CONSECUTIVE ;/* Booleans config or default, cascades to other settings */ -repo_cfg_bool (r,"feature.manyfiles" , &manyfiles,0 ); -repo_cfg_bool (r,"feature.experimental" , &experimental,0 ); +repo_cfg_bool (r,"feature.manyfiles" , &manyfiles,0 ,0 ); +repo_cfg_bool (r,"feature.experimental" , &experimental,0 ,0 );/* Defaults modified by feature.* */ if (experimental) {@@ -42,13 +44,13 @@ void prepare_repo_settings(struct repository *r) }/* Boolean config or default, does not cascade (simple) */ -repo_cfg_bool (r,"core.commitgraph" , &r->settings.core_commit_graph,1 ); -repo_cfg_bool (r,"commitgraph.readchangedpaths" , &r->settings.commit_graph_read_changed_paths,1 ); -repo_cfg_bool (r,"gc.writecommitgraph" , &r->settings.gc_write_commit_graph,1 ); -repo_cfg_bool (r,"fetch.writecommitgraph" , &r->settings.fetch_write_commit_graph,0 );
repo-settings.c void prepare_repo_settings(struct repository *r) · line 53
Quoted 10 lines
repo-settings.c +7
-3-repo_cfg_bool (r,"pack.usesparse" , &r->settings.pack_use_sparse,1 ); -repo_cfg_bool (r,"core.multipackindex" , &r->settings.core_multi_pack_index,1 ); -repo_cfg_bool (r,"index.sparse" , &r->settings.sparse_index,0 ); +repo_cfg_bool (r,"core.commitgraph" , &r->settings.core_commit_graph,1 ,0 ); +repo_cfg_bool (r,"commitgraph.readchangedpaths" , &r->settings.commit_graph_read_changed_paths,1 ,0 ); +repo_cfg_bool (r,"gc.writecommitgraph" , &r->settings.gc_write_commit_graph,1 ,0 ); +repo_cfg_bool (r,"fetch.writecommitgraph" , &r->settings.fetch_write_commit_graph,0 ,0 ); +repo_cfg_bool (r,"pack.usesparse" , &r->settings.pack_use_sparse,1 ,0 ); +repo_cfg_bool (r,"core.multipackindex" , &r->settings.core_multi_pack_index,1 ,1 ); +repo_cfg_bool (r,"index.sparse" , &r->settings.sparse_index,0 ,0 );
So if I'm reading this right, this will only print the value of "core.multipackindex".
You can get this without the changes here using
GIT_TRACE2_CONFIG_PARAMS="core.multipackindex,...any_other_values_of_interest..." before running the command (or use the config setting) before running your commandss.
Alternatively, just emit a "trace2_data_intmax()" on the one value your want to print here.
Jeff
Quoted 4 lines
repo-settings.c +0
-0/* * The GIT_TEST_MULTI_PACK_INDEX variable is special in that
Re: [PATCH v3 5/5] bitmap: add trace2 outputs during open "bitmap" file
Quoted 7 lines · Jeff Hostetler
(I just sent a response to your V2 before I saw your V3, so I'll my response here so that it doesn't get lost.)
We should not be doing this. This would dump every repo-related boolean value on every command. I see that in V3 that you have a "trace" flag to control this. But again, this seems wrong here.
I noticed that "flag" while merging this with another topic to "seen" and found it strange, too. It "allows" the caller to choose which one gets logged per variable, but in a very hard-coded way; we probably would have been much better off to have a table of what gets logged and have repo_cfg_*() calls consult it. Even without configurability, at least it would move the hard-coded choice of what gets logged from code to data.
But ...
Quoted 4 lines
I already have a
GIT_TRACE2_CONFIG_PARAMSand trace2.configparams that will dump "interesting" config values to the trace2 log. Just set one of them to a list of regex's. Look at the comment abovetrace2_cmd_list_config()intrace2.hfor details.
... that does sound like the right way to go.
Quoted 9 lines
...
So if I'm reading this right, this will only print the value of "core.multipackindex".You can get this without the changes here using
GIT_TRACE2_CONFIG_PARAMS="core.multipackindex,...any_other_values_of_interest..." before running the command (or use the config setting) before running your commandss.
Thanks.
v5 cover tr2: avoid to print "interesting" config repeatedly
In reply to Teng LongChanges since v4:
By Jeff Hostetler's suggestion, we can use GIT_TRACE2_CONFIG_PARAMS and trace2.configparams to dump "interesting" config values to the trace2 log. So I drop the commit [5/5] 'bitmap: add trace2 outputs during open "bitmap" file' in v4.
Then I found if a config key exist multiple config values in different scope the trace2 log will print the value repeatly but not only print the final active value, so a new commit "avoid to print "interesting" config repeatedly" is appended in v5.
Finally, compared with [3/5] in v4, commit "pack-bitmap.c: retrieve missing i18n translations" fix every place which was missing translation in pack-bitmap.c.
Thanks.
Teng Long (5):
pack-bitmap.c: continue looping when first MIDX bitmap is found pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: using error() instead of silently returning -1 pack-bitmap.c: retrieve missing i18n translations
tr2: avoid to print "interesting" config repeatedly
pack-bitmap.c | 105 ++++++++++++++++++++++++----------------------- trace2/tr2_cfg.c | 9 +++- 2 files changed, 61 insertions(+), 53 deletions(-)
Range-diff against v4:
1: d8dfe53dd4 < -: ---------- pack-bitmap.c: make warnings support i18N when opening bitmap -: ---------- > 1: 589e3f4075 pack-bitmap.c: continue looping when first MIDX bitmap is found -: ---------- > 2: b6b30047fc pack-bitmap.c: rename "idx_name" to "bitmap_name" 2: 917551f2b5 = 3: 82d4493a6e pack-bitmap.c: using error() instead of silently returning -1 3: 8735ae9979 < -: ---------- bitmap: add trace2 outputs during open "bitmap" file -: ---------- > 4: 065b7c9ccb pack-bitmap.c: retrieve missing i18n translations -: ---------- > 5: f3b87a33da tr2: avoid to print "interesting" config repeatedly
Signature · 1 line
2.35.1.582.gf3b87a33da
PATCH v5 1/5 pack-bitmap.c: continue looping when first MIDX bitmap is found
In "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to continue the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The reason for this is, there may exist more than one MIDX file in a repo. The "multi_pack_index" struct is actually designed as a singly linked list, and if a MIDX file has been already opened successfully, then the other MIDX files will be skipped and left with a warning "ignoring extra bitmap file." to the output.
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -494,15 +494,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.1.582.gf3b87a33da
PATCH v5 2/5 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index, let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -313,10 +313,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -368,14 +368,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.1.582.gf3b87a33da
PATCH v5 3/5 pack-bitmap.c: using error() instead of silently returning -1
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
pack-bitmap.c +5 -5
@@ -323,7 +323,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot stat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -361,7 +361,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error (_ ("cannot open midx bitmap file" )); }static int open_pack_bitmap_1 (struct bitmap_index *bitmap_git,struct packed_git *packfile)
@@ -382,7 +382,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot stat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -394,7 +394,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {close (fd); -return -1 ; +return error (_ ("packfile is invalid" )); } bitmap_git->pack = packfile;
@@ -409,7 +409,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_git->map_size =0 ; bitmap_git->map_pos =0 ; bitmap_git->pack =NULL ; -return -1 ; +return error (_ ("bitmap header is invalid" )); }return 0 ;
Signature · 1 line
2.35.1.582.gf3b87a33da
Re: [PATCH v5 3/5] pack-bitmap.c: using error() instead of silently returning -1
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 326
Quoted 4 lines · Teng Long
pack-bitmap.c +1
-1if (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot stat bitmap file" ));
"stat" -> "fstat" perhaps?
Not a new problem, but before this hunk, we have
fd = git_open(...); if (fd < 0) return -1;
where it probably is legitimate to ignore ENOENT silently. In practice, it may be OK to treat a file that exists but cannot be opened for whatever reason, but I do not think it would hurt to report such a rare anomaly with a warning, e.g.
if (fd < 0) {
if (errno != ENOENT)
warning("'%s' cannot open '%s'",
strerror(errno), bitmap_name);
free(bitmap_name);
return -1;
}or something like that perhaps.
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 364
Quoted 6 lines
pack-bitmap.c +1
-1@@ -361,7 +361,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error (_ ("cannot open midx bitmap file" ));
This is not exactly "cannot open". We opened the file, and found there was something we do not like in it. If we failed to find midx lacks revindex chunk, we already would have given a warning, and the error is not just misleading but is redundant. load_bitmap_header() also gives an error() on its own.
I think this patch aims for a good end result, but needs more work. At least, checksum mismatch that goes to cleanup also needs to issue its own error() and then we can remove this "cannot open" at the end.
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 385
Quoted 6 lines
pack-bitmap.c +1
-1@@ -382,7 +382,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot stat bitmap file" ));
"stat" -> "fstat"
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 397
Quoted 6 lines
pack-bitmap.c +1
-1@@ -394,7 +394,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {close (fd); -return -1 ; +return error (_ ("packfile is invalid" ));
Same "sometimes redundant" comment applies here, but not due to this part of the code but due to the helpers called from is_pack_valid().
Namely, packfile.c::open_packed_git_1() is mostly chatty, but is silent upon "unable to open" and "unable to fstat" (which I think is safe to make chatty as well---we do not want to special case ENOENT in open_packed_git(), so "cannot open because it is not there" is an error).
And then, this "invalid" will become redundant and fuzzier message than what is_pack_valid() would have already given, so you can leave it to silently return -1 here.
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 412
Quoted 6 lines
pack-bitmap.c +1
-1@@ -409,7 +409,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_git->map_size =0 ; bitmap_git->map_pos =0 ; bitmap_git->pack =NULL ; -return -1 ; +return error (_ ("bitmap header is invalid" ));
Having shown how to analize these kind of things twice in the above, I would not bother checking myself, but you should look at the existing error messages from load_bitmap_header() and see if this extra message should really be here. I suspect not.
Quoted 3 lines
pack-bitmap.c +0
-0}return 0 ;
Re: [PATCH v5 3/5] pack-bitmap.c: using error() instead of silently returning -1
Quoted 1 line · Junio C Hamano
"stat" -> "fstat" perhaps?
Yes.
Quoted 19 lines
Not a new problem, but before this hunk, we have
fd = git_open(...); if (fd < 0) return -1;where it probably is legitimate to ignore ENOENT silently. In practice, it may be OK to treat a file that exists but cannot be opened for whatever reason, but I do not think it would hurt to report such a rare anomaly with a warning, e.g.
if (fd < 0) { if (errno != ENOENT) warning("'%s' cannot open '%s'", strerror(errno), bitmap_name); free(bitmap_name); return -1; }
Quoted 1 line
or something like that perhaps.
Yes.
It's more accurate here with your suggestion. At the same time I found there actually exists many similar place like "ignore ENOENT silently" in repo. And I think it's not worth to impove them right now in this patch, if you want to do that it could in another pathset and please tell me.
Quoted 13 lines
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 364
Quoted 6 lines
pack-bitmap.c +1
-1@@ -361,7 +361,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, bitmap_git->map_pos =0 ; bitmap_git->map =NULL ; bitmap_git->midx =NULL ; -return -1 ; +return error (_ ("cannot open midx bitmap file" ));This is not exactly "cannot open". We opened the file, and found there was something we do not like in it. If we failed to find midx lacks revindex chunk, we already would have given a warning, and the error is not just misleading but is redundant.
load_bitmap_header()also gives anerror()on its own.
Quoted 3 lines
I think this patch aims for a good end result, but needs more work. At least, checksum mismatch that goes to cleanup also needs to issue its own
error()and then we can remove this "cannot open" at the end.
Yes, It's detailed here and I missed it, I will fix this in next
patch: return error() when checksum doesn't match and let "cleanup" return "-1" but not an inaccurate error().
Quoted 1 line
"stat" -> "fstat"
Yes.
Quoted 9 lines
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 397
Quoted 6 lines
pack-bitmap.c +1
-1@@ -394,7 +394,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {close (fd); -return -1 ; +return error (_ ("packfile is invalid" ));Same "sometimes redundant" comment applies here, but not due to this part of the code but due to the helpers called from
is_pack_valid().
Let me try to know about it, the "helpers" means the place where invoke is_pack_valid() like here. In fact, in is_pack_valid() they already return the errors in various scenarios, so it would be no need to return another error.
Quoted 5 lines
Namely,
packfile.c::open_packed_git_1()is mostly chatty, but is silent upon "unable to open" and "unable to fstat" (which I think is safe to make chatty as well---we do not want to special case ENOENT inopen_packed_git(), so "cannot open because it is not there" is an error).
"chatty" means the code that regularly gives information about its internal operations. Did I understand it right because I'm actually firstly know this description.
"cannot open because it is not there" is an error, but I think it will also could be a BUG, actually I'm not very sure for clarify the difference bwtween the use of the two, but I will look into it to dig something out.
Quoted 3 lines
And then, this "invalid" will become redundant and fuzzier message than what
is_pack_valid()would have already given, so you can leave it to silently return -1 here.
Clear now.
Thanks.
Re: [PATCH v5 3/5] pack-bitmap.c: using error() instead of silently returning -1
Quoted 5 lines · Teng Long
It's more accurate here with your suggestion. At the same time I found there actually exists many similar place like "ignore ENOENT silently" in repo. And I think it's not worth to impove them right now in this patch, if you want to do that it could in another pathset and please tell me.
It is sufficien to just mark it as #leftoverbits to find and fix places where the code means to ignore only a missing optional file but ignores all other errors it gets from open/fopen instead.
Quoted 10 lines
Quoted 9 lines
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 397
Quoted 6 lines
pack-bitmap.c +1
-1@@ -394,7 +394,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (!is_pack_valid (packfile)) {close (fd); -return -1 ; +return error (_ ("packfile is invalid" ));Same "sometimes redundant" comment applies here, but not due to this part of the code but due to the helpers called from
is_pack_valid().
Quoted 4 lines
Let me try to know about it, the "helpers" means the place where invoke
is_pack_valid()like here. In fact, inis_pack_valid()they already return the errors in various scenarios, so it would be no need to return another error.
Yup. is_pack_valid() calls open_packed_git() and the helper functions that are called from that call chain are full of calls to error() that tell specifically what exactly went wrong. But ...
Quoted 5 lines
Quoted 5 lines
Namely,
packfile.c::open_packed_git_1()is mostly chatty, but is silent upon "unable to open" and "unable to fstat" (which I think is safe to make chatty as well---we do not want to special case ENOENT inopen_packed_git(), so "cannot open because it is not there" is an error).
... the error coverage is not complete. There are some (rare) code paths that silently "return -1", not "return error(_("..."))". They should be updated to say something; otherwise we will silently fail in these "rare" codepaths.
Quoted 4 lines
"cannot open because it is not there" is an error, but I think it will also could be a BUG, actually I'm not very sure for clarify the difference bwtween the use of the two, but I will look into it to dig something out.
The code may have many reasons to believe that a file should exist there and try to open it, but it may find the file missing, but I would suspect that it is never a BUG. You may have run stat() on the path earlier and you know it existed, but by the time you try to open it, some other process may have removed it. You may have found the .idx file and expect that the corresponding .pack file to exist, but the .pack file may be missing. You may have just written a file and you expect to be able to open it, but some other process may have removed it already, or you may have run out of file descriptors and cannot open it. These are all runtime errors that deserve to be reported via error() or die(), but never BUG().
Thanks.
PATCH v5 4/5 pack-bitmap.c: retrieve missing i18n translations
In reply to Teng LongIn pack-bitmap.c, some printed texts are translated,some are not. Let's support the translations of the bitmap related output.
--- pack-bitmap.c | 76 +++++++++++++++++++++++++-------------------------- 1 file changed, 38 insertions(+), 38 deletions(-)
pack-bitmap.c +38 -38
@@ -138,7 +138,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("Failed to load bitmap index (corrupted?)" ); +error (_ ("Failed to load bitmap index (corrupted?)" ));ewah_pool_free (b);return NULL ; }
@@ -160,14 +160,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("Corrupted bitmap index (too small)" ); +return error (_ ("Corrupted bitmap index (too small)" ));if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("Corrupted bitmap index file (wrong header)" ); +return error (_ ("Corrupted bitmap index file (wrong header)" )); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("Unsupported version for bitmap index file (%d)" , index->version); +return error (_ ("Unsupported version for bitmap index file (%d)" ), index->version);/* Parse known bitmap format options */ {
@@ -176,12 +176,12 @@ static int load_bitmap_header(struct bitmap_index *index)unsigned char *index_end = index->map + index->map_size - the_hash_algo->rawsz;if ((flags &BITMAP_OPT_FULL_DAG ) ==0 ) -return error ("Unsupported options for bitmap index file " -"(Git requires BITMAP_OPT_FULL_DAG)" ); +return error (_ ("Unsupported options for bitmap index file " +"(Git requires BITMAP_OPT_FULL_DAG)" ));if (flags &BITMAP_OPT_HASH_CACHE ) {if (cache_size > index_end - index->map - header_size) -return error ("corrupted bitmap index file (too short to fit hash cache)" ); +return error (_ ("corrupted bitmap index file (too short to fit hash cache)" )); index->hashes = (void *)(index_end - cache_size); index_end -= cache_size; }
@@ -215,7 +215,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("Duplicate entry in bitmap index: %s" ,oid_to_hex (oid)); +error (_ ("Duplicate entry in bitmap index: %s" ),oid_to_hex (oid));return NULL ; }
@@ -259,14 +259,14 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)struct object_id oid;if (index->map_size - index->map_pos <6 ) -return error ("corrupt ewah bitmap: truncated header for entry %d" , i); +return error (_ ("corrupt ewah bitmap: truncated header for entry %d)" ), i); commit_idx_pos =read_be32 (index->map, &index->map_pos); xor_offset =read_u8 (index->map, &index->map_pos); flags =read_u8 (index->map, &index->map_pos);if (nth_bitmap_object_oid (index, &oid, commit_idx_pos) <0 ) -return error ("corrupt ewah bitmap: commit index %u out of range" , +return error (_ ("corrupt ewah bitmap: commit index %u out of range" ), (unsigned )commit_idx_pos); bitmap =read_bitmap_1 (index);
@@ -274,13 +274,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("Corrupted bitmap pack index" ); +return error (_ ("Corrupted bitmap pack index" ));if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("Invalid XOR offset in bitmap pack index" ); +return error (_ ("Invalid XOR offset in bitmap pack index" )); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -305,7 +305,7 @@ char *pack_bitmap_filename(struct packed_git *p)size_t len;if (!strip_suffix (p->pack_name,".pack" , &len)) -BUG ("pack_name does not end in .pack" ); +BUG (_ ("pack_name does not end in .pack" ));return xstrfmt ("%.*s.bitmap" , (int )len, p->pack_name); }
@@ -330,7 +330,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning (_ ("ignoring extra bitmap file: %s" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +387,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: %s" ), packfile->pack_name);close (fd);return -1 ; }
@@ -819,7 +819,7 @@ static struct bitmap *find_objects(struct bitmap_index *bitmap_git, revs->include_check_data = &incdata;if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); show_data.bitmap_git = bitmap_git; show_data.base = base;
@@ -1134,7 +1134,7 @@ static void filter_bitmap_tree_depth(struct bitmap_index *bitmap_git, unsigned long limit) {if (limit) -BUG ("filter_bitmap_tree_depth given non-zero limit" ); +BUG (_ ("filter_bitmap_tree_depth given non-zero limit" ));filter_bitmap_exclude_type (bitmap_git, tip_objects, to_filter,OBJ_TREE );
@@ -1148,7 +1148,7 @@ static void filter_bitmap_object_type(struct bitmap_index *bitmap_git, enum object_type object_type) {if (object_type <OBJ_COMMIT || object_type >OBJ_TAG ) -BUG ("filter_bitmap_object_type given invalid object" ); +BUG (_ ("filter_bitmap_object_type given invalid object" ));if (object_type !=OBJ_TAG )filter_bitmap_exclude_type (bitmap_git, tip_objects, to_filter,OBJ_TAG );
@@ -1304,14 +1304,14 @@ struct bitmap_index *prepare_bitmap_walk(struct rev_info *revs, revs->ignore_missing_links =0 ;if (haves_bitmap ==NULL ) -BUG ("failed to perform bitmap walk" ); +BUG (_ ("failed to perform bitmap walk" )); } wants_bitmap =find_objects (bitmap_git, revs, wants, haves_bitmap, filter);if (!wants_bitmap) -BUG ("failed to perform bitmap walk" ); +BUG (_ ("failed to perform bitmap walk" ));if (haves_bitmap)bitmap_and_not (wants_bitmap, haves_bitmap);
@@ -1432,7 +1432,7 @@ uint32_t midx_preferred_pack(struct bitmap_index *bitmap_git) {struct multi_pack_index *m = bitmap_git->midx;if (!m) -BUG ("midx_preferred_pack: requires non-empty MIDX" ); +BUG (_ ("midx_preferred_pack: requires non-empty MIDX" ));return nth_midxed_pack_int_id (m,pack_pos_to_midx (bitmap_git->midx,0 )); }
@@ -1629,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object %s not found in type bitmaps" , +die (_ ("object %s not found in type bitmaps" ),oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object %s does not have a unique type" , +die (_ ("object %s does not have a unique type" ),oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object %s: real type %s, expected: %s" , +die (_ ("object %s: real type %s, expected: %s" ),oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1651,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&object->oid)); +die (_ ("Object not in bitmap: %s\n" ),oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1666,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&commit->object.oid)); +die (_ ("Object not in bitmap: %s\n" ),oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1683,26 +1683,26 @@ void test_bitmap_walk(struct rev_info *revs)struct ewah_bitmap *bm;if (!(bitmap_git =prepare_bitmap_git (revs->repo))) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" ));if (revs->pending.nr !=1 ) -die ("you must specify exactly one commit to test" ); +die (_ ("you must specify exactly one commit to test" )); -fprintf (stderr,"Bitmap v%d test (%d entries loaded)\n" , +fprintf (stderr,_ ("Bitmap v%d test (%d entries loaded)\n" ), bitmap_git->version, bitmap_git->entry_count); root = revs->pending.objects[0 ].item; bm =bitmap_for_commit (bitmap_git, (struct commit *)root);if (bm) { -fprintf (stderr,"Found bitmap for %s. %d bits / %08x checksum\n" , +fprintf (stderr,_ ("Found bitmap for %s. %d bits / %08x checksum\n" ),oid_to_hex (&root->oid), (int )bm->bit_size,ewah_checksum (bm)); result =ewah_to_bitmap (bm); }if (result ==NULL ) -die ("Commit %s doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die (_ ("Commit %s doesn't have an indexed bitmap" ),oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1711,7 +1711,7 @@ void test_bitmap_walk(struct rev_info *revs) result_popcnt =bitmap_popcount (result);if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); tdata.bitmap_git = bitmap_git; tdata.base =bitmap_new ();
@@ -1719,7 +1719,7 @@ void test_bitmap_walk(struct rev_info *revs) tdata.trees =ewah_to_bitmap (bitmap_git->trees); tdata.blobs =ewah_to_bitmap (bitmap_git->blobs); tdata.tags =ewah_to_bitmap (bitmap_git->tags); - tdata.prg =start_progress ("Verifying bitmap entries" , result_popcnt); + tdata.prg =start_progress (_ ("Verifying bitmap entries" ), result_popcnt); tdata.seen =0 ;traverse_commit_list (revs, &test_show_commit, &test_show_object, &tdata);
@@ -1727,9 +1727,9 @@ void test_bitmap_walk(struct rev_info *revs)stop_progress (&tdata.prg);if (bitmap_equals (result, tdata.base)) -fprintf (stderr,"OK!\n" ); +fprintf (stderr,_ ("OK!\n" ));else -die ("mismatch in bitmap results" ); +die (_ ("mismatch in bitmap results" ));bitmap_free (result);bitmap_free (tdata.base);
@@ -1747,7 +1747,7 @@ int test_bitmap_commits(struct repository *r) MAYBE_UNUSED void *value;if (!bitmap_git) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" )); kh_foreach(bitmap_git->bitmaps, oid, value, { printf("%s\n", oid_to_hex(&oid));
@@ -1825,8 +1825,8 @@ uint32_t *create_bitmap_mapping(struct bitmap_index *bitmap_git,if (!bitmap_is_midx (bitmap_git))load_reverse_index (bitmap_git);else if (load_midx_revindex (bitmap_git->midx) <0 ) -BUG ("rebuild_existing_bitmaps: missing required rev-cache " -"extension" ); +BUG (_ ("rebuild_existing_bitmaps: missing required rev-cache " +"extension" )); num_objects =bitmap_num_objects (bitmap_git);CALLOC_ARRAY (reposition, num_objects);
Signature · 1 line
2.35.1.582.gf3b87a33da
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
Quoted 2 lines · Teng Long
In
pack-bitmap.c, some printed texts are translated,some are not. Let's support the translations of the bitmap related output.
Usually we don't go for cleanup-while-at-it, but in this case we're marking messages that don't conform to our CodingGudielines for translation, mostly because they're error messages that start with an upper-case letter.
So I think we should fix those issues first, to avoid double-work for translators (well, a bit less, since they're the translation memory, but it's quite a bit of churn...).
pack-bitmap.c static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) · line 141
Quoted 2 lines
pack-bitmap.c +1
-1-error ("Failed to load bitmap index (corrupted?)" ); +error (_ ("Failed to load bitmap index (corrupted?)" ));
e.g. here.
pack-bitmap.c static int load_bitmap_header(struct bitmap_index *index) · line 163
Quoted 2 lines
pack-bitmap.c +1
-1-return error ("Corrupted bitmap index (too small)" ); +return error (_ ("Corrupted bitmap index (too small)" ));
..and here, etc. etc.
pack-bitmap.c static int load_bitmap_header(struct bitmap_index *index) · line 170
Quoted 2 lines
pack-bitmap.c +1
-1-return error ("Unsupported version for bitmap index file (%d)" , index->version); +return error (_ ("Unsupported version for bitmap index file (%d)" ), index->version);
Let's say "unsupported version '%d' for ..." instead?
pack-bitmap.c static int load_bitmap_header(struct bitmap_index *index) · line 180
Quoted 4 lines
pack-bitmap.c +2
-2-return error ("Unsupported options for bitmap index file " -"(Git requires BITMAP_OPT_FULL_DAG)" ); +return error (_ ("Unsupported options for bitmap index file " +"(Git requires BITMAP_OPT_FULL_DAG)" ));
I'm not sure, but shouldn't this be a BUG()?
pack-bitmap.c static struct stored_bitmap *store_bitmap(struct bitmap_index *index, · line 218
Quoted 2 lines
pack-bitmap.c +1
-1-error ("Duplicate entry in bitmap index: %s" ,oid_to_hex (oid)); +error (_ ("Duplicate entry in bitmap index: %s" ),oid_to_hex (oid));
Ditto upper-case, but add '%s' while at it.
pack-bitmap.c char *pack_bitmap_filename(struct packed_git *p) · line 308
Quoted 3 lines
pack-bitmap.c +1
-1if (!strip_suffix (p->pack_name,".pack" , &len)) -BUG ("pack_name does not end in .pack" ); +BUG (_ ("pack_name does not end in .pack" ));
Do not translate BUG() messages.
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 333
Quoted 2 lines
pack-bitmap.c +1
-1-warning ("ignoring extra bitmap file: %s" , buf.buf); +warning (_ ("ignoring extra bitmap file: %s" ), buf.buf);
Quote the name.
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 390
Quoted 2 lines
pack-bitmap.c +1
-1-warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: %s" ), packfile->pack_name);
ditto.
pack-bitmap.c static struct bitmap *find_objects(struct bitmap_index *bitmap_git, · line 822
Quoted 3 lines
pack-bitmap.c +1
-1if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" ));
Looks good, but aside from this we should really have a tree-wide xprepare_revision_walk() or something, we have this copy/pasted all over the place...
pack-bitmap.c static void filter_bitmap_tree_depth(struct bitmap_index *bitmap_git, · line 1137
Quoted 2 lines
pack-bitmap.c +1
-1-BUG ("filter_bitmap_tree_depth given non-zero limit" ); +BUG (_ ("filter_bitmap_tree_depth given non-zero limit" ));
Ditto BUG.
Quoted 13 lines
pack-bitmap.c +1
-1filter_bitmap_exclude_type (bitmap_git, tip_objects, to_filter,OBJ_TREE );@@ -1148,7 +1148,7 @@ static void filter_bitmap_object_type(struct bitmap_index *bitmap_git, enum object_type object_type) {if (object_type <OBJ_COMMIT || object_type >OBJ_TAG ) -BUG ("filter_bitmap_object_type given invalid object" ); +BUG (_ ("filter_bitmap_object_type given invalid object" ));if (object_type !=OBJ_TAG )filter_bitmap_exclude_type (bitmap_git, tip_objects, to_filter,OBJ_TAG );
pack-bitmap.c struct bitmap_index *prepare_bitmap_walk(struct rev_info *revs, · line 1308
Quoted 6 lines
pack-bitmap.c +1
-1@@ -1304,14 +1304,14 @@ struct bitmap_index *prepare_bitmap_walk(struct rev_info *revs, revs->ignore_missing_links =0 ;if (haves_bitmap ==NULL ) -BUG ("failed to perform bitmap walk" ); +BUG (_ ("failed to perform bitmap walk" )); }
etc. etc.
Quoted 11 lines
pack-bitmap.c +1
-1wants_bitmap =find_objects (bitmap_git, revs, wants, haves_bitmap, filter);if (!wants_bitmap) -BUG ("failed to perform bitmap walk" ); +BUG (_ ("failed to perform bitmap walk" ));if (haves_bitmap)bitmap_and_not (wants_bitmap, haves_bitmap);
pack-bitmap.c uint32_t midx_preferred_pack(struct bitmap_index *bitmap_git) · line 1435
Quoted 5 lines
pack-bitmap.c +1
-1@@ -1432,7 +1432,7 @@ uint32_t midx_preferred_pack(struct bitmap_index *bitmap_git) {struct multi_pack_index *m = bitmap_git->midx;if (!m) -BUG ("midx_preferred_pack: requires non-empty MIDX" ); +BUG (_ ("midx_preferred_pack: requires non-empty MIDX" ));
etc. etc.
Quoted 12 lines
pack-bitmap.c +1
-1return nth_midxed_pack_int_id (m,pack_pos_to_midx (bitmap_git->midx,0 )); }@@ -1629,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object %s not found in type bitmaps" , +die (_ ("object %s not found in type bitmaps" ),oid_to_hex (&obj->oid));if (bitmaps_nr >1 )
pack-bitmap.c static void test_bitmap_type(struct bitmap_test_data *tdata, · line 1643
Quoted 10 lines
pack-bitmap.c +2
-2-die ("object %s does not have a unique type" , +die (_ ("object %s does not have a unique type" ),oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object %s: real type %s, expected: %s" , +die (_ ("object %s: real type %s, expected: %s" ),oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
quote %s for these.
pack-bitmap.c static void test_show_object(struct object *object, const char *name, · line 1654
Quoted 6 lines
pack-bitmap.c +1
-1@@ -1651,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&object->oid)); +die (_ ("Object not in bitmap: %s\n" ),oid_to_hex (&object->oid));
Lose the \n here, in addition to lower-case & quote %s.
pack-bitmap.c static void test_show_commit(struct commit *commit, void *data) · line 1669
Quoted 9 lines
pack-bitmap.c +1
-1test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);@@ -1666,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&commit->object.oid)); +die (_ ("Object not in bitmap: %s\n" ),oid_to_hex (&commit->object.oid));
Ditto.
Quoted 45 lines
pack-bitmap.c +6
-6test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);@@ -1683,26 +1683,26 @@ void test_bitmap_walk(struct rev_info *revs)struct ewah_bitmap *bm;if (!(bitmap_git =prepare_bitmap_git (revs->repo))) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" ));if (revs->pending.nr !=1 ) -die ("you must specify exactly one commit to test" ); +die (_ ("you must specify exactly one commit to test" )); -fprintf (stderr,"Bitmap v%d test (%d entries loaded)\n" , +fprintf (stderr,_ ("Bitmap v%d test (%d entries loaded)\n" ), bitmap_git->version, bitmap_git->entry_count); root = revs->pending.objects[0 ].item; bm =bitmap_for_commit (bitmap_git, (struct commit *)root);if (bm) { -fprintf (stderr,"Found bitmap for %s. %d bits / %08x checksum\n" , +fprintf (stderr,_ ("Found bitmap for %s. %d bits / %08x checksum\n" ),oid_to_hex (&root->oid), (int )bm->bit_size,ewah_checksum (bm)); result =ewah_to_bitmap (bm); }if (result ==NULL ) -die ("Commit %s doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die (_ ("Commit %s doesn't have an indexed bitmap" ),oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;@@ -1711,7 +1711,7 @@ void test_bitmap_walk(struct rev_info *revs) result_popcnt =bitmap_popcount (result);if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); tdata.bitmap_git = bitmap_git; tdata.base =bitmap_new ();
pack-bitmap.c void test_bitmap_walk(struct rev_info *revs) · line 1722
Quoted 5 lines
pack-bitmap.c +1
-1@@ -1719,7 +1719,7 @@ void test_bitmap_walk(struct rev_info *revs) tdata.trees =ewah_to_bitmap (bitmap_git->trees); tdata.blobs =ewah_to_bitmap (bitmap_git->blobs); tdata.tags =ewah_to_bitmap (bitmap_git->tags); - tdata.prg =start_progress ("Verifying bitmap entries" , result_popcnt); + tdata.prg =start_progress (_ ("Verifying bitmap entries" ), result_popcnt);
Good catch!
pack-bitmap.c void test_bitmap_walk(struct rev_info *revs) · line 1730
Quoted 9 lines
pack-bitmap.c +1
-1tdata.seen =0 ;traverse_commit_list (revs, &test_show_commit, &test_show_object, &tdata);@@ -1727,9 +1727,9 @@ void test_bitmap_walk(struct rev_info *revs)stop_progress (&tdata.prg);if (bitmap_equals (result, tdata.base)) -fprintf (stderr,"OK!\n" ); +fprintf (stderr,_ ("OK!\n" ));
Ditto don't include \n.
Quoted 16 lines
pack-bitmap.c +2
-2else -die ("mismatch in bitmap results" ); +die (_ ("mismatch in bitmap results" ));bitmap_free (result);bitmap_free (tdata.base);@@ -1747,7 +1747,7 @@ int test_bitmap_commits(struct repository *r) MAYBE_UNUSED void *value;if (!bitmap_git) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" )); kh_foreach(bitmap_git->bitmaps, oid, value, { printf("%s\n", oid_to_hex(&oid));
pack-bitmap.c uint32_t *create_bitmap_mapping(struct bitmap_index *bitmap_git, · line 1830
Quoted 8 lines
pack-bitmap.c +2
-2@@ -1825,8 +1825,8 @@ uint32_t *create_bitmap_mapping(struct bitmap_index *bitmap_git,if (!bitmap_is_midx (bitmap_git))load_reverse_index (bitmap_git);else if (load_midx_revindex (bitmap_git->midx) <0 ) -BUG ("rebuild_existing_bitmaps: missing required rev-cache " -"extension" ); +BUG (_ ("rebuild_existing_bitmaps: missing required rev-cache " +"extension" ));
Ditto don't translate BUG().
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
Quoted 4 lines · Ævar Arnfjörð Bjarmason
pack-bitmap.c void test_bitmap_walk(struct rev_info *revs) · line 1730
Quoted 3 lines · Teng Long
if (
bitmap_equals(result, tdata.base))pack-bitmap.c +1
-1-fprintf (stderr,"OK!\n" ); +fprintf (stderr,_ ("OK!\n" ));
As a minor additional bit of help, you can use fprintf_ln() from strbuf.h which will automatically output the trailing "\n":
frpintf_ln(stderr, _("OK!"));(Aside: Use of fprintf() here is a bit odd since there are no formatting directives in the argument, thus fputs() would have been a better choice, but that's a cleanup for another day and a different patch series, probably.)
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
Quoted 7 lines · Eric Sunshine
Quoted 4 lines · Teng Long
pack-bitmap.c void test_bitmap_walk(struct rev_info *revs) · line 1730
Quoted 3 lines
if (
bitmap_equals(result, tdata.base))pack-bitmap.c +1
-1-fprintf (stderr,"OK!\n" ); +fprintf (stderr,_ ("OK!\n" ));As a minor additional bit of help, you can use
fprintf_ln()fromstrbuf.hwhich will automatically output the trailing "\n":
Yes, agree.
Quoted 4 lines
(Aside: Use of
fprintf()here is a bit odd since there are no formatting directives in the argument, thusfputs()would have been a better choice, but that's a cleanup for another day and a different patch series, probably.)
*nod*, by the way, I think I will remove the translation on "OK!", currently I think it doesnt mean much.
Thanks.
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
In reply to Ævar Arnfjörð BjarmasonQuoted 8 lines
Usually we don't go for cleanup-while-at-it, but in this case we're marking messages that don't conform to our CodingGudielines for translation, mostly because they're error messages that start with an upper-case letter.
So I think we should fix those issues first, to avoid double-work for translators (well, a bit less, since they're the translation memory, but it's quite a bit of churn...).
Yes. I think it's need to make a cleanup commit first.
Quoted 6 lines
Lose the \n here, in addition to lower-case & quote %s.
...
Ditto don't include \n.
...
Ditto don't translateBUG().
..
etc.Will fix.
Thanks.
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
In reply to Teng LongQuoted 1 line · Teng Long
Subject: Re: [PATCH v5 4/5]
pack-bitmap.c: retrieve missing i18n translations
The verb "retrieve" is puzzling.
Quoted 1 line
In
pack-bitmap.c, some printed texts are translated,some are not.
"," -> ", ".
pack-bitmap.c static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) · line 141
Quoted 5 lines
Let's support the translations of the bitmap related output.
pack-bitmap.c +1
-1if (bitmap_size <0 ) { -error ("Failed to load bitmap index (corrupted?)" ); +error (_ ("Failed to load bitmap index (corrupted?)" ));
If we were to do this, to avoid burdening translators with double work, we probably would want to fix the "C" locale version of the string, either as a preliminary clean-up before this step, or as part of this step. From Documentation/CodingGuidelines:
Error Messages
- Do not end error messages with a full stop.
- Do not capitalize the first word, only because it is the first word in the message ("unable to open %s", not "Unable to open %s"). But "SHA-3 not supported" is fine, because the reason the first word is capitalized is not because it is at the beginning of the sentence, but because the word would be spelled in capital letters even when it appeared in the middle of the sentence.
- Say what the error is first ("cannot open %s", not "%s: cannot open")
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
Quoted 1 line · Junio C Hamano
The verb "retrieve" is puzzling.
I use "retrieve" because I think they should be there but actually missing. But If it's not appropriate here I will change another word like "add".
Quoted 1 line
"," -> ", ".
Yes.
Quoted 4 lines
If we were to do this, to avoid burdening translators with double work, we probably would want to fix the "C" locale version of the string, either as a preliminary clean-up before this step, or as part of this step. From
Documentation/CodingGuidelines:
Yes.
Does git have any NOT "C" Locale string?
Another doublt is I found something like in:
File: ./contrib/completion/git-completion.bash
923 LANG=C LC_ALL=C git merge -s help 2>&1
I think LC_ALL=C will override LANG=C in these cases, so I think LC_ALL=C git merge -s help 2>&1 is OK here.
Quoted 4 lines
Error Messages
- Do not end error messages with a full stop.
Quoted 8 lines
- Do not capitalize the first word, only because it is the first word in the message ("unable to open %s", not "Unable to open %s"). But "SHA-3 not supported" is fine, because the reason the first word is capitalized is not because it is at the beginning of the sentence, but because the word would be spelled in capital letters even when it appeared in the middle of the sentence.
- Say what the error is first ("cannot open %s", not "%s: cannot open")
Helpful, fill in blind spots.
Actually, in tr2 some text are printed as capitalized the first word. Is this entry appropriate for this situation? Whether we should unify the both?
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
Quoted 7 lines · Teng Long
Quoted 1 line · Junio C Hamano
The verb "retrieve" is puzzling.
I use "retrieve" because I think they should be there but actually missing. But If it's not appropriate here I will change another word like "add".
To retrieve is to get/bring something back and regaining possession of, which implies that the thing existed somewhere already but at a wrong/different place, and the only thing you are doing is to move it to the right place, but in this case, the translations did not exist. The patch is marking more strings for translation. And the act of marking them for translation will cause i18n/l10n folks to translate these strings, which will (finally) allow _("...") to retrieve the translated strings at runtime.
So "retrieve" is indeed involved somewhere in the process, but using the verb skips a few steps.
Subject: [PATCH 4/5] pack-bitmap.c: mark more strings for translations
perhaps?
Quoted 8 lines
Quoted 4 lines
If we were to do this, to avoid burdening translators with double work, we probably would want to fix the "C" locale version of the string, either as a preliminary clean-up before this step, or as part of this step. From
Documentation/CodingGuidelines:Yes.
Does git have any NOT "C" Locale string?
Sorry, but I am not sure what you are asking. What I meant is that a hunk like this from the patch in discussion:
if (bitmap_size < 0) {
- error("Failed to load bitmap index (corrupted?)");
+ error(_("Failed to load bitmap index (corrupted?)"));
ewah_pool_free(b);
return NULL;
}
makes translators to first translate the above string, but we will fix the "C" locale version (that is, the string inside _() that is used as the key to the .po database when retrieving the translated version) to follow our error message formatting convention to read something like
error(_("failed to load bitmap index (corrupted?)"));or even
error(_("failed to load bitmap index (corrupted?): '%s'"),
filename);And the translators have to redo the work. If a preliminary patch fixed these up before bothering translators with more strings to translate, they do not need to translate the current, known to be faulty, version of messages.
Quoted 7 lines
Another doublt is I found something like in:
File: ./contrib/completion/git-completion.bash
923 LANG=C LC_ALL=C git merge -s help 2>&1I think LC_ALL=C will override LANG=C in these cases, so I think
LC_ALL=C git merge -s help 2>&1is OK here.
In practice, yes, but the code is following the convention to reduce common confusion caused by leaving some lower precedence but common environment variables (i.e. LANG) as their original values.
Does the line in the completion script have anything to do with [PATCH 4/5], or is this merely your curiosity? Avoid mixing in unrelated things into the topic, which will only make the review cycle unnecessarily longer, but raise a separate discussion if you have to.
Thanks.
Re: [PATCH v5 4/5] pack-bitmap.c: retrieve missing i18n translations
Quoted 9 lines
To retrieve is to get/bring something back and regaining possession of, which implies that the thing existed somewhere already but at a wrong/different place, and the only thing you are doing is to move it to the right place, but in this case, the translations did not exist. The patch is marking more strings for translation. And the act of marking them for translation will cause i18n/l10n folks to translate these strings, which will (finally) allow
_("...")to retrieve the translated strings at runtime.
Quoted 6 lines
So "retrieve" is indeed involved somewhere in the process, but using the verb skips a few steps.
Subject: [PATCH 4/5]
pack-bitmap.c: mark more strings for translationsperhaps?
Yes. The explanation is clear.
Quoted 20 lines
Sorry, but I am not sure what you are asking. What I meant is that a hunk like this from the patch in discussion:
pack-bitmap.c +1
-1if (bitmap_size <0 ) { -error ("Failed to load bitmap index (corrupted?)" ); +error (_ ("Failed to load bitmap index (corrupted?)" ));ewah_pool_free (b);return NULL ; }makes translators to first translate the above string, but we will fix the "C" locale version (that is, the string inside
_()that is used as the key to the .po database when retrieving the translated version) to follow our error message formatting convention to read something likeerror(_("failed to load bitmap index (corrupted?)"));or even
Quoted 7 lines
error(_("failed to load bitmap index (corrupted?): '%s'"), filename);And the translators have to redo the work. If a preliminary patch fixed these up before bothering translators with more strings to translate, they do not need to translate the current, known to be faulty, version of messages.
Yes. I understand a bit of that, maybe. So, if the string is not C locale, translator will redo because it cannot be a translate key. Another scence is, if the string is not following the guideline like capitalized first letter, translator will redo too, we should avoid that.
Quoted 3 lines
In practice, yes, but the code is following the convention to reduce common confusion caused by leaving some lower precedence but common environment variables (i.e. LANG) as their original values.
OK.
Quoted 5 lines
Does the line in the completion script have anything to do with [PATCH 4/5], or is this merely your curiosity? Avoid mixing in unrelated things into the topic, which will only make the review cycle unnecessarily longer, but raise a separate discussion if you have to.
Curiosity.
Sorry for that, I will raise a separate one next time.
Thanks.
PATCH v5 5/5 tr2: avoid to print "interesting" config repeatedly
In reply to Teng LongWe can use GIT_TRACE2_CONFIG_PARAMS and trace2.configparams to dump the config which we are inteseted in to the tr2 log. If an "interesting" config exists in multiple scopes, it will be dumped multiple times. So, let's fix it to only print the final value instead.
--- trace2/tr2_cfg.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)
trace2/tr2_cfg.c +8 -1
@@ -1,5 +1,6 @@#include "cache.h" #include "config.h" +#include "strmap.h" #include "trace2/tr2_cfg.h" #include "trace2/tr2_sysenv.h"
@@ -10,6 +11,7 @@ static int tr2_cfg_loaded;static struct strbuf **tr2_cfg_env_vars;static int tr2_cfg_env_vars_count;static int tr2_cfg_env_vars_loaded; +static struct strset tr_cfg_set =STRSET_INIT ;/* * Parse a string containing a comma-delimited list of config keys
@@ -101,12 +103,17 @@ static int tr2_cfg_cb(const char *key, const char *value, void *d) {struct strbuf **s;struct tr2_cfg_data *data = (struct tr2_cfg_data *)d; +const char *prior_value;for (s = tr2_cfg_patterns; *s; s++) {struct strbuf *buf = *s;int wm =wildmatch (buf->buf, key,WM_CASEFOLD );if (wm ==WM_MATCH ) { -trace2_def_param_fl (data->file, data->line, key, value); +if (strset_contains (&tr_cfg_set, key) + ||git_config_get_value (key, &prior_value)) +continue ; +trace2_def_param_fl (data->file, data->line, key, prior_value); +strset_add (&tr_cfg_set, key);return 0 ; } }
Signature · 1 line
2.35.1.582.gf3b87a33da
Re: [PATCH v5 5/5] tr2: avoid to print "interesting" config repeatedly
Quoted 40 lines · Teng Long
We can use
GIT_TRACE2_CONFIG_PARAMSand trace2.configparams to dump the config which we are inteseted in to the tr2 log. If an "interesting" config exists in multiple scopes, it will be dumped multiple times. So, let's fix it to only print the final value instead.Signed-off-by Teng Long <[email protected]>Helped-by Jeff Hostetler <[email protected]>--- trace2/tr2_cfg.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-)trace2/tr2_cfg.c +3
-0@@ -1,5 +1,6 @@#include "cache.h" #include "config.h" +#include "strmap.h" #include "trace2/tr2_cfg.h" #include "trace2/tr2_sysenv.h" @@ -10,6 +11,7 @@ static int tr2_cfg_loaded;static struct strbuf **tr2_cfg_env_vars;static int tr2_cfg_env_vars_count;static int tr2_cfg_env_vars_loaded; +static struct strset tr_cfg_set =STRSET_INIT ;/* * Parse a string containing a comma-delimited list of config keys @@ -101,12 +103,17 @@ static int tr2_cfg_cb(const char *key, const char *value, void *d) {struct strbuf **s;struct tr2_cfg_data *data = (struct tr2_cfg_data *)d; +const char *prior_value;for (s = tr2_cfg_patterns; *s; s++) {struct strbuf *buf = *s;int wm =wildmatch (buf->buf, key,WM_CASEFOLD );
trace2/tr2_cfg.c static int tr2_cfg_cb(const char *key, const char *value, void *d) · line 119
Quoted 10 lines
trace2/tr2_cfg.c +5
-1if (wm ==WM_MATCH ) { -trace2_def_param_fl (data->file, data->line, key, value); +if (strset_contains (&tr_cfg_set, key) + ||git_config_get_value (key, &prior_value)) +continue ; +trace2_def_param_fl (data->file, data->line, key, prior_value); +strset_add (&tr_cfg_set, key);return 0 ; } }
Is tr2_cfg_load_patterns() run at early startup, but this perhaps at runtime? I wonder if this is OK under threading, i.e. concurrent access to your strset.
The assumption you're making here doesn't hold in general, some config is "last vaule wins", but for some other config all configured values are important.
Do we care in this case? I really don't know, but perhaps we can declare "dedup" using the same facility we're using to wildcard-match keys, and either make that optional or the default, e.g.:
GIT_TRACE2_CONFIG_PARAMS='*:dedup,core.*:'
I.e. to make it a list of <glob>[:<options>].
Maybe not worth the effort...
Re: [PATCH v5 5/5] tr2: avoid to print "interesting" config repeatedly
Quoted 9 lines · Ævar Arnfjörð Bjarmason
Do we care in this case? I really don't know, but perhaps we can declare "dedup" using the same facility we're using to wildcard-match keys, and either make that optional or the default, e.g.:
GIT_TRACE2_CONFIG_PARAMS='*:dedup,core.*:'I.e. to make it a list of <glob>[:<options>].
Maybe not worth the effort...
I happen to think that the trace output is primarily for machine consumption (i.e. if you want to make it readable by humans, it is OK to require them to pre-process).
How does this "duplicate output" come about? If it is because end-user configuration files have multiple entries that are either list-valued or relying on last-one-wins semantics, I suspect that it is better not to prematurely filter these at the trace generation side (which by definition is an opration that loses information).
So, to me, it smells that the whole "dedup at the source" thing is not just not worth the effort but is misguided.
Re: [PATCH v5 5/5] tr2: avoid to print "interesting" config repeatedly
Quoted 16 lines · Junio C Hamano
Quoted 9 lines · Ævar Arnfjörð Bjarmason
Do we care in this case? I really don't know, but perhaps we can declare "dedup" using the same facility we're using to wildcard-match keys, and either make that optional or the default, e.g.:
GIT_TRACE2_CONFIG_PARAMS='*:dedup,core.*:'I.e. to make it a list of <glob>[:<options>].
Maybe not worth the effort...
I happen to think that the trace output is primarily for machine consumption (i.e. if you want to make it readable by humans, it is OK to require them to pre-process).
Quoted 9 lines
How does this "duplicate output" come about? If it is because end-user configuration files have multiple entries that are either list-valued or relying on last-one-wins semantics, I suspect that it is better not to prematurely filter these at the trace generation side (which by definition is an opration that loses information).
So, to me, it smells that the whole "dedup at the source" thing is not just not worth the effort but is misguided.
I agree. Let's not try to dedup these. IIRC, the tr2_cfg_cb()
machinery runs during a "read early" or "read very early" scan of
the config values and the program is still starting up (in some
sense). For example, the command line may not have been fully
processed yet (again, IIRC). So injecting a call here to force an explicit lookup could cause problems.
And you'll be arbitrarily taking the first value, which is probably the system level setting rather than the last value, which is probably the local setting. So the output would be misleading.
And I don't think it is worth the effort. Let's just log the context (system, global, local) as you described in the previous version and be happy that we get multiple-but-now-qualified values.
This was intended as a machine-readable feature to allow telemetry post-processing to group or filter command events. For example, if we want to do some "feature-x" is ON or OFF and compute averages for the ONs and averages for the OFFs and compare them. Such post- processing has (I'm assuming) in the past always looked at the last value logged. And I assume that that would not change here with the additional qualification.
Also, your new qualification would help us answer support questions using telemetry where a user thought they had a feature on (say globally) but was actually off at another level (say locally).
Jeff
Re: [PATCH v5 5/5] tr2: avoid to print "interesting" config repeatedly
Quoted 7 lines
I agree. Let's not try to dedup these. IIRC, the
tr2_cfg_cb()machinery runs during a "read early" or "read very early" scan of the config values and the program is still starting up (in some sense). For example, the command line may not have been fully processed yet (again, IIRC). So injecting a call here to force an explicit lookup could cause problems.
The "read early" or "read very early" is a little abtract for me I think, I will take a further look.
Quoted 3 lines
And I don't think it is worth the effort. Let's just log the context (system, global, local) as you described in the previous version and be happy that we get multiple-but-now-qualified values.
Yes.
Quoted 3 lines
Also, your new qualification would help us answer support questions using telemetry where a user thought they had a feature on (say globally) but was actually off at another level (say locally).
That's certainly my original thought.
Thanks.
Re: [PATCH v5 5/5] tr2: avoid to print "interesting" config repeatedly
In reply to Ævar Arnfjörð BjarmasonQuoted 4 lines
Is
tr2_cfg_load_patterns()run at early startup, but this perhaps at runtime? I wonder if this is OK under threading, i.e. concurrent access to your strset.
Yes. I think it may have at runtime.
Not for that reason, for now I think it's better to dump each config and scope name is a better way in terms of other replied suggestions.
Quoted 3 lines
The assumption you're making here doesn't hold in general, some config is "last vaule wins", but for some other config all configured values are important.
Yes. Also based on the above replies.
Quoted 9 lines
Do we care in this case? I really don't know, but perhaps we can declare "dedup" using the same facility we're using to wildcard-match keys, and either make that optional or the default, e.g.:
GIT_TRACE2_CONFIG_PARAMS='*:dedup,core.*:'I.e. to make it a list of <glob>[:<options>].
Maybe not worth the effort...
I think if we could print config and scope names is enough maybe. When users specify such as GIT_TRACE2_CONFIG_PARAMS, no one will cares about the extra few lines config output in my opinion, but it may be useful for some troubleshooting, so in my opinion this may not be implemented in this patch for the time being.
Thanks.
v6 cover trace2: dump scope when print "interesting" config
In reply to Teng LongSince v5:
1. [1/7] New commit, pre-clean the issues of the strings.
2. [2/7] Optimize commit subject (the word "retrieve").
3. [4/7] New commit, do not ignore ENOENT silently when fail to open file.
4. [5/7] Replace "stat" to "fstat" in output string and let "cleanup" to return -1 instead of an unaccurate error tip as "cannot open".
5. [7/7] dump corresponding scope-name when "interesting" config is print to trace2 log.
Thanks.
Teng Long (7):
clean: fixed issues related to text output format pack-bitmap.c: mark more strings for translations pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: don't ignore ENOENT silently pack-bitmap.c: using error() instead of silently returning -1 pack-bitmap.c: continue looping when first MIDX bitmap is found
tr2: dump names if config exist in multiple scopes
pack-bitmap.c | 105 ++++++++++++++++++++++------------------ trace2/tr2_tgt_event.c | 3 ++ trace2/tr2_tgt_normal.c | 5 +- trace2/tr2_tgt_perf.c | 9 +++- 4 files changed, 72 insertions(+), 50 deletions(-)
Range-diff against v5:
-: ---------- > 1: 94a64ba895 clean: fixed issues related to text output format
4: 065b7c9ccb ! 2: cadecd6b84 pack-bitmap.c: retrieve missing i18n translations
@@ Metadata
Author: Teng Long <[email protected]>
## Commit message ##
- pack-bitmap.c: retrieve missing i18n translations
+ pack-bitmap.c: mark more strings for translations
- In pack-bitmap.c, some printed texts are translated,some are not.
+ In pack-bitmap.c, some printed texts are translated, some are not.
Let's support the translations of the bitmap related output.
Signed-off-by: Teng Long <[email protected]>
@@ pack-bitmap.c: static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *ind
index->map_size - index->map_pos);
if (bitmap_size < 0) {
-- error("Failed to load bitmap index (corrupted?)");
-+ error(_("Failed to load bitmap index (corrupted?)"));
+- error("failed to load bitmap index (corrupted?)");
++ error(_("failed to load bitmap index (corrupted?)"));
ewah_pool_free(b);
return NULL;
}
@@ pack-bitmap.c: static int load_bitmap_header(struct bitmap_index *index)
size_t header_size = sizeof(*header) - GIT_MAX_RAWSZ + the_hash_algo->rawsz;
if (index->map_size < header_size + the_hash_algo->rawsz)
-- return error("Corrupted bitmap index (too small)");
-+ return error(_("Corrupted bitmap index (too small)"));
+- return error("corrupted bitmap index (too small)");
++ return error(_("corrupted bitmap index (too small)"));
if (memcmp(header->magic, BITMAP_IDX_SIGNATURE, sizeof(BITMAP_IDX_SIGNATURE)) != 0)
-- return error("Corrupted bitmap index file (wrong header)");
-+ return error(_("Corrupted bitmap index file (wrong header)"));
+- return error("corrupted bitmap index file (wrong header)");
++ return error(_("corrupted bitmap index file (wrong header)"));
index->version = ntohs(header->version);
if (index->version != 1)
-- return error("Unsupported version for bitmap index file (%d)", index->version);
-+ return error(_("Unsupported version for bitmap index file (%d)"), index->version);
+- return error("unsupported version '%d' for bitmap index file", index->version);
++ return error(_("unsupported version '%d' for bitmap index file"), index->version);
/* Parse known bitmap format options */
{
@@ pack-bitmap.c: static int load_bitmap_header(struct bitmap_index *index)
- unsigned char *index_end = index->map + index->map_size - the_hash_algo->rawsz;
-
- if ((flags & BITMAP_OPT_FULL_DAG) == 0)
-- return error("Unsupported options for bitmap index file "
-- "(Git requires BITMAP_OPT_FULL_DAG)");
-+ return error(_("Unsupported options for bitmap index file "
-+ "(Git requires BITMAP_OPT_FULL_DAG)"));
if (flags & BITMAP_OPT_HASH_CACHE) {
if (cache_size > index_end - index->map - header_size)
@@ pack-bitmap.c: static struct stored_bitmap *store_bitmap(struct bitmap_index *in
* because the SHA1 already existed on the map. this is bad, there
* shouldn't be duplicated commits in the index */
if (ret == 0) {
-- error("Duplicate entry in bitmap index: %s", oid_to_hex(oid));
-+ error(_("Duplicate entry in bitmap index: %s"), oid_to_hex(oid));
+- error("duplicate entry in bitmap index: '%s'", oid_to_hex(oid));
++ error(_("duplicate entry in bitmap index: '%s'"), oid_to_hex(oid));
return NULL;
}
@@ pack-bitmap.c: static int load_bitmap_entries_v1(struct bitmap_index *index)
if (index->map_size - index->map_pos < 6)
- return error("corrupt ewah bitmap: truncated header for entry %d", i);
-+ return error(_("corrupt ewah bitmap: truncated header for entry %d)"), i);
++ return error(_("corrupt ewah bitmap: truncated header for entry %d"), i);
commit_idx_pos = read_be32(index->map, &index->map_pos);
xor_offset = read_u8(index->map, &index->map_pos);
@@ pack-bitmap.c: static int load_bitmap_entries_v1(struct bitmap_index *index)
return -1;
if (xor_offset > MAX_XOR_OFFSET || xor_offset > i)
-- return error("Corrupted bitmap pack index");
-+ return error(_("Corrupted bitmap pack index"));
+- return error("corrupted bitmap pack index");
++ return error(_("corrupted bitmap pack index"));
if (xor_offset > 0) {
xor_bitmap = recent_bitmaps[(i - xor_offset) % MAX_XOR_OFFSET];
if (xor_bitmap == NULL)
-- return error("Invalid XOR offset in bitmap pack index");
-+ return error(_("Invalid XOR offset in bitmap pack index"));
+- return error("invalid XOR offset in bitmap pack index");
++ return error(_("invalid XOR offset in bitmap pack index"));
}
recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
-@@ pack-bitmap.c: char *pack_bitmap_filename(struct packed_git *p)
- size_t len;
-
- if (!strip_suffix(p->pack_name, ".pack", &len))
-- BUG("pack_name does not end in .pack");
-+ BUG(_("pack_name does not end in .pack"));
- return xstrfmt("%.*s.bitmap", (int)len, p->pack_name);
- }
-
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
struct strbuf buf = STRBUF_INIT;
get_midx_filename(&buf, midx->object_dir);
/* ignore extra bitmap file; we can only handle one */
-- warning("ignoring extra bitmap file: %s", buf.buf);
-+ warning(_("ignoring extra bitmap file: %s"), buf.buf);
+- warning("ignoring extra bitmap file: '%s'", buf.buf);
++ warning(_("ignoring extra bitmap file: '%s'"), buf.buf);
close(fd);
strbuf_release(&buf);
return -1;
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, st
if (bitmap_git->pack || bitmap_git->midx) {
/* ignore extra bitmap file; we can only handle one */
-- warning("ignoring extra bitmap file: %s", packfile->pack_name);
-+ warning(_("ignoring extra bitmap file: %s"), packfile->pack_name);
+- warning("ignoring extra bitmap file: '%s'", packfile->pack_name);
++ warning(_("ignoring extra bitmap file: '%s'"), packfile->pack_name);
close(fd);
return -1;
}
@@ pack-bitmap.c: static struct bitmap *find_objects(struct bitmap_index *bitmap_gi
show_data.bitmap_git = bitmap_git;
show_data.base = base;
-@@ pack-bitmap.c: static void filter_bitmap_tree_depth(struct bitmap_index *bitmap_git,
- unsigned long limit)
- {
- if (limit)
-- BUG("filter_bitmap_tree_depth given non-zero limit");
-+ BUG(_("filter_bitmap_tree_depth given non-zero limit"));
-
- filter_bitmap_exclude_type(bitmap_git, tip_objects, to_filter,
- OBJ_TREE);
-@@ pack-bitmap.c: static void filter_bitmap_object_type(struct bitmap_index *bitmap_git,
- enum object_type object_type)
- {
- if (object_type < OBJ_COMMIT || object_type > OBJ_TAG)
-- BUG("filter_bitmap_object_type given invalid object");
-+ BUG(_("filter_bitmap_object_type given invalid object"));
-
- if (object_type != OBJ_TAG)
- filter_bitmap_exclude_type(bitmap_git, tip_objects, to_filter, OBJ_TAG);
-@@ pack-bitmap.c: struct bitmap_index *prepare_bitmap_walk(struct rev_info *revs,
- revs->ignore_missing_links = 0;
-
- if (haves_bitmap == NULL)
-- BUG("failed to perform bitmap walk");
-+ BUG(_("failed to perform bitmap walk"));
- }
-
- wants_bitmap = find_objects(bitmap_git, revs, wants, haves_bitmap,
- filter);
-
- if (!wants_bitmap)
-- BUG("failed to perform bitmap walk");
-+ BUG(_("failed to perform bitmap walk"));
-
- if (haves_bitmap)
- bitmap_and_not(wants_bitmap, haves_bitmap);
-@@ pack-bitmap.c: uint32_t midx_preferred_pack(struct bitmap_index *bitmap_git)
- {
- struct multi_pack_index *m = bitmap_git->midx;
- if (!m)
-- BUG("midx_preferred_pack: requires non-empty MIDX");
-+ BUG(_("midx_preferred_pack: requires non-empty MIDX"));
- return nth_midxed_pack_int_id(m, pack_pos_to_midx(bitmap_git->midx, 0));
- }
-
@@ pack-bitmap.c: static void test_bitmap_type(struct bitmap_test_data *tdata,
}
if (bitmap_type == OBJ_NONE)
-- die("object %s not found in type bitmaps",
-+ die(_("object %s not found in type bitmaps"),
+- die("object '%s' not found in type bitmaps",
++ die(_("object '%s' not found in type bitmaps"),
oid_to_hex(&obj->oid));
if (bitmaps_nr > 1)
-- die("object %s does not have a unique type",
-+ die(_("object %s does not have a unique type"),
+- die("object '%s' does not have a unique type",
++ die(_("object '%s' does not have a unique type"),
oid_to_hex(&obj->oid));
if (bitmap_type != obj->type)
-- die("object %s: real type %s, expected: %s",
-+ die(_("object %s: real type %s, expected: %s"),
+- die("object '%s': real type '%s', expected: '%s'",
++ die(_("object '%s': real type '%s', expected: '%s'"),
oid_to_hex(&obj->oid),
type_name(obj->type),
type_name(bitmap_type));
@@ pack-bitmap.c: static void test_show_object(struct object *object, const char *n
bitmap_pos = bitmap_position(tdata->bitmap_git, &object->oid);
if (bitmap_pos < 0)
-- die("Object not in bitmap: %s\n", oid_to_hex(&object->oid));
-+ die(_("Object not in bitmap: %s\n"), oid_to_hex(&object->oid));
+- die("object not in bitmap: '%s'", oid_to_hex(&object->oid));
++ die(_("object not in bitmap: '%s'"), oid_to_hex(&object->oid));
test_bitmap_type(tdata, object, bitmap_pos);
bitmap_set(tdata->base, bitmap_pos);
@@ pack-bitmap.c: static void test_show_commit(struct commit *commit, void *data)
bitmap_pos = bitmap_position(tdata->bitmap_git,
&commit->object.oid);
if (bitmap_pos < 0)
-- die("Object not in bitmap: %s\n", oid_to_hex(&commit->object.oid));
-+ die(_("Object not in bitmap: %s\n"), oid_to_hex(&commit->object.oid));
+- die("object not in bitmap: '%s'", oid_to_hex(&commit->object.oid));
++ die(_("object not in bitmap: '%s'"), oid_to_hex(&commit->object.oid));
test_bitmap_type(tdata, &commit->object, bitmap_pos);
bitmap_set(tdata->base, bitmap_pos);
@@ pack-bitmap.c: void test_bitmap_walk(struct rev_info *revs)
- die("you must specify exactly one commit to test");
+ die(_("you must specify exactly one commit to test"));
-- fprintf(stderr, "Bitmap v%d test (%d entries loaded)\n",
-+ fprintf(stderr, _("Bitmap v%d test (%d entries loaded)\n"),
+ fprintf_ln(stderr, "Bitmap v%d test (%d entries loaded)",
bitmap_git->version, bitmap_git->entry_count);
-
- root = revs->pending.objects[0].item;
- bm = bitmap_for_commit(bitmap_git, (struct commit *)root);
-
- if (bm) {
-- fprintf(stderr, "Found bitmap for %s. %d bits / %08x checksum\n",
-+ fprintf(stderr, _("Found bitmap for %s. %d bits / %08x checksum\n"),
- oid_to_hex(&root->oid), (int)bm->bit_size, ewah_checksum(bm));
-
- result = ewah_to_bitmap(bm);
+@@ pack-bitmap.c: void test_bitmap_walk(struct rev_info *revs)
}
if (result == NULL)
-- die("Commit %s doesn't have an indexed bitmap", oid_to_hex(&root->oid));
-+ die(_("Commit %s doesn't have an indexed bitmap"), oid_to_hex(&root->oid));
+- die("commit '%s' doesn't have an indexed bitmap", oid_to_hex(&root->oid));
++ die(_("commit '%s' doesn't have an indexed bitmap"), oid_to_hex(&root->oid));
revs->tag_objects = 1;
revs->tree_objects = 1;
@@ pack-bitmap.c: void test_bitmap_walk(struct rev_info *revs)
tdata.bitmap_git = bitmap_git;
tdata.base = bitmap_new();
@@ pack-bitmap.c: void test_bitmap_walk(struct rev_info *revs)
- tdata.trees = ewah_to_bitmap(bitmap_git->trees);
- tdata.blobs = ewah_to_bitmap(bitmap_git->blobs);
- tdata.tags = ewah_to_bitmap(bitmap_git->tags);
-- tdata.prg = start_progress("Verifying bitmap entries", result_popcnt);
-+ tdata.prg = start_progress(_("Verifying bitmap entries"), result_popcnt);
- tdata.seen = 0;
-
- traverse_commit_list(revs, &test_show_commit, &test_show_object, &tdata);
-@@ pack-bitmap.c: void test_bitmap_walk(struct rev_info *revs)
- stop_progress(&tdata.prg);
-
if (bitmap_equals(result, tdata.base))
-- fprintf(stderr, "OK!\n");
-+ fprintf(stderr, _("OK!\n"));
+ fprintf_ln(stderr, "OK!");
else
- die("mismatch in bitmap results");
+ die(_("mismatch in bitmap results"));
@@ pack-bitmap.c: int test_bitmap_commits(struct repository *r)
+ die(_("failed to load bitmap indexes"));
kh_foreach(bitmap_git->bitmaps, oid, value, {
- printf("%s\n", oid_to_hex(&oid));
-@@ pack-bitmap.c: uint32_t *create_bitmap_mapping(struct bitmap_index *bitmap_git,
- if (!bitmap_is_midx(bitmap_git))
- load_reverse_index(bitmap_git);
- else if (load_midx_revindex(bitmap_git->midx) < 0)
-- BUG("rebuild_existing_bitmaps: missing required rev-cache "
-- "extension");
-+ BUG(_("rebuild_existing_bitmaps: missing required rev-cache "
-+ "extension"));
-
- num_objects = bitmap_num_objects(bitmap_git);
- CALLOC_ARRAY(reposition, num_objects);
+ printf_ln("%s", oid_to_hex(&oid));
2: b6b30047fc = 3: d8a2235cb0 pack-bitmap.c: rename "idx_name" to "bitmap_name"
-: ---------- > 4: 009cc49a18 pack-bitmap.c: don't ignore ENOENT silently
-: ---------- > 5: 52783555e2 pack-bitmap.c: using error() instead of silently returning -1
1: 589e3f4075 = 6: 95832190b8 pack-bitmap.c: continue looping when first MIDX bitmap is found
3: 82d4493a6e < -: ---------- pack-bitmap.c: using error() instead of silently returning -1
5: f3b87a33da < -: ---------- tr2: avoid to print "interesting" config repeatedly
-: ---------- > 7: c45ead51ff tr2: dump names if config exist in multiple scopesSignature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
PATCH v6 1/7 clean: fixed issues related to text output format
There are some text output issues in 'pack-bitmap.c', they exist in die(), error() etc. This includes issues with capitalization the first letter, newlines, error() instead of BUG(), and substitution that don't have quotes around them.
--- pack-bitmap.c | 47 ++++++++++++++++++++++++----------------------- 1 file changed, 24 insertions(+), 23 deletions(-)
pack-bitmap.c +24 -23
@@ -1,5 +1,6 @@#include "cache.h" #include "commit.h" +#include "strbuf.h" #include "tag.h" #include "diff.h" #include "revision.h"
@@ -138,7 +139,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("Failed to load bitmap index (corrupted?)" ); +error ("failed to load bitmap index (corrupted?)" );ewah_pool_free (b);return NULL ; }
@@ -160,14 +161,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("Corrupted bitmap index (too small)" ); +return error ("corrupted bitmap index (too small)" );if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("Corrupted bitmap index file (wrong header)" ); +return error ("corrupted bitmap index file (wrong header)" ); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("Unsupported version for bitmap index file (%d)" , index->version); +return error ("unsupported version '%d' for bitmap index file" , index->version);/* Parse known bitmap format options */ {
@@ -176,7 +177,7 @@ static int load_bitmap_header(struct bitmap_index *index)unsigned char *index_end = index->map + index->map_size - the_hash_algo->rawsz;if ((flags &BITMAP_OPT_FULL_DAG ) ==0 ) -return error ("Unsupported options for bitmap index file " +BUG ("unsupported options for bitmap index file " "(Git requires BITMAP_OPT_FULL_DAG)" );if (flags &BITMAP_OPT_HASH_CACHE ) {
@@ -215,7 +216,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("Duplicate entry in bitmap index: %s" ,oid_to_hex (oid)); +error ("duplicate entry in bitmap index: '%s'" ,oid_to_hex (oid));return NULL ; }
@@ -274,13 +275,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("Corrupted bitmap pack index" ); +return error ("corrupted bitmap pack index" );if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("Invalid XOR offset in bitmap pack index" ); +return error ("invalid XOR offset in bitmap pack index" ); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -330,7 +331,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning ("ignoring extra bitmap file: '%s'" , buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +388,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning ("ignoring extra bitmap file: '%s'" , packfile->pack_name);close (fd);return -1 ; }
@@ -1628,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object %s not found in type bitmaps" , +die ("object '%s' not found in type bitmaps" ,oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object %s does not have a unique type" , +die ("object '%s' does not have a unique type" ,oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object %s: real type %s, expected: %s" , +die ("object '%s': real type '%s', expected: '%s'" ,oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1650,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&object->oid)); +die ("object not in bitmap: '%s'" ,oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1665,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&commit->object.oid)); +die ("object not in bitmap: '%s'" ,oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1687,21 +1688,21 @@ void test_bitmap_walk(struct rev_info *revs)if (revs->pending.nr !=1 )die ("you must specify exactly one commit to test" ); -fprintf (stderr,"Bitmap v%d test (%d entries loaded)\n" , +fprintf_ln (stderr,"Bitmap v%d test (%d entries loaded)" , bitmap_git->version, bitmap_git->entry_count); root = revs->pending.objects[0 ].item; bm =bitmap_for_commit (bitmap_git, (struct commit *)root);if (bm) { -fprintf (stderr,"Found bitmap for %s. %d bits / %08x checksum\n" , +fprintf_ln (stderr,"Found bitmap for '%s'. %d bits / %08x checksum" ,oid_to_hex (&root->oid), (int )bm->bit_size,ewah_checksum (bm)); result =ewah_to_bitmap (bm); }if (result ==NULL ) -die ("Commit %s doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die ("commit '%s' doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1726,7 +1727,7 @@ void test_bitmap_walk(struct rev_info *revs)stop_progress (&tdata.prg);if (bitmap_equals (result, tdata.base)) -fprintf (stderr,"OK!\n" ); +fprintf_ln (stderr,"OK!" );else die ("mismatch in bitmap results" );
@@ -1749,7 +1750,7 @@ int test_bitmap_commits(struct repository *r)die ("failed to load bitmap indexes" );kh_foreach (bitmap_git->bitmaps, oid, value, { -printf ("%s\n" ,oid_to_hex (&oid)); +printf_ln ("%s" ,oid_to_hex (&oid)); });free_bitmap_index (bitmap_git);
@@ -1774,7 +1775,7 @@ int test_bitmap_hashes(struct repository *r)nth_bitmap_object_oid (bitmap_git, &oid, index_pos); -printf ("%s %" PRIu32"\n" , +printf_ln ("%s %" PRIu32"" ,oid_to_hex (&oid),get_be32 (bitmap_git->hashes + index_pos)); }
@@ -1936,7 +1937,7 @@ static off_t get_disk_usage_for_type(struct bitmap_index *bitmap_git,struct object_id oid;nth_midxed_object_oid (&oid, bitmap_git->midx, midx_pos); -die (_ ("could not find %s in pack %s at offset %" PRIuMAX), +die (_ ("could not find '%s' in pack '%s' at offset %" PRIuMAX),oid_to_hex (&oid), pack->pack_name, (uintmax_t )offset);
@@ -1972,7 +1973,7 @@ static off_t get_disk_usage_for_extended(struct bitmap_index *bitmap_git)continue ;if (oid_object_info_extended (the_repository, &obj->oid, &oi,0 ) <0 ) -die (_ ("unable to get disk usage of %s" ), +die (_ ("unable to get disk usage of '%s'" ),oid_to_hex (&obj->oid)); total += object_size;
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
Re: [PATCH v6 1/7] clean: fixed issues related to text output format
Quoted 1 line · Teng Long
Subject: Re: [PATCH v6 1/7] clean: fixed issues related to text output format
This does not seem to be about "git clean". Mistitled?
pack-bitmap.c: fix formatting of error messages
perhaps?
Re: [PATCH v6 1/7] clean: fixed issues related to text output format
Quoted 1 line · On Mon, 11 Jul 2022 14:08:13 -070 Junio C Hamano wrote:
This does not seem to be about "git clean". Mistitled?
Yes, will fix.
Quoted 3 lines
pack-bitmap.c: fix formatting of error messagesperhaps?
Will accept.
Thanks.
PATCH v6 2/7 pack-bitmap.c: mark more strings for translations
In reply to Teng LongIn pack-bitmap.c, some printed texts are translated, some are not. Let's support the translations of the bitmap related output.
--- pack-bitmap.c | 48 ++++++++++++++++++++++++------------------------ 1 file changed, 24 insertions(+), 24 deletions(-)
pack-bitmap.c +24 -24
@@ -139,7 +139,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("failed to load bitmap index (corrupted?)" ); +error (_ ("failed to load bitmap index (corrupted?)" ));ewah_pool_free (b);return NULL ; }
@@ -161,14 +161,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("corrupted bitmap index (too small)" ); +return error (_ ("corrupted bitmap index (too small)" ));if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("corrupted bitmap index file (wrong header)" ); +return error (_ ("corrupted bitmap index file (wrong header)" )); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("unsupported version '%d' for bitmap index file" , index->version); +return error (_ ("unsupported version '%d' for bitmap index file" ), index->version);/* Parse known bitmap format options */ {
@@ -182,7 +182,7 @@ static int load_bitmap_header(struct bitmap_index *index)if (flags &BITMAP_OPT_HASH_CACHE ) {if (cache_size > index_end - index->map - header_size) -return error ("corrupted bitmap index file (too short to fit hash cache)" ); +return error (_ ("corrupted bitmap index file (too short to fit hash cache)" )); index->hashes = (void *)(index_end - cache_size); index_end -= cache_size; }
@@ -216,7 +216,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("duplicate entry in bitmap index: '%s'" ,oid_to_hex (oid)); +error (_ ("duplicate entry in bitmap index: '%s'" ),oid_to_hex (oid));return NULL ; }
@@ -260,14 +260,14 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)struct object_id oid;if (index->map_size - index->map_pos <6 ) -return error ("corrupt ewah bitmap: truncated header for entry %d" , i); +return error (_ ("corrupt ewah bitmap: truncated header for entry %d" ), i); commit_idx_pos =read_be32 (index->map, &index->map_pos); xor_offset =read_u8 (index->map, &index->map_pos); flags =read_u8 (index->map, &index->map_pos);if (nth_bitmap_object_oid (index, &oid, commit_idx_pos) <0 ) -return error ("corrupt ewah bitmap: commit index %u out of range" , +return error (_ ("corrupt ewah bitmap: commit index %u out of range" ), (unsigned )commit_idx_pos); bitmap =read_bitmap_1 (index);
@@ -275,13 +275,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("corrupted bitmap pack index" ); +return error (_ ("corrupted bitmap pack index" ));if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("invalid XOR offset in bitmap pack index" ); +return error (_ ("invalid XOR offset in bitmap pack index" )); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -331,7 +331,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: '%s'" , buf.buf); +warning (_ ("ignoring extra bitmap file: '%s'" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -388,7 +388,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: '%s'" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: '%s'" ), packfile->pack_name);close (fd);return -1 ; }
@@ -819,7 +819,7 @@ static struct bitmap *find_objects(struct bitmap_index *bitmap_git, revs->include_check_data = &incdata;if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); show_data.bitmap_git = bitmap_git; show_data.base = base;
@@ -1629,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object '%s' not found in type bitmaps" , +die (_ ("object '%s' not found in type bitmaps" ),oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object '%s' does not have a unique type" , +die (_ ("object '%s' does not have a unique type" ),oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object '%s': real type '%s', expected: '%s'" , +die (_ ("object '%s': real type '%s', expected: '%s'" ),oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1651,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("object not in bitmap: '%s'" ,oid_to_hex (&object->oid)); +die (_ ("object not in bitmap: '%s'" ),oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1666,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("object not in bitmap: '%s'" ,oid_to_hex (&commit->object.oid)); +die (_ ("object not in bitmap: '%s'" ),oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1683,10 +1683,10 @@ void test_bitmap_walk(struct rev_info *revs)struct ewah_bitmap *bm;if (!(bitmap_git =prepare_bitmap_git (revs->repo))) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" ));if (revs->pending.nr !=1 ) -die ("you must specify exactly one commit to test" ); +die (_ ("you must specify exactly one commit to test" ));fprintf_ln (stderr,"Bitmap v%d test (%d entries loaded)" , bitmap_git->version, bitmap_git->entry_count);
@@ -1702,7 +1702,7 @@ void test_bitmap_walk(struct rev_info *revs) }if (result ==NULL ) -die ("commit '%s' doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die (_ ("commit '%s' doesn't have an indexed bitmap" ),oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1711,7 +1711,7 @@ void test_bitmap_walk(struct rev_info *revs) result_popcnt =bitmap_popcount (result);if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); tdata.bitmap_git = bitmap_git; tdata.base =bitmap_new ();
@@ -1729,7 +1729,7 @@ void test_bitmap_walk(struct rev_info *revs)if (bitmap_equals (result, tdata.base))fprintf_ln (stderr,"OK!" );else -die ("mismatch in bitmap results" ); +die (_ ("mismatch in bitmap results" ));bitmap_free (result);bitmap_free (tdata.base);
@@ -1747,7 +1747,7 @@ int test_bitmap_commits(struct repository *r) MAYBE_UNUSED void *value;if (!bitmap_git) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" )); kh_foreach(bitmap_git->bitmaps, oid, value, { printf_ln("%s", oid_to_hex(&oid));
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
PATCH v6 3/7 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index, let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -314,10 +314,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -369,14 +369,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
PATCH v6 4/7 pack-bitmap.c: don't ignore ENOENT silently
In reply to Teng LongWhen finished call git_open(), instead of ignoring ENOENT silently and return error_errno(_("cannot stat...")), it's better to check the ENOENT before then output the warning.
--- pack-bitmap.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-)
pack-bitmap.c +12 -5
@@ -317,10 +317,13 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,char *bitmap_name =midx_bitmap_filename (midx);int fd =git_open (bitmap_name); -free (bitmap_name); - -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning ("'%s' cannot open '%s'" ,strerror (errno), bitmap_name); +free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
@@ -376,10 +379,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_name =pack_bitmap_filename (packfile); fd =git_open (bitmap_name); -free (bitmap_name); -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning ("'%s' cannot open '%s'" ,strerror (errno), bitmap_name); +free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
Re: [PATCH v6 4/7] pack-bitmap.c: don't ignore ENOENT silently
Quoted 3 lines · Teng Long
When finished call
git_open(), instead of ignoring ENOENT silently and returnerror_errno(_("cannot stat...")), it's better to check the ENOENT before then output the warning.
Makes sense, but...
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 322
Quoted 4 lines
pack-bitmap.c +3
-1-if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning ("'%s' cannot open '%s'" ,strerror (errno), bitmap_name);
This should just be warning_errno("cannot open '%s'", bitmap_name), unless I'm missing something...
Quoted 8 lines
pack-bitmap.c +3
-0+free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 385
Quoted 9 lines
pack-bitmap.c +3
-2@@ -376,10 +379,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_name =pack_bitmap_filename (packfile); fd =git_open (bitmap_name); -free (bitmap_name); -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning ("'%s' cannot open '%s'" ,strerror (errno), bitmap_name);
...ditto.
Quoted 7 lines
pack-bitmap.c +3
-0+free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
Re: [PATCH v6 4/7] pack-bitmap.c: don't ignore ENOENT silently
Quoted 7 lines · Ævar Arnfjörð Bjarmason
Makes sense, but...
...
This should just bewarning_errno("cannot open '%s'", bitmap_name), unless I'm missing something...
...
...ditto.
You are right, I didn't notice warning_errno() before, will improve.
Thanks.
Re: [PATCH v6 4/7] pack-bitmap.c: don't ignore ENOENT silently
In reply to Teng LongQuoted 3 lines · Teng Long
When finished call
git_open(), instead of ignoring ENOENT silently and returnerror_errno(_("cannot stat...")), it's better to check the ENOENT before then output the warning.
Both the title and the body describes a complete opposite of what the patch does, doesn't it?
pack-bitmap.c: do not ignore error when opening a bitmap file
Calls to git_open() to open the pack bitmap file and multi-pack bitmap file do not report any error when they fail. These files are optional and it is not an error if open failed due to ENOENT, but we shouldn't be ignoring other kinds of errors.
or something?
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 323
Quoted 4 lines
pack-bitmap.c +4
-0+if (fd <0 ) { +if (errno !=ENOENT ) +warning ("'%s' cannot open '%s'" ,strerror (errno), bitmap_name); +free (bitmap_name);
Showing the errno is good, but I do not think it should be enclosed in single quotes.
One thing you should consider when writing a "this is an optional file and not finding it is perfectly fine" logic like this one is that you may or may not want to ignore ENOTDIR. ENOTDIR is what you get when you say open("a/b") and "a" exists as something other than a directory. If you have a path with a slash in bitmap_name, and if in a sane and healthy repository the leading path should always exist (i.e. the file "a/b" may be missing, but the directory "a/" should be there), then getting ENOTDIR is a noteworthy event. There may be cases where ENOTDIR can justifiably and should be ignored.
Re: [PATCH v6 4/7] pack-bitmap.c: don't ignore ENOENT silently
Quoted 5 lines · Junio C Hamano
Both the title and the body describes a complete opposite of what the patch does, doesn't it?
pack-bitmap.c: do not ignore error when opening a bitmap file
Quoted 7 lines
Calls to
git_open()to open the pack bitmap file and multi-pack bitmap file do not report any error when they fail. These files are optional and it is not an error if open failed due to ENOENT, but we shouldn't be ignoring other kinds of errors.or something?
Yes, will apply, thanks for this detailed assistent.
Quoted 7 lines
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 323
Quoted 4 lines
pack-bitmap.c +4
-0+if (fd <0 ) { +if (errno !=ENOENT ) +warning ("'%s' cannot open '%s'" ,strerror (errno), bitmap_name); +free (bitmap_name);Showing the errno is good, but I do not think it should be enclosed in single quotes.
Yes, it should not be enclosed around and by Ævar Arnfjörð Bjarmason's advice the "warning" should be replaced by "warning_errno", I agree with that, so, I think at the same time it solves this problem.
Quoted 3 lines
One thing you should consider when writing a "this is an optional file and not finding it is perfectly fine" logic like this one is that you may or may not want to ignore ENOTDIR.
$ errno ENOTDIR ENOTDIR 20 Not a directory
Quoted 6 lines
ENOTDIR is what you
get when you sayopen("a/b")and "a" exists as something other than a directory. If you have a path with a slash inbitmap_name, and if in a sane and healthy repository the leading path should always exist (i.e. the file "a/b" may be missing, but the directory "a/" should be there), then getting ENOTDIR is a noteworthy event.
Make sense, and I tried to move "pack" away and touch a new file named "pack", then executed "git fsck", it will show the result about ENOTDIR:
$ git fsck Checking object directories: 100% (256/256), done. error: unable to open object pack directory: /Users/tenglong.tl/Downloads/dyrone/.git/objects/pack: Not a directory
So, this is how it works in git I think, caller should also print the "errno" when deal error and warning, append errno string as the last part and the first part is to describe what's wrong happened in git.
Quoted 1 line
There may be cases where ENOTDIR can justifiably and should be ignored.
Yes. If I understand it correctly, I think it's the current scene.
For example, if this happens in case like "./pack" is a non-directory, then we print warning which contains ENOTDIR description under current change.
Thanks.
PATCH v6 5/7 pack-bitmap.c: using error() instead of silently returning -1
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)
pack-bitmap.c +5 -3
@@ -327,7 +327,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot fstat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
@@ -350,8 +350,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (load_bitmap_header (bitmap_git) <0 ) goto cleanup; -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +error (_ ("checksum doesn't match in MIDX and bitmap" )); goto cleanup; + }if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" ));
@@ -390,7 +392,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot fstat bitmap file" )); }if (bitmap_git->pack || bitmap_git->midx) {
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
Re: [PATCH v6 5/7] pack-bitmap.c: using error() instead of silently returning -1
On Mon, Jul 11 2022, Teng Long wrote:
Good to report errno now, however...
Quoted 10 lines
Signed-off-by Teng Long <[email protected]>--- pack-bitmap.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-)pack-bitmap.c +0
-0
pack-bitmap.c static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, · line 330
Quoted 5 lines
pack-bitmap.c +1
-1@@ -327,7 +327,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot fstat bitmap file" ));
This is a logic error, as fstat() failed, but you're reproting errno() after our call to close(), at which point it's anyone's guess what "errno" is. It's only meaningful immediately after a system call.
So either:
error_errno(....);
close(fd);
return -1;Or even better:
error_errno(...)
if (close(fd))
warning_errno("cannot close() bitmap file");
return -1;Although that last one may be too pedantic.
Quoted 16 lines
pack-bitmap.c +3
-1}if (bitmap_git->pack || bitmap_git->midx) {@@ -350,8 +350,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (load_bitmap_header (bitmap_git) <0 ) goto cleanup; -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +error (_ ("checksum doesn't match in MIDX and bitmap" )); goto cleanup; + }if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" ));
pack-bitmap.c static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git · line 395
Quoted 5 lines
pack-bitmap.c +1
-1@@ -390,7 +392,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (fstat (fd, &st)) {close (fd); -return -1 ; +return error_errno (_ ("cannot fstat bitmap file" ));
Same issue here.
Re: [PATCH v6 5/7] pack-bitmap.c: using error() instead of silently returning -1
Quoted 12 lines · Ævar Arnfjörð Bjarmason
This is a logic error, as
fstat()failed, but you're reprotingerrno()after our call toclose(), at which point it's anyone's guess what "errno" is. It's only meaningful immediately after a system call.So either:
error_errno(....); close(fd); return -1;Or even better:
Quoted 8 lines
error_errno(...) if (close(fd)) warning_errno("cannot close() bitmap file"); return -1;Although that last one may be too pedantic.
...
Same issue here.
Yes, this will be fixed in next patch.
Thanks.
PATCH v6 6/7 pack-bitmap.c: continue looping when first MIDX bitmap is found
In reply to Teng LongIn "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to continue the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The reason for this is, there may exist more than one MIDX file in a repo. The "multi_pack_index" struct is actually designed as a singly linked list, and if a MIDX file has been already opened successfully, then the other MIDX files will be skipped and left with a warning "ignoring extra bitmap file." to the output.
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -504,15 +504,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
PATCH v6 7/7 tr2: dump names if config exist in multiple scopes
In reply to Teng LongWhen we specify GIT_TRACE2_CONFIG_PARAMS or trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):
...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false
As the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \
Quoted 2 lines
GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \
git rev-list --test-bitmap HEAD"
The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ... def_param scope:system core.multipackindex=false ... git.c:461 ... def_param scope:global core.multipackindex=false ... git.c:461 ... def_param scope:local core.multipackindex=false
Format perf:
... | def_param | ... | scope:system | core.multipackindex:false ... | def_param | ... | scope:global | core.multipackindex:false ... | def_param | ... | scope:local | core.multipackindex:false
Format event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
--- trace2/tr2_tgt_event.c | 3 +++ trace2/tr2_tgt_normal.c | 5 ++++- trace2/tr2_tgt_perf.c | 9 +++++++-- 3 files changed, 14 insertions(+), 3 deletions(-)
trace2/tr2_tgt_event.c +3 -0
@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);jw_object_begin (&jw,0 );event_fmt_prepare (event_name, file, line,NULL , &jw); +jw_object_string (&jw,"scope" , scope_name);jw_object_string (&jw,"param" , param);jw_object_string (&jw,"value" , value);jw_end (&jw);
trace2/tr2_tgt_normal.c +4 -1
@@ -298,8 +298,11 @@ static void fn_param_fl(const char *file, int line, const char *param, const char *value) {struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope); -strbuf_addf (&buf_payload,"def_param %s=%s" , param, value); +strbuf_addf (&buf_payload,"def_param scope:%s %s=%s" , scope_name, param, + value);normal_io_write_fl (file, line, &buf_payload);strbuf_release (&buf_payload); }
trace2/tr2_tgt_perf.c +7 -2
@@ -441,12 +441,17 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct strbuf buf_payload =STRBUF_INIT ; +struct strbuf scope_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);strbuf_addf (&buf_payload,"%s:%s" , param, value); +strbuf_addf (&scope_payload,"%s:%s" ,"scope" , scope_name); -perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL ,NULL , - &buf_payload); +perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL , + scope_payload.buf, &buf_payload);strbuf_release (&buf_payload); +strbuf_release (&scope_payload); } static void fn_repo_fl(const char *file, int line,
Signature · 1 line
2.35.0.rc0.676.g60105b7097.dirty
Re: [PATCH v6 7/7] tr2: dump names if config exist in multiple scopes
Quoted 32 lines · Teng Long
When we specify
GIT_TRACE2_CONFIG_PARAMSor trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:falseAs the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \Quoted 2 lines
GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \
git rev-list --test-bitmap HEAD"The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ...def_paramscope:system core.multipackindex=false ... git.c:461 ...def_paramscope:global core.multipackindex=false ... git.c:461 ...def_paramscope:local core.multipackindex=falseFormat perf:
... | def_param | ... | scope:system | core.multipackindex:false ... | def_param | ... | scope:global | core.multipackindex:false ... | def_param | ... | scope:local | core.multipackindex:false
Quoted 5 lines
Format event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
This seems sensible on its face, but...
Quoted 6 lines
Signed-off-by Teng Long <[email protected]>--- trace2/tr2_tgt_event.c | 3 +++ trace2/tr2_tgt_normal.c | 5 ++++- trace2/tr2_tgt_perf.c | 9 +++++++-- 3 files changed, 14 insertions(+), 3 deletions(-)
... we really should update Documentation/technical/api-trace2.txt here too.
It does say "..." currently, so we're not lying there, but since we now add "scope" unconditionally...
Re: [PATCH v6 7/7] tr2: dump names if config exist in multiple scopes
In reply to Teng LongQuoted 94 lines · Teng Long
When we specify
GIT_TRACE2_CONFIG_PARAMSor trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:falseAs the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \ > GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \ > git rev-list --test-bitmap HEAD"The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ...def_paramscope:system core.multipackindex=false ... git.c:461 ...def_paramscope:global core.multipackindex=false ... git.c:461 ...def_paramscope:local core.multipackindex=falseFormat perf:
... | def_param | ... | scope:system | core.multipackindex:false ... | def_param | ... | scope:global | core.multipackindex:false ... | def_param | ... | scope:local | core.multipackindex:falseFormat event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
Signed-off-by Teng Long <[email protected]>--- trace2/tr2_tgt_event.c | 3 +++ trace2/tr2_tgt_normal.c | 5 ++++- trace2/tr2_tgt_perf.c | 9 +++++++-- 3 files changed, 14 insertions(+), 3 deletions(-)trace2/tr2_tgt_event.c +3
-0@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);jw_object_begin (&jw,0 );event_fmt_prepare (event_name, file, line,NULL , &jw); +jw_object_string (&jw,"scope" , scope_name);jw_object_string (&jw,"param" , param);jw_object_string (&jw,"value" , value);jw_end (&jw);trace2/tr2_tgt_normal.c +4
-1@@ -298,8 +298,11 @@ static void fn_param_fl(const char *file, int line, const char *param, const char *value) {struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope); -strbuf_addf (&buf_payload,"def_param %s=%s" , param, value); +strbuf_addf (&buf_payload,"def_param scope:%s %s=%s" , scope_name, param, + value);normal_io_write_fl (file, line, &buf_payload);strbuf_release (&buf_payload); }trace2/tr2_tgt_perf.c +4
-0@@ -441,12 +441,17 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct strbuf buf_payload =STRBUF_INIT ; +struct strbuf scope_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);strbuf_addf (&buf_payload,"%s:%s" , param, value); +strbuf_addf (&scope_payload,"%s:%s" ,"scope" , scope_name);
trace2/tr2_tgt_perf.c static void fn_param_fl(const char *file, int line, const char *param, · line 457
Quoted 10 lines
trace2/tr2_tgt_perf.c +3
-2-perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL ,NULL , - &buf_payload); +perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL , + scope_payload.buf, &buf_payload);strbuf_release (&buf_payload); +strbuf_release (&scope_payload); } static void fn_repo_fl(const char *file, int line,
Nicely done.
Thanks for your attention on this.
Jeff
Re: [PATCH v6 0/7] trace2: dump scope when print "interesting" config
In reply to Teng LongQuoted 9 lines · Teng Long
Since v5:
1. [1/7] New commit, pre-clean the issues of the strings.
2. [2/7] Optimize commit subject (the word "retrieve").
3. [4/7] New commit, do not ignore ENOENT silently when fail to open file.
4. [5/7] Replace "stat" to "fstat" in output string and let "cleanup" to return -1 instead of an unaccurate error tip as "cannot open".
5. [7/7] dump corresponding scope-name when "interesting" config is print to trace2 log.
I left some minor comments, another thing I'd just like to note, but it largely pre-dates this series, so we can leave it for now is that I found our mismatching reporting of the *.bitmap and/or *.pack kind of odd.
I.e. this bit (a diff on top):
diff --git a/pack-bitmap.c b/pack-bitmap.c
index 7e69093d5a8..8a8be5f6cad 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -381,23 +381,25 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
bitmap_name = pack_bitmap_filename(packfile);
fd = git_open(bitmap_name);
+ free(bitmap_name);
if (fd < 0) {
if (errno != ENOENT)
- warning("'%s' cannot open '%s'", strerror(errno), bitmap_name);
- free(bitmap_name);
+ warning_errno("cannot open bitmap for pack '%s'",
+ packfile->pack_name);
return -1;
}
- free(bitmap_name);
if (fstat(fd, &st)) {
close(fd);
- return error_errno(_("cannot fstat bitmap file"));
+ return error_errno(_("cannot fstat() bitmap file for pack '%s'"),
+ packfile->pack_name);
}
if (bitmap_git->pack || bitmap_git->midx) {
/* ignore extra bitmap file; we can only handle one */
warning(_("ignoring extra bitmap file: '%s'"), packfile->pack_name);
+ /* ^^ because here we refer to the pack file, not the bitmap file */
close(fd);
return -1;
}As tha shows we sometimes talk about the path for the pack when something is wrong with the bitmap, but other times the *.bitmap associated with the pack.
It would be a tad simpler to always talk about the *.pack, as we can free() the bitamp filename right away, but arguably we shoudl do it the other way around. After all we have issues with the bitmap file, let's mention it (or mention a filename at all).
Also a pre-existing issue: Why are we getting as far as opening a *.bitmap file, if we then find that we have "bitmap_git->pack || bitmap_git->midx" already, and then ignore it? Maybe I've missed something...
All of the same oddness applies to open_midx_bitmap_1, except with "path to the midx" in place of "*.pack" above.
Re: [PATCH v6 0/7] trace2: dump scope when print "interesting" config
Quoted 3 lines · Ævar Arnfjörð Bjarmason
I left some minor comments, another thing I'd just like to note, but it largely pre-dates this series, so we can leave it for now is that I found our mismatching reporting of the *.bitmap and/or *.pack kind of odd.
Ok.
Quoted 37 lines
I.e. this bit (a diff on top):
diff --git a/pack-bitmap.c b/pack-bitmap.c index 7e69093d5a8..8a8be5f6cad 100644 --- a/pack-bitmap.c +++ b/pack-bitmap.c @@ -381,23 +381,25 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_name = pack_bitmap_filename(packfile); fd = git_open(bitmap_name); + free(bitmap_name); if (fd < 0) { if (errno != ENOENT) - warning("'%s' cannot open '%s'", strerror(errno), bitmap_name); - free(bitmap_name); + warning_errno("cannot open bitmap for pack '%s'", + packfile->pack_name); return -1; } - free(bitmap_name); if (fstat(fd, &st)) { close(fd); - return error_errno(_("cannot fstat bitmap file")); + return error_errno(_("cannot fstat() bitmap file for pack '%s'"), + packfile->pack_name); } if (bitmap_git->pack || bitmap_git->midx) { /* ignore extra bitmap file; we can only handle one */ warning(_("ignoring extra bitmap file: '%s'"), packfile->pack_name); + /* ^^ because here we refer to the pack file, not the bitmap file */ close(fd); return -1; }
Quoted 3 lines
As tha shows we sometimes talk about the path for the pack when something is wrong with the bitmap, but other times the *.bitmap associated with the pack.
I think the reason is, the single-pack-bitmap filename based on the pack filename as "pack-$hash.bitmap" format and the multi-pack-bitmap filename based on the multi-pack-index as "multi-pack-index-$hash.bitmap" format. So, if want to open single-pack-bitmap, the corresponding packfile is needed. if want to open multi-pack-bitmap, the multi-pack-index file is needed, then initialize fields in "bitmap_index" like mmap, etc.
That'"s probably where these two methods and inputs come from and make the "path" different:
static int open_pack_bitmap_1(struct bitmap_index *bitmap_git,
struct packed_git *packfile)
static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
struct multi_pack_index *midx)
By the way, I think maybe the "multi-pack-index" filename maybe could change to "multi-pack-index-$hash.idx", it could map the current multi-pack-bitmap filename "multi-pack-index-$sha.bitmap", and ".idx" could be consistent with single-pack-index file. Maybe it's better I think, but it also should consider the backend compatibility.
Quoted 7 lines
Also a pre-existing issue: Why are we getting as far as opening a *.bitmap file, if we then find that we have "
bitmap_git->pack||bitmap_git->midx" already, and then ignore it? Maybe I've missed something...All of the same oddness applies to
open_midx_bitmap_1, except with "path to the midx" in place of "*.pack" above.
I'm not quite sure but in the code I think it's purposed to get the "fd" first by "open..()", then get the bitmap file size by "fstat()", finally "mmap" the bitmap file for some struct data initialization work.
Thanks.
v7 cover trace2: dump scope when print "interesting" config
In reply to Teng LongChanges since v6:
1. [1/7] Fixed the mistitled commit mesage.
[PATCH v6 1/7] "clean: fixed issues related to text output format"
to:
[PATCH v7 1/7] "pack-bitmap.c: fix formatting of error messages"
2. [4/7] replace "warning()" to "warning_errno()" and rewrite commit message.
3. [5/7] fix the logic error, move "error_errno()" before close(fd) to avoid errno lost.
4. [7/7] update Documentation/technical/api-trace2.txt here too.
Thanks.
Teng Long (7):
pack-bitmap.c: fix formatting of error messages pack-bitmap.c: mark more strings for translations pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: do not ignore error when opening a bitmap file pack-bitmap.c: using error() instead of silently returning -1 pack-bitmap.c: continue looping when first MIDX bitmap is found
tr2: dump names if config exist in multiple scopes
Documentation/technical/api-trace2.txt | 45 +++++++++++ pack-bitmap.c | 103 ++++++++++++++----------- trace2/tr2_tgt_event.c | 3 + trace2/tr2_tgt_normal.c | 5 +- trace2/tr2_tgt_perf.c | 9 ++- 5 files changed, 117 insertions(+), 48 deletions(-)
Range-diff against v6:
1: 94a64ba895 ! 1: 7426b277ba clean: fixed issues related to text output format
@@ Metadata
Author: Teng Long <[email protected]>
## Commit message ##
- clean: fixed issues related to text output format
+ pack-bitmap.c: fix formatting of error messages
There are some text output issues in 'pack-bitmap.c', they exist in die(), error() etc. This includes issues with capitalization the
2: cadecd6b84 = 2: 4e493426a0 pack-bitmap.c: mark more strings for translations
3: d8a2235cb0 = 3: 263f45ba96 pack-bitmap.c: rename "idx_name" to "bitmap_name"
4: 009cc49a18 ! 4: d11ea092d5 pack-bitmap.c: don't ignore ENOENT silently
@@ Metadata
Author: Teng Long <[email protected]>
## Commit message ##
- pack-bitmap.c: don't ignore ENOENT silently
+ pack-bitmap.c: do not ignore error when opening a bitmap file
- When finished call git_open(), instead of ignoring ENOENT silently
- and return error_errno(_("cannot stat...")), it's better to check
- the ENOENT before then output the warning.
+ Calls to git_open() to open the pack bitmap file and
+ multi-pack bitmap file do not report any error when they
+ fail. These files are optional and it is not an error if
+ open failed due to ENOENT, but we shouldn't be ignoring
+ other kinds of errors.
Signed-off-by: Teng Long <[email protected]>
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
- if (fd < 0)
+ if (fd < 0) {
+ if (errno != ENOENT)
-+ warning("'%s' cannot open '%s'", strerror(errno), bitmap_name);
++ warning_errno("cannot open '%s'", bitmap_name);
+ free(bitmap_name);
return -1;
+ }
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, st
- if (fd < 0)
+ if (fd < 0) {
+ if (errno != ENOENT)
-+ warning("'%s' cannot open '%s'", strerror(errno), bitmap_name);
++ warning_errno("cannot open '%s'", bitmap_name);
+ free(bitmap_name);
return -1;
+ }
5: 52783555e2 ! 5: f60efe78d6 pack-bitmap.c: using error() instead of silently returning -1
@@ Commit message
## pack-bitmap.c ##
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
+ free(bitmap_name);
if (fstat(fd, &st)) {
++ error_errno(_("cannot fstat bitmap file"));
close(fd);
-- return -1;
-+ return error_errno(_("cannot fstat bitmap file"));
+ return -1;
}
-
- if (bitmap_git->pack || bitmap_git->midx) {
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
if (load_bitmap_header(bitmap_git) < 0)
goto cleanup;
@@ pack-bitmap.c: static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,
if (load_midx_revindex(bitmap_git->midx) < 0) {
warning(_("multi-pack bitmap is missing required reverse index"));
@@ pack-bitmap.c: static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git
+ free(bitmap_name);
if (fstat(fd, &st)) {
++ error_errno(_("cannot fstat bitmap file"));
close(fd);
-- return -1;
-+ return error_errno(_("cannot fstat bitmap file"));
+ return -1;
}
-
- if (bitmap_git->pack || bitmap_git->midx) {
6: 95832190b8 = 6: 83090308ad pack-bitmap.c: continue looping when first MIDX bitmap is found
7: c45ead51ff ! 7: a01ae8478d tr2: dump names if config exist in multiple scopes@@ Commit message
+ ## Documentation/technical/api-trace2.txt ##
+@@ Documentation/technical/api-trace2.txt: at offset 508.
+ This example also shows that thread names are assigned in a racy manner
+ as each thread starts and allocates TLS storage.
+
++Print Configs::
++
++ Dump "interesting" config values to trace2 log.
+++
++The environment variable GIT_TRACE2_CONFIG_PARAMS and configuration ++trace2.configparams can be used to output config values which you care ++about. For example, assume that we want to config different color.ui
++values in multiple scopes, such as:
+++
++----------------
++$ git config --system color.ui never
++$ git config --global color.ui always
++$ git config --local color.ui auto
++$ git config --list --show-scope | grep 'color.ui'
++system color.ui=never
++global color.ui=always
++local color.ui=auto
++----------------
+++
++Then, mark the config color.ui as "interesting" config with ++GIT_TRACE2_CONFIG_PARAMS:
+++
++----------------
++$ export GIT_TRACE2_PERF_BRIEF=1
++$ export GIT_TRACE2_PERF=~/log.perf
++$ export GIT_TRACE2_CONFIG_PARAMS=color.ui
++$ git version
++...
++$ cat ~/log.perf
++d0 | main | version | | | | | ...
++d0 | main | start | | 0.000284 | | | /opt/git/master/bin/git version
++d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
++d0 | main | cmd_name | | | | | version (version)
++d0 | main | exit | | 0.000419 | | | code:0
++d0 | main | atexit | | 0.000426 | | | code:0
++d0 | main | version | | | | | ...
++d0 | main | start | | 0.000275 | | | /opt/git/master/bin/git version
++d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
++d0 | main | cmd_name | | | | | version (version)
++d0 | main | def_param | | | | | color.ui:never
++d0 | main | def_param | | | | | color.ui:always
++d0 | main | def_param | | | | | color.ui:auto
++d0 | main | exit | | 0.000543 | | | code:0
++d0 | main | atexit | | 0.000549 | | | code:0
++----------------
+ == Future Work
+
+ === Relationship to the Existing Trace Api (api-trace.txt)
+
## trace2/tr2_tgt_event.c ##
@@ trace2/tr2_tgt_event.c: static void fn_param_fl(const char *file, int line, const char *param,
{
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 1/7 pack-bitmap.c: fix formatting of error messages
There are some text output issues in 'pack-bitmap.c', they exist in die(), error() etc. This includes issues with capitalization the first letter, newlines, error() instead of BUG(), and substitution that don't have quotes around them.
--- pack-bitmap.c | 47 ++++++++++++++++++++++++----------------------- 1 file changed, 24 insertions(+), 23 deletions(-)
pack-bitmap.c +24 -23
@@ -1,5 +1,6 @@#include "cache.h" #include "commit.h" +#include "strbuf.h" #include "tag.h" #include "diff.h" #include "revision.h"
@@ -138,7 +139,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("Failed to load bitmap index (corrupted?)" ); +error ("failed to load bitmap index (corrupted?)" );ewah_pool_free (b);return NULL ; }
@@ -160,14 +161,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("Corrupted bitmap index (too small)" ); +return error ("corrupted bitmap index (too small)" );if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("Corrupted bitmap index file (wrong header)" ); +return error ("corrupted bitmap index file (wrong header)" ); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("Unsupported version for bitmap index file (%d)" , index->version); +return error ("unsupported version '%d' for bitmap index file" , index->version);/* Parse known bitmap format options */ {
@@ -176,7 +177,7 @@ static int load_bitmap_header(struct bitmap_index *index)unsigned char *index_end = index->map + index->map_size - the_hash_algo->rawsz;if ((flags &BITMAP_OPT_FULL_DAG ) ==0 ) -return error ("Unsupported options for bitmap index file " +BUG ("unsupported options for bitmap index file " "(Git requires BITMAP_OPT_FULL_DAG)" );if (flags &BITMAP_OPT_HASH_CACHE ) {
@@ -215,7 +216,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("Duplicate entry in bitmap index: %s" ,oid_to_hex (oid)); +error ("duplicate entry in bitmap index: '%s'" ,oid_to_hex (oid));return NULL ; }
@@ -274,13 +275,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("Corrupted bitmap pack index" ); +return error ("corrupted bitmap pack index" );if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("Invalid XOR offset in bitmap pack index" ); +return error ("invalid XOR offset in bitmap pack index" ); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -330,7 +331,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning ("ignoring extra bitmap file: '%s'" , buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +388,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning ("ignoring extra bitmap file: '%s'" , packfile->pack_name);close (fd);return -1 ; }
@@ -1628,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object %s not found in type bitmaps" , +die ("object '%s' not found in type bitmaps" ,oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object %s does not have a unique type" , +die ("object '%s' does not have a unique type" ,oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object %s: real type %s, expected: %s" , +die ("object '%s': real type '%s', expected: '%s'" ,oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1650,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&object->oid)); +die ("object not in bitmap: '%s'" ,oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1665,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&commit->object.oid)); +die ("object not in bitmap: '%s'" ,oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1687,21 +1688,21 @@ void test_bitmap_walk(struct rev_info *revs)if (revs->pending.nr !=1 )die ("you must specify exactly one commit to test" ); -fprintf (stderr,"Bitmap v%d test (%d entries loaded)\n" , +fprintf_ln (stderr,"Bitmap v%d test (%d entries loaded)" , bitmap_git->version, bitmap_git->entry_count); root = revs->pending.objects[0 ].item; bm =bitmap_for_commit (bitmap_git, (struct commit *)root);if (bm) { -fprintf (stderr,"Found bitmap for %s. %d bits / %08x checksum\n" , +fprintf_ln (stderr,"Found bitmap for '%s'. %d bits / %08x checksum" ,oid_to_hex (&root->oid), (int )bm->bit_size,ewah_checksum (bm)); result =ewah_to_bitmap (bm); }if (result ==NULL ) -die ("Commit %s doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die ("commit '%s' doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1726,7 +1727,7 @@ void test_bitmap_walk(struct rev_info *revs)stop_progress (&tdata.prg);if (bitmap_equals (result, tdata.base)) -fprintf (stderr,"OK!\n" ); +fprintf_ln (stderr,"OK!" );else die ("mismatch in bitmap results" );
@@ -1749,7 +1750,7 @@ int test_bitmap_commits(struct repository *r)die ("failed to load bitmap indexes" );kh_foreach (bitmap_git->bitmaps, oid, value, { -printf ("%s\n" ,oid_to_hex (&oid)); +printf_ln ("%s" ,oid_to_hex (&oid)); });free_bitmap_index (bitmap_git);
@@ -1774,7 +1775,7 @@ int test_bitmap_hashes(struct repository *r)nth_bitmap_object_oid (bitmap_git, &oid, index_pos); -printf ("%s %" PRIu32"\n" , +printf_ln ("%s %" PRIu32"" ,oid_to_hex (&oid),get_be32 (bitmap_git->hashes + index_pos)); }
@@ -1936,7 +1937,7 @@ static off_t get_disk_usage_for_type(struct bitmap_index *bitmap_git,struct object_id oid;nth_midxed_object_oid (&oid, bitmap_git->midx, midx_pos); -die (_ ("could not find %s in pack %s at offset %" PRIuMAX), +die (_ ("could not find '%s' in pack '%s' at offset %" PRIuMAX),oid_to_hex (&oid), pack->pack_name, (uintmax_t )offset);
@@ -1972,7 +1973,7 @@ static off_t get_disk_usage_for_extended(struct bitmap_index *bitmap_git)continue ;if (oid_object_info_extended (the_repository, &obj->oid, &oi,0 ) <0 ) -die (_ ("unable to get disk usage of %s" ), +die (_ ("unable to get disk usage of '%s'" ),oid_to_hex (&obj->oid)); total += object_size;
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 2/7 pack-bitmap.c: mark more strings for translations
In reply to Teng LongIn pack-bitmap.c, some printed texts are translated, some are not. Let's support the translations of the bitmap related output.
--- pack-bitmap.c | 48 ++++++++++++++++++++++++------------------------ 1 file changed, 24 insertions(+), 24 deletions(-)
pack-bitmap.c +24 -24
@@ -139,7 +139,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("failed to load bitmap index (corrupted?)" ); +error (_ ("failed to load bitmap index (corrupted?)" ));ewah_pool_free (b);return NULL ; }
@@ -161,14 +161,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("corrupted bitmap index (too small)" ); +return error (_ ("corrupted bitmap index (too small)" ));if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("corrupted bitmap index file (wrong header)" ); +return error (_ ("corrupted bitmap index file (wrong header)" )); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("unsupported version '%d' for bitmap index file" , index->version); +return error (_ ("unsupported version '%d' for bitmap index file" ), index->version);/* Parse known bitmap format options */ {
@@ -182,7 +182,7 @@ static int load_bitmap_header(struct bitmap_index *index)if (flags &BITMAP_OPT_HASH_CACHE ) {if (cache_size > index_end - index->map - header_size) -return error ("corrupted bitmap index file (too short to fit hash cache)" ); +return error (_ ("corrupted bitmap index file (too short to fit hash cache)" )); index->hashes = (void *)(index_end - cache_size); index_end -= cache_size; }
@@ -216,7 +216,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("duplicate entry in bitmap index: '%s'" ,oid_to_hex (oid)); +error (_ ("duplicate entry in bitmap index: '%s'" ),oid_to_hex (oid));return NULL ; }
@@ -260,14 +260,14 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)struct object_id oid;if (index->map_size - index->map_pos <6 ) -return error ("corrupt ewah bitmap: truncated header for entry %d" , i); +return error (_ ("corrupt ewah bitmap: truncated header for entry %d" ), i); commit_idx_pos =read_be32 (index->map, &index->map_pos); xor_offset =read_u8 (index->map, &index->map_pos); flags =read_u8 (index->map, &index->map_pos);if (nth_bitmap_object_oid (index, &oid, commit_idx_pos) <0 ) -return error ("corrupt ewah bitmap: commit index %u out of range" , +return error (_ ("corrupt ewah bitmap: commit index %u out of range" ), (unsigned )commit_idx_pos); bitmap =read_bitmap_1 (index);
@@ -275,13 +275,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("corrupted bitmap pack index" ); +return error (_ ("corrupted bitmap pack index" ));if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("invalid XOR offset in bitmap pack index" ); +return error (_ ("invalid XOR offset in bitmap pack index" )); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -331,7 +331,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: '%s'" , buf.buf); +warning (_ ("ignoring extra bitmap file: '%s'" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -388,7 +388,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: '%s'" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: '%s'" ), packfile->pack_name);close (fd);return -1 ; }
@@ -819,7 +819,7 @@ static struct bitmap *find_objects(struct bitmap_index *bitmap_git, revs->include_check_data = &incdata;if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); show_data.bitmap_git = bitmap_git; show_data.base = base;
@@ -1629,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object '%s' not found in type bitmaps" , +die (_ ("object '%s' not found in type bitmaps" ),oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object '%s' does not have a unique type" , +die (_ ("object '%s' does not have a unique type" ),oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object '%s': real type '%s', expected: '%s'" , +die (_ ("object '%s': real type '%s', expected: '%s'" ),oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1651,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("object not in bitmap: '%s'" ,oid_to_hex (&object->oid)); +die (_ ("object not in bitmap: '%s'" ),oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1666,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("object not in bitmap: '%s'" ,oid_to_hex (&commit->object.oid)); +die (_ ("object not in bitmap: '%s'" ),oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1683,10 +1683,10 @@ void test_bitmap_walk(struct rev_info *revs)struct ewah_bitmap *bm;if (!(bitmap_git =prepare_bitmap_git (revs->repo))) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" ));if (revs->pending.nr !=1 ) -die ("you must specify exactly one commit to test" ); +die (_ ("you must specify exactly one commit to test" ));fprintf_ln (stderr,"Bitmap v%d test (%d entries loaded)" , bitmap_git->version, bitmap_git->entry_count);
@@ -1702,7 +1702,7 @@ void test_bitmap_walk(struct rev_info *revs) }if (result ==NULL ) -die ("commit '%s' doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die (_ ("commit '%s' doesn't have an indexed bitmap" ),oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1711,7 +1711,7 @@ void test_bitmap_walk(struct rev_info *revs) result_popcnt =bitmap_popcount (result);if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); tdata.bitmap_git = bitmap_git; tdata.base =bitmap_new ();
@@ -1729,7 +1729,7 @@ void test_bitmap_walk(struct rev_info *revs)if (bitmap_equals (result, tdata.base))fprintf_ln (stderr,"OK!" );else -die ("mismatch in bitmap results" ); +die (_ ("mismatch in bitmap results" ));bitmap_free (result);bitmap_free (tdata.base);
@@ -1747,7 +1747,7 @@ int test_bitmap_commits(struct repository *r) MAYBE_UNUSED void *value;if (!bitmap_git) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" )); kh_foreach(bitmap_git->bitmaps, oid, value, { printf_ln("%s", oid_to_hex(&oid));
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 3/7 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index, let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -314,10 +314,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -369,14 +369,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 4/7 pack-bitmap.c: do not ignore error when opening a bitmap file
In reply to Teng LongCalls to git_open() to open the pack bitmap file and multi-pack bitmap file do not report any error when they fail. These files are optional and it is not an error if open failed due to ENOENT, but we shouldn't be ignoring other kinds of errors.
--- pack-bitmap.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-)
pack-bitmap.c +12 -5
@@ -317,10 +317,13 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,char *bitmap_name =midx_bitmap_filename (midx);int fd =git_open (bitmap_name); -free (bitmap_name); - -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning_errno ("cannot open '%s'" , bitmap_name); +free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
@@ -376,10 +379,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_name =pack_bitmap_filename (packfile); fd =git_open (bitmap_name); -free (bitmap_name); -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning_errno ("cannot open '%s'" , bitmap_name); +free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 5/7 pack-bitmap.c: using error() instead of silently returning -1
In reply to Teng LongIn "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
pack-bitmap.c +5 -1
@@ -326,6 +326,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,free (bitmap_name);if (fstat (fd, &st)) { +error_errno (_ ("cannot fstat bitmap file" ));close (fd);return -1 ; }
@@ -350,8 +351,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (load_bitmap_header (bitmap_git) <0 ) goto cleanup; -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +error (_ ("checksum doesn't match in MIDX and bitmap" )); goto cleanup; + }if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" ));
@@ -389,6 +392,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitfree (bitmap_name);if (fstat (fd, &st)) { +error_errno (_ ("cannot fstat bitmap file" ));close (fd);return -1 ; }
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 6/7 pack-bitmap.c: continue looping when first MIDX bitmap is found
In reply to Teng LongIn "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to continue the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The reason for this is, there may exist more than one MIDX file in a repo. The "multi_pack_index" struct is actually designed as a singly linked list, and if a MIDX file has been already opened successfully, then the other MIDX files will be skipped and left with a warning "ignoring extra bitmap file." to the output.
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -506,15 +506,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v7 7/7 tr2: dump names if config exist in multiple scopes
In reply to Teng LongWhen we specify GIT_TRACE2_CONFIG_PARAMS or trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):
...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false
As the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \
Quoted 2 lines
GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \
git rev-list --test-bitmap HEAD"
The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ... def_param scope:system core.multipackindex=false ... git.c:461 ... def_param scope:global core.multipackindex=false ... git.c:461 ... def_param scope:local core.multipackindex=false
Format perf:
... | def_param | ... | scope:system | core.multipackindex:false ... | def_param | ... | scope:global | core.multipackindex:false ... | def_param | ... | scope:local | core.multipackindex:false
Format event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
--- Documentation/technical/api-trace2.txt | 45 ++++++++++++++++++++++++++ trace2/tr2_tgt_event.c | 3 ++ trace2/tr2_tgt_normal.c | 5 ++- trace2/tr2_tgt_perf.c | 9 ++++-- 4 files changed, 59 insertions(+), 3 deletions(-)
Documentation/technical/api-trace2.txt +45 -0
@@ -1207,6 +1207,51 @@ at offset 508. This example also shows that thread names are assigned in a racy manner as each thread starts and allocates TLS storage. +Print Configs:: + + Dump "interesting" config values to trace2 log. ++ +The environment variable `GIT_TRACE2_CONFIG_PARAMS` and configuration +`trace2.configparams` can be used to output config values which you care +about. For example, assume that we want to config different `color.ui` +values in multiple scopes, such as: ++ +---------------- +$ git config --system color.ui never +$ git config --global color.ui always +$ git config --local color.ui auto +$ git config --list --show-scope | grep 'color.ui' +system color.ui=never +global color.ui=always +local color.ui=auto +---------------- ++ +Then, mark the config `color.ui` as "interesting" config with +`GIT_TRACE2_CONFIG_PARAMS`: ++ +---------------- +$ export GIT_TRACE2_PERF_BRIEF=1 +$ export GIT_TRACE2_PERF=~/log.perf +$ export GIT_TRACE2_CONFIG_PARAMS=color.ui +$ git version +... +$ cat ~/log.perf +d0 | main | version | | | | | ... +d0 | main | start | | 0.000284 | | | /opt/git/master/bin/git version +d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] +d0 | main | cmd_name | | | | | version (version) +d0 | main | exit | | 0.000419 | | | code:0 +d0 | main | atexit | | 0.000426 | | | code:0 +d0 | main | version | | | | | ... +d0 | main | start | | 0.000275 | | | /opt/git/master/bin/git version +d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] +d0 | main | cmd_name | | | | | version (version) +d0 | main | def_param | | | | | color.ui:never +d0 | main | def_param | | | | | color.ui:always +d0 | main | def_param | | | | | color.ui:auto +d0 | main | exit | | 0.000543 | | | code:0 +d0 | main | atexit | | 0.000549 | | | code:0 +---------------- == Future Work === Relationship to the Existing Trace Api (api-trace.txt)
trace2/tr2_tgt_event.c +3 -0
@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);jw_object_begin (&jw,0 );event_fmt_prepare (event_name, file, line,NULL , &jw); +jw_object_string (&jw,"scope" , scope_name);jw_object_string (&jw,"param" , param);jw_object_string (&jw,"value" , value);jw_end (&jw);
trace2/tr2_tgt_normal.c +4 -1
@@ -298,8 +298,11 @@ static void fn_param_fl(const char *file, int line, const char *param, const char *value) {struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope); -strbuf_addf (&buf_payload,"def_param %s=%s" , param, value); +strbuf_addf (&buf_payload,"def_param scope:%s %s=%s" , scope_name, param, + value);normal_io_write_fl (file, line, &buf_payload);strbuf_release (&buf_payload); }
trace2/tr2_tgt_perf.c +7 -2
@@ -441,12 +441,17 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct strbuf buf_payload =STRBUF_INIT ; +struct strbuf scope_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);strbuf_addf (&buf_payload,"%s:%s" , param, value); +strbuf_addf (&scope_payload,"%s:%s" ,"scope" , scope_name); -perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL ,NULL , - &buf_payload); +perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL , + scope_payload.buf, &buf_payload);strbuf_release (&buf_payload); +strbuf_release (&scope_payload); } static void fn_repo_fl(const char *file, int line,
Signature · 1 line
2.35.0.rc0.679.gc613175da2
Re: [PATCH v7 7/7] tr2: dump names if config exist in multiple scopes
Quoted 86 lines · Teng Long
When we specify
GIT_TRACE2_CONFIG_PARAMSor trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:false ...| def_param | ... | core.multipackindex:falseAs the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \ > GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \ > git rev-list --test-bitmap HEAD"The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ...def_paramscope:system core.multipackindex=false ... git.c:461 ...def_paramscope:global core.multipackindex=false ... git.c:461 ...def_paramscope:local core.multipackindex=falseFormat perf:
... | def_param | ... | scope:system | core.multipackindex:false ... | def_param | ... | scope:global | core.multipackindex:false ... | def_param | ... | scope:local | core.multipackindex:falseFormat event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
Signed-off-by Teng Long <[email protected]>--- Documentation/technical/api-trace2.txt | 45 ++++++++++++++++++++++++++ trace2/tr2_tgt_event.c | 3 ++ trace2/tr2_tgt_normal.c | 5 ++- trace2/tr2_tgt_perf.c | 9 ++++-- 4 files changed, 59 insertions(+), 3 deletions(-)Documentation/technical/api-trace2.txt +32
-0@@ -1207,6 +1207,51 @@ at offset 508. This example also shows that thread names are assigned in a racy manner as each thread starts and allocates TLS storage. +Print Configs:: + + Dump "interesting" config values to trace2 log. ++ +The environment variable `GIT_TRACE2_CONFIG_PARAMS` and configuration +`trace2.configparams` can be used to output config values which you care +about. For example, assume that we want to config different `color.ui` +values in multiple scopes, such as: ++ +---------------- +$ git config --system color.ui never +$ git config --global color.ui always +$ git config --local color.ui auto +$ git config --list --show-scope | grep 'color.ui' +system color.ui=never +global color.ui=always +local color.ui=auto +---------------- ++ +Then, mark the config `color.ui` as "interesting" config with +`GIT_TRACE2_CONFIG_PARAMS`: ++ +---------------- +$ export GIT_TRACE2_PERF_BRIEF=1 +$ export GIT_TRACE2_PERF=~/log.perf +$ export GIT_TRACE2_CONFIG_PARAMS=color.ui +$ git version +... +$ cat ~/log.perf +d0 | main | version | | | | | ... +d0 | main | start | | 0.000284 | | | /opt/git/master/bin/git version +d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd]
Documentation/technical/api-trace2.txt at offset 508. · line 1251
Quoted 10 lines
Documentation/technical/api-trace2.txt +10
-0+d0 | main | cmd_name | | | | | version (version) +d0 | main | exit | | 0.000419 | | | code:0 +d0 | main | atexit | | 0.000426 | | | code:0 +d0 | main | version | | | | | ... +d0 | main | start | | 0.000275 | | | /opt/git/master/bin/git version +d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] +d0 | main | cmd_name | | | | | version (version) +d0 | main | def_param | | | | | color.ui:never +d0 | main | def_param | | | | | color.ui:always +d0 | main | def_param | | | | | color.ui:auto
shouldn't this example in the docs have the scope value in the next to last column as you noted in the cover letter ?
Documentation/technical/api-trace2.txt at offset 508. · line 1255
Quoted 4 lines
Documentation/technical/api-trace2.txt +3
-0+d0 | main | exit | | 0.000543 | | | code:0 +d0 | main | atexit | | 0.000549 | | | code:0 +---------------- == Future Work
...
Otherwise, this all LGTM.
Thanks
Jeff
Re: [PATCH v7 7/7] tr2: dump names if config exist in multiple scopes
Quoted 2 lines · Jeff Hostetler
shouldn't this example in the docs have the scope value in the next to last column as you noted in the cover letter ?
oops, sorry for that. I will send a quick fix today.
Thanks.
Re: [PATCH v7 7/7] tr2: dump names if config exist in multiple scopes
In reply to Teng LongQuoted 1 line · Teng Long
Subject: Re: [PATCH v7 7/7] tr2: dump names if config exist in multiple scopes
This make it sound as if we do not show the scope if a configuration variable appears only in one scope, but I do not see how the updated code achieves that. Instead, it seems that it shows the scope unconditionally in addition to the key-value pair.
If that is the case, a retitle is in order. Also this should probably be a separate topic, unrelated from the other 6 patches.
Re: [PATCH v7 7/7] tr2: dump names if config exist in multiple scopes
Quoted 4 lines · Junio C Hamano
This make it sound as if we do not show the scope if a configuration variable appears only in one scope, but I do not see how the updated code achieves that. Instead, it seems that it shows the scope unconditionally in addition to the key-value pair.
Yes. There is a problem with this statement. I will fix it, maybe like:
tr2: print corresponding scope with config key-value
Quoted 2 lines
If that is the case, a retitle is in order. Also this should probably be a separate topic, unrelated from the other 6 patches.
That's the case. I will send the [0-6] as the v8, and separate [7/7] as a new patch series.
Thanks.
Re: [PATCH v7 0/7] trace2: dump scope when print "interesting" config
In reply to Teng LongQuoted 19 lines · Teng Long
Changes since v6:
1. [1/7] Fixed the mistitled commit mesage.
[PATCH v6 1/7] "clean: fixed issues related to text output format"
to:
[PATCH v7 1/7] "
pack-bitmap.c: fix formatting of error messages"2. [4/7] replace "
warning()" to "warning_errno()" and rewrite commit message.3. [5/7] fix the logic error, move "
error_errno()" before close(fd) to avoid errno lost.4. [7/7] update
Documentation/technical/api-trace2.txthere too.Thanks.
Quoted 8 lines
Teng Long (7):
pack-bitmap.c: fix formatting of error messagespack-bitmap.c: mark more strings for translationspack-bitmap.c: rename "idx_name" to "bitmap_name"pack-bitmap.c: do not ignore error when opening a bitmap filepack-bitmap.c: usingerror()instead of silently returning -1pack-bitmap.c: continue looping when first MIDX bitmap is found
tr2: dump names if config exist in multiple scopes
This has been "cooking" on the list for quite some time and I found that all parts that I had comments on earlier iterations are now in good shape.
Fellow reviewers, how does this round look? The only gripe I have is that the last one seems totally disconnected from the rest, but That's minor.
Thanks.
Re: [PATCH v7 0/7] trace2: dump scope when print "interesting" config
Quoted 30 lines · Junio C Hamano
Quoted 19 lines · Teng Long
Changes since v6:
1. [1/7] Fixed the mistitled commit mesage.
[PATCH v6 1/7] "clean: fixed issues related to text output format"
to:
[PATCH v7 1/7] "
pack-bitmap.c: fix formatting of error messages"2. [4/7] replace "
warning()" to "warning_errno()" and rewrite commit message.3. [5/7] fix the logic error, move "
error_errno()" before close(fd) to avoid errno lost.4. [7/7] update
Documentation/technical/api-trace2.txthere too.Thanks.
Quoted 8 lines
Teng Long (7):
pack-bitmap.c: fix formatting of error messagespack-bitmap.c: mark more strings for translationspack-bitmap.c: rename "idx_name" to "bitmap_name"pack-bitmap.c: do not ignore error when opening a bitmap filepack-bitmap.c: usingerror()instead of silently returning -1pack-bitmap.c: continue looping when first MIDX bitmap is found
tr2: dump names if config exist in multiple scopes
Quoted 7 lines
This has been "cooking" on the list for quite some time and I found that all parts that I had comments on earlier iterations are now in good shape.
Fellow reviewers, how does this round look? The only gripe I have is that the last one seems totally disconnected from the rest, but That's minor.
Yeah likewise, it even applies directly on master. But I can live with it :)
One minor nit is that something like this (which needs to be fleshened up) should be fixed up into 7/7 (and maybe we want to keep the "..."?):
Documentation/technical/api-trace2.txt +1 -1
@@ -716,7 +716,7 @@ The "exec_id" field is a command-unique id and is only useful if the ------------ { "event":"def_param", - ... + "scope": ... "param":"core.abbrev", "value":"7" }
And that the addition to api-trace2.txt seems to partially be something that should just link to Documentation/config/trace2.txt, i.e. it's generally documenting an existing facility.
I think it would be great in any case to have that 7/7 split into what we do now & docs for that, and then the minor addition of "scope".
The rest all looks good to me, and seems to fully address the feedback in previous rounds.
Re: [PATCH v7 0/7] trace2: dump scope when print "interesting" config
Quoted 2 lines
Yeah likewise, it even applies directly on master. But I can live with it :)
Sorry, may I ask what's the recommanded way to deal this, alway use tag or master on update?
Quoted 8 lines
One minor nit is that something like this (which needs to be fleshened up) should be fixed up into 7/7 (and maybe we want to keep the "..."?):
Documentation/technical/api-trace2.txt +0
-0
Documentation/technical/api-trace2.txt The "exec_id" field is a command-unique id and is only useful if the · line 722
Quoted 8 lines
Documentation/technical/api-trace2.txt +1
-1@@ -716,7 +716,7 @@ The "exec_id" field is a command-unique id and is only useful if the ------------ { "event":"def_param", - ... + "scope": ... "param":"core.abbrev", "value":"7" }
I do a little check, the json format in this scenario is like :
{
"event": "def_param",
"sid": "20220719T075535.279369Z-H1b0a19dc-P000093db", "thread": "main",
"time": "2022-07-19T07:55:35.280720Z", "file": "git.c",
"line": 461,
"scope": "global",
"param": "color.ui",
"value": "always"
}
So, I think this is ok maybe:
Documentation/technical/api-trace2.txt +1 -0
@@ -717,6 +717,7 @@ The "exec_id" field is a command-unique id and is only useful if the { "event":"def_param", ... + "scope":... "param":"core.abbrev", "value":"7" }
Quoted 3 lines
And that the addition to api-trace2.txt seems to partially be something that should just link to
Documentation/config/trace2.txt, i.e. it's generally documenting an existing facility.
Do you mean the modification about trace2.configParams and you suggest to make a link to Documentation/config/trace2.txt(actually as git-config[1]) in /Documentation/technical/api-trace2.txt?
Quoted 2 lines
I think it would be great in any case to have that 7/7 split into what we do now & docs for that, and then the minor addition of "scope".
Let me try to understand this, it's better to split [7/7] into two commits.
First commit is to add docs in /Documentation/technical/api-trace2.txt to let reader to find the print-config ability, by the way link the doc about the GIT_TRACE2_CONFIG_PARAM and trace2.configparami in git-config[1].
Like:
commit 2db47572d4462e3788a92fd355b97df13b9bcc39
Author: Teng Long <[email protected]> Date: Tue Jul 19 17:30:35 2022 +0800
api-trace2.txt: add docs to print config
It's supported to print "interesting" config value to tr2 log by setting the "GIT_TRACE2_CONFIG_PARAMS" environment variable and the "trace2.configparam" config, let's add the related docs.
Documentation/technical/api-trace2.txt +39 -0
@@ -1207,6 +1207,45 @@ at offset 508. This example also shows that thread names are assigned in a racy manner as each thread starts and allocates TLS storage. +Print Configs:: + + Dump "interesting" config values to trace2 log. ++ +The environment variable `GIT_TRACE2_CONFIG_PARAMS` and configuration +`trace2.configparams` can be used to output config values which you care +about(see linkgit:git-config[1). For example, assume that we want to +config different `color.ui` values in multiple scopes, such as: ++ +---------------- +$ git config --system color.ui never +$ git config --global color.ui always +$ git config --local color.ui auto +$ git config --list --show-scope | grep 'color.ui' +system color.ui=never +global color.ui=always +local color.ui=auto +---------------- ++ +Then, mark the config `color.ui` as "interesting" config with +`GIT_TRACE2_CONFIG_PARAMS`: ++ +---------------- +$ export GIT_TRACE2_PERF_BRIEF=1 +$ export GIT_TRACE2_PERF=~/log.perf +$ export GIT_TRACE2_CONFIG_PARAMS=color.ui +$ git version +... +$ cat ~/log.perf +d0 | main | version | | | | | ... +d0 | main | start | | 0.000260 | | | /opt/git/master/bin/git version +d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] +d0 | main | cmd_name | | | | | version (version) +d0 | main | def_param | | | | | color.ui:never +d0 | main | def_param | | | | | color.ui:always +d0 | main | def_param | | | | | color.ui:auto +d0 | main | exit | | 0.000470 | | | code:0 +d0 | main | atexit | | 0.000477 | | | code:0 +---------------- == Future Work
Second is actually what we do in [7/7], modify to support print scope names and update the docs which based on what we just add in the first commit
Like:
commit 615ab4864fce6b8042778aa78799ac2656785710 (HEAD -> tl/bitmap-append-trace2-outputs)
Author: Teng Long <[email protected]> Date: Thu Jul 7 21:48:32 2022 +0800
tr2: dump names if config exist in multiple scopes
When we specify GIT_TRACE2_CONFIG_PARAMS or trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):
...| def_param | ... | core.multipackindex:false
...| def_param | ... | core.multipackindex:false
...| def_param | ... | core.multipackindex:false
As the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \
> GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \
> git rev-list --test-bitmap HEAD"
The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ... def_param scope:system core.multipackindex=false ... git.c:461 ... def_param scope:global core.multipackindex=false ... git.c:461 ... def_param scope:local core.multipackindex=false
Format perf:
... | def_param | ... | scope:system | core.multipackindex:false
... | def_param | ... | scope:global | core.multipackindex:false
... | def_param | ... | scope:local | core.multipackindex:false
Format event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
Documentation/technical/api-trace2.txt +4 -3
@@ -717,6 +717,7 @@ The "exec_id" field is a command-unique id and is only useful if the { "event":"def_param", ... + "scope":... "param":"core.abbrev", "value":"7" }
@@ -1240,9 +1241,9 @@ d0 | main | version | | | | d0 | main | start | | 0.000260 | | | /opt/git/master/bin/git version d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] d0 | main | cmd_name | | | | | version (version) -d0 | main | def_param | | | | | color.ui:never -d0 | main | def_param | | | | | color.ui:always -d0 | main | def_param | | | | | color.ui:auto +d0 | main | def_param | | | | scope:system | color.ui:never +d0 | main | def_param | | | | scope:global | color.ui:always +d0 | main | def_param | | | | scope:local | color.ui:auto d0 | main | exit | | 0.000470 | | | code:0 d0 | main | atexit | | 0.000477 | | | code:0 ----------------
trace2/tr2_tgt_event.c +3 -0
@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);jw_object_begin (&jw,0 );event_fmt_prepare (event_name, file, line,NULL , &jw); +jw_object_string (&jw,"scope" , scope_name);jw_object_string (&jw,"param" , param);jw_object_string (&jw,"value" , value);jw_end (&jw);
trace2/tr2_tgt_normal.c +4 -1
@@ -298,8 +298,11 @@ static void fn_param_fl(const char *file, int line, const char *param, const char *value) {struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope); -strbuf_addf (&buf_payload,"def_param %s=%s" , param, value); +strbuf_addf (&buf_payload,"def_param scope:%s %s=%s" , scope_name, param, + value);normal_io_write_fl (file, line, &buf_payload);strbuf_release (&buf_payload); }
trace2/tr2_tgt_perf.c +7 -2
@@ -441,12 +441,17 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct strbuf buf_payload =STRBUF_INIT ; +struct strbuf scope_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);strbuf_addf (&buf_payload,"%s:%s" , param, value); +strbuf_addf (&scope_payload,"%s:%s" ,"scope" , scope_name); -perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL ,NULL , - &buf_payload); +perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL , + scope_payload.buf, &buf_payload);strbuf_release (&buf_payload); +strbuf_release (&scope_payload); } static void fn_repo_fl(const char *file, int line,
Am I understand accurately?
Thanks.
Re: [PATCH v7 0/7] trace2: dump scope when print "interesting" config
Quoted 5 lines · tenglong.tl
Quoted 2 lines
Yeah likewise, it even applies directly on master. But I can live with it :)
Sorry, may I ask what's the recommanded way to deal this, alway use tag or master on update?
We generally submit patches on a single "topic", what a "topic" is is often fuzzy, and sometimes a topic that's mostly trying to do X will fix or change some unrelated Y "while at it".
But patch or patches at the end of a series don't depend on anything that comes before them, and could be "cherry-pick"'d directly on top of "master" that's generally a sign that you should be submitting two sets of patches, not one.
Per
https://lore.kernel.org/git/2016ef2e342c2ec6517afa8ec3e57035021fb965.1650547400.git.dyroneteng@gmail.com/
the "let's log config" is just something you happened to run into on this topic, but it might have just as well been some other command.
So I think it's better to split it up into its own topic.
Quoted 38 lines
Quoted 8 lines
One minor nit is that something like this (which needs to be fleshened up) should be fixed up into 7/7 (and maybe we want to keep the "..."?):
Documentation/technical/api-trace2.txt +0
-0Documentation/technical/api-trace2.txt The "exec_id" field is a command-unique id and is only useful if the · line 722
Quoted 8 lines
Documentation/technical/api-trace2.txt +1
-1@@ -716,7 +716,7 @@ The "exec_id" field is a command-unique id and is only useful if the ------------ { "event":"def_param", - ... + "scope": ... "param":"core.abbrev", "value":"7" }I do a little check, the json format in this scenario is like :
{
"event": "def_param",
"sid": "20220719T075535.279369Z-H1b0a19dc-P000093db", "thread": "main",
"time": "2022-07-19T07:55:35.280720Z", "file": "git.c",
"line": 461,
"scope": "global",
"param": "color.ui",
"value": "always"
}So, I think this is ok maybe:
Documentation/technical/api-trace2.txt +0
-0
Documentation/technical/api-trace2.txt The "exec_id" field is a command-unique id and is only useful if the · line 722
Quoted 6 lines
Documentation/technical/api-trace2.txt +1
-0@@ -717,6 +717,7 @@ The "exec_id" field is a command-unique id and is only useful if the { "event":"def_param", + "scope":... "param":"core.abbrev", "value":"7"
Yes, that sound good, although I'd make that "scope" line be:
"scope": "global" | "worktree" | <add more things to the list here>
Or just say:
"scope": <a string that 'git config --show-scope' would return>,
Which covers all the possibilities, without hardcoding them there.
Quoted 9 lines
}
Quoted 3 lines
And that the addition to api-trace2.txt seems to partially be something that should just link to
Documentation/config/trace2.txt, i.e. it's generally documenting an existing facility.Do you mean the modification about
trace2.configParamsand you suggest to make a link to Documentation/config/trace2.txt(actually as git-config[1]) in /Documentation/technical/api-trace2.txt?
I mean that part of what you're adding is about this new "scope" feature, but another part just seems to be explaining how the trace2.configParams works in general.
For the "works in general" let's either link to git-config(1), or if that explanation is lacking improve it & link to it.
Quoted 61 lines
Quoted 2 lines
I think it would be great in any case to have that 7/7 split into what we do now & docs for that, and then the minor addition of "scope".
Let me try to understand this, it's better to split [7/7] into two commits.
First commit is to add docs in /Documentation/technical/api-trace2.txt to let reader to find the print-config ability, by the way link the doc about the
GIT_TRACE2_CONFIG_PARAMand trace2.configparami in git-config[1].Like:
commit
2db47572d4462e3788a92fd355b97df13b9bcc39Author: Teng Long <[email protected]> Date: Tue Jul 19 17:30:35 2022 +0800api-trace2.txt: add docs to print config
It's supported to print "interesting" config value to tr2 log by setting the "
GIT_TRACE2_CONFIG_PARAMS" environment variable and the "trace2.configparam" config, let's add the related docs.Signed-off-by Teng Long <[email protected]>Documentation/technical/api-trace2.txt +30
-0@@ -1207,6 +1207,45 @@ at offset 508. This example also shows that thread names are assigned in a racy manner as each thread starts and allocates TLS storage. +Print Configs:: + + Dump "interesting" config values to trace2 log. ++ +The environment variable `GIT_TRACE2_CONFIG_PARAMS` and configuration +`trace2.configparams` can be used to output config values which you care +about(see linkgit:git-config[1). For example, assume that we want to +config different `color.ui` values in multiple scopes, such as: ++ +---------------- +$ git config --system color.ui never +$ git config --global color.ui always +$ git config --local color.ui auto +$ git config --list --show-scope | grep 'color.ui' +system color.ui=never +global color.ui=always +local color.ui=auto +---------------- ++ +Then, mark the config `color.ui` as "interesting" config with +`GIT_TRACE2_CONFIG_PARAMS`: ++ +---------------- +$ export GIT_TRACE2_PERF_BRIEF=1 +$ export GIT_TRACE2_PERF=~/log.perf +$ export GIT_TRACE2_CONFIG_PARAMS=color.ui +$ git version +... +$ cat ~/log.perf +d0 | main | version | | | | | ...
Documentation/technical/api-trace2.txt at offset 508. · line 1249
Quoted 10 lines
Documentation/technical/api-trace2.txt +9
-0+d0 | main | start | | 0.000260 | | | /opt/git/master/bin/git version +d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] +d0 | main | cmd_name | | | | | version (version) +d0 | main | def_param | | | | | color.ui:never +d0 | main | def_param | | | | | color.ui:always +d0 | main | def_param | | | | | color.ui:auto +d0 | main | exit | | 0.000470 | | | code:0 +d0 | main | atexit | | 0.000477 | | | code:0 +---------------- == Future Work
Yes, something like that, although it's a bit odd to discuss "scope" here and not have the trace show it yet, but that's fixed below.:
Quoted 136 lines
Second is actually what we do in [7/7], modify to support print scope names and update the docs which based on what we just add in the first commit
Like:
commit
615ab4864fce6b8042778aa78799ac2656785710(HEAD -> tl/bitmap-append-trace2-outputs)Author: Teng Long <[email protected]> Date: Thu Jul 7 21:48:32 2022 +0800tr2: dump names if config exist in multiple scopes
When we specify
GIT_TRACE2_CONFIG_PARAMSor trace2.configparams, trace2 will prints "interesting" config values to log. Sometimes, when a config set in multiple scope files, the following output looks like (the irrelevant fields are omitted here as "..."):...|
def_param| ... | core.multipackindex:false
...|def_param| ... | core.multipackindex:false
...|def_param| ... | core.multipackindex:falseAs the log shows, even each config in different scope is dumped, but we don't know which scope it comes from. Therefore, it's better to add the scope names as well to make them be more recognizable. For example, when execute:
$ GIT_TRACE2_PERF=1 \
> GIT_TRACE2_CONFIG_PARAMS=core.multipackIndex \
> git rev-list --test-bitmap HEAD"The following is the ouput (the irrelevant fields are omitted here as "..."):
Format normal:
... git.c:461 ...def_paramscope:system core.multipackindex=false ... git.c:461 ...def_paramscope:global core.multipackindex=false ... git.c:461 ...def_paramscope:local core.multipackindex=falseFormat perf:
... |
def_param| ... | scope:system | core.multipackindex:false
... |def_param| ... | scope:global | core.multipackindex:false
... |def_param| ... | scope:local | core.multipackindex:falseFormat event:
{"event":"def_param", ... ,"scope":"system","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"global","param":"core.multipackindex","value":"false"} {"event":"def_param", ... ,"scope":"local","param":"core.multipackindex","value":"false"}
Signed-off-by Teng Long <[email protected]>Documentation/technical/api-trace2.txt +4
-3@@ -717,6 +717,7 @@ The "exec_id" field is a command-unique id and is only useful if the { "event":"def_param", + "scope":... "param":"core.abbrev", "value":"7" }@@ -1240,9 +1241,9 @@ d0 | main | version | | | | d0 | main | start | | 0.000260 | | | /opt/git/master/bin/git version d0 | main | cmd_ancestry | | | | | ancestry:[bash sshd sshd sshd systemd] d0 | main | cmd_name | | | | | version (version) -d0 | main | def_param | | | | | color.ui:never -d0 | main | def_param | | | | | color.ui:always -d0 | main | def_param | | | | | color.ui:auto +d0 | main | def_param | | | | scope:system | color.ui:never +d0 | main | def_param | | | | scope:global | color.ui:always +d0 | main | def_param | | | | scope:local | color.ui:auto d0 | main | exit | | 0.000470 | | | code:0 d0 | main | atexit | | 0.000477 | | | code:0 ----------------trace2/tr2_tgt_event.c +3
-0@@ -479,9 +479,12 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct json_writer jw =JSON_WRITER_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);jw_object_begin (&jw,0 );event_fmt_prepare (event_name, file, line,NULL , &jw); +jw_object_string (&jw,"scope" , scope_name);jw_object_string (&jw,"param" , param);jw_object_string (&jw,"value" , value);jw_end (&jw);trace2/tr2_tgt_normal.c +4
-1@@ -298,8 +298,11 @@ static void fn_param_fl(const char *file, int line, const char *param, const char *value) {struct strbuf buf_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope); -strbuf_addf (&buf_payload,"def_param %s=%s" , param, value); +strbuf_addf (&buf_payload,"def_param scope:%s %s=%s" , scope_name, param, + value);normal_io_write_fl (file, line, &buf_payload);strbuf_release (&buf_payload); }trace2/tr2_tgt_perf.c +7
-2@@ -441,12 +441,17 @@ static void fn_param_fl(const char *file, int line, const char *param, {const char *event_name ="def_param" ;struct strbuf buf_payload =STRBUF_INIT ; +struct strbuf scope_payload =STRBUF_INIT ; +enum config_scope scope =current_config_scope (); +const char *scope_name =config_scope_name (scope);strbuf_addf (&buf_payload,"%s:%s" , param, value); +strbuf_addf (&scope_payload,"%s:%s" ,"scope" , scope_name); -perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL ,NULL , - &buf_payload); +perf_io_write_fl (file, line, event_name,NULL ,NULL ,NULL , + scope_payload.buf, &buf_payload);strbuf_release (&buf_payload); +strbuf_release (&scope_payload); } static void fn_repo_fl(const char *file, int line,
Quoted 2 lines
Am I understand accurately?
Thanks.
Yes, exactly! That makes it much clearer what the functional change was about, i.e. we can see what parts of the trace are now different (the scope is added to the trace).
Re: [PATCH v7 0/7] trace2: dump scope when print "interesting" config
Quoted 9 lines · Ævar Arnfjörð Bjarmason
We generally submit patches on a single "topic", what a "topic" is is often fuzzy, and sometimes a topic that's mostly trying to do X will fix or change some unrelated Y "while at it".
But patch or patches at the end of a series don't depend on anything that comes before them, and could be "cherry-pick"'d directly on top of "master" that's generally a sign that you should be submitting two sets of patches, not one.
Quoted 6 lines
Per
https://lore.kernel.org/git/2016ef2e342c2ec6517afa8ec3e57035021fb965.1650547400.git.dyroneteng@gmail.com/
the "let's log config" is just something you happened to run into on this topic, but it might have just as well been some other command.So I think it's better to split it up into its own topic.
Make sense.
Quoted 9 lines
Yes, that sound good, although I'd make that "scope" line be:
"scope": "global" | "worktree" | <add more things to the list here>
Or just say:
"scope": <a string that 'git config --show-scope' would return>,
Which covers all the possibilities, without hardcoding them there.
I think I'd like to prefer the second way, thanks for the input.
Quoted 6 lines
I mean that part of what you're adding is about this new "scope" feature, but another part just seems to be explaining how the trace2.configParams works in general.
For the "works in general" let's either link to git-config(1), or if that explanation is lacking improve it & link to it.
Yes, I think the current explanation is OK for me in git-config[1], so
I add the link as in previous reply (Search: 2db47572d4462e3788a92fd355b97df13b9bcc39) :
+trace2.configparams can be used to output config values which you care +about(see linkgit:git-config[1).
Quoted 2 lines
Yes, something like that, although it's a bit odd to discuss "scope" here and not have the trace show it yet, but that's fixed below.:
Yes, because I want the result to be more obvious, if a config only in single scope maybe it's a litter harder to remember "Wo, what looks like if config is in multiple scopes?(although it's intentional)"
Quoted 3 lines
Yes, exactly! That makes it much clearer what the functional change was about, i.e. we can see what parts of the trace are now different (the scope is added to the trace).
Yes.
Thanks.
v8 cover pack-bitmap.c: optimize error messages
In reply to Teng LongFrom: Teng Long <[email protected]>
Changes since v7:
Remove [7/7] and will move to another separate topic.
Thanks.
Teng Long (6):
pack-bitmap.c: fix formatting of error messages pack-bitmap.c: mark more strings for translations pack-bitmap.c: rename "idx_name" to "bitmap_name" pack-bitmap.c: do not ignore error when opening a bitmap file pack-bitmap.c: using error() instead of silently returning -1 pack-bitmap.c: continue looping when first MIDX bitmap is found
pack-bitmap.c | 103 ++++++++++++++++++++++++++++---------------------- 1 file changed, 58 insertions(+), 45 deletions(-)
Range-diff against v7:
1: a01ae8478d < -: ---------- tr2: dump names if config exist in multiple scopes -: ---------- > 1: 7426b277ba pack-bitmap.c: fix formatting of error messages -: ---------- > 2: 4e493426a0 pack-bitmap.c: mark more strings for translations -: ---------- > 3: 263f45ba96 pack-bitmap.c: rename "idx_name" to "bitmap_name" -: ---------- > 4: d11ea092d5 pack-bitmap.c: do not ignore error when opening a bitmap file -: ---------- > 5: f60efe78d6 pack-bitmap.c: using error() instead of silently returning -1 -: ---------- > 6: 83090308ad pack-bitmap.c: continue looping when first MIDX bitmap is found
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v8 1/6 pack-bitmap.c: fix formatting of error messages
From: Teng Long <[email protected]>
There are some text output issues in 'pack-bitmap.c', they exist in die(), error() etc. This includes issues with capitalization the first letter, newlines, error() instead of BUG(), and substitution that don't have quotes around them.
--- pack-bitmap.c | 47 ++++++++++++++++++++++++----------------------- 1 file changed, 24 insertions(+), 23 deletions(-)
pack-bitmap.c +24 -23
@@ -1,5 +1,6 @@#include "cache.h" #include "commit.h" +#include "strbuf.h" #include "tag.h" #include "diff.h" #include "revision.h"
@@ -138,7 +139,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("Failed to load bitmap index (corrupted?)" ); +error ("failed to load bitmap index (corrupted?)" );ewah_pool_free (b);return NULL ; }
@@ -160,14 +161,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("Corrupted bitmap index (too small)" ); +return error ("corrupted bitmap index (too small)" );if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("Corrupted bitmap index file (wrong header)" ); +return error ("corrupted bitmap index file (wrong header)" ); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("Unsupported version for bitmap index file (%d)" , index->version); +return error ("unsupported version '%d' for bitmap index file" , index->version);/* Parse known bitmap format options */ {
@@ -176,7 +177,7 @@ static int load_bitmap_header(struct bitmap_index *index)unsigned char *index_end = index->map + index->map_size - the_hash_algo->rawsz;if ((flags &BITMAP_OPT_FULL_DAG ) ==0 ) -return error ("Unsupported options for bitmap index file " +BUG ("unsupported options for bitmap index file " "(Git requires BITMAP_OPT_FULL_DAG)" );if (flags &BITMAP_OPT_HASH_CACHE ) {
@@ -215,7 +216,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("Duplicate entry in bitmap index: %s" ,oid_to_hex (oid)); +error ("duplicate entry in bitmap index: '%s'" ,oid_to_hex (oid));return NULL ; }
@@ -274,13 +275,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("Corrupted bitmap pack index" ); +return error ("corrupted bitmap pack index" );if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("Invalid XOR offset in bitmap pack index" ); +return error ("invalid XOR offset in bitmap pack index" ); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -330,7 +331,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , buf.buf); +warning ("ignoring extra bitmap file: '%s'" , buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -387,7 +388,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: %s" , packfile->pack_name); +warning ("ignoring extra bitmap file: '%s'" , packfile->pack_name);close (fd);return -1 ; }
@@ -1628,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object %s not found in type bitmaps" , +die ("object '%s' not found in type bitmaps" ,oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object %s does not have a unique type" , +die ("object '%s' does not have a unique type" ,oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object %s: real type %s, expected: %s" , +die ("object '%s': real type '%s', expected: '%s'" ,oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1650,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&object->oid)); +die ("object not in bitmap: '%s'" ,oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1665,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("Object not in bitmap: %s\n" ,oid_to_hex (&commit->object.oid)); +die ("object not in bitmap: '%s'" ,oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1687,21 +1688,21 @@ void test_bitmap_walk(struct rev_info *revs)if (revs->pending.nr !=1 )die ("you must specify exactly one commit to test" ); -fprintf (stderr,"Bitmap v%d test (%d entries loaded)\n" , +fprintf_ln (stderr,"Bitmap v%d test (%d entries loaded)" , bitmap_git->version, bitmap_git->entry_count); root = revs->pending.objects[0 ].item; bm =bitmap_for_commit (bitmap_git, (struct commit *)root);if (bm) { -fprintf (stderr,"Found bitmap for %s. %d bits / %08x checksum\n" , +fprintf_ln (stderr,"Found bitmap for '%s'. %d bits / %08x checksum" ,oid_to_hex (&root->oid), (int )bm->bit_size,ewah_checksum (bm)); result =ewah_to_bitmap (bm); }if (result ==NULL ) -die ("Commit %s doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die ("commit '%s' doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1726,7 +1727,7 @@ void test_bitmap_walk(struct rev_info *revs)stop_progress (&tdata.prg);if (bitmap_equals (result, tdata.base)) -fprintf (stderr,"OK!\n" ); +fprintf_ln (stderr,"OK!" );else die ("mismatch in bitmap results" );
@@ -1749,7 +1750,7 @@ int test_bitmap_commits(struct repository *r)die ("failed to load bitmap indexes" );kh_foreach (bitmap_git->bitmaps, oid, value, { -printf ("%s\n" ,oid_to_hex (&oid)); +printf_ln ("%s" ,oid_to_hex (&oid)); });free_bitmap_index (bitmap_git);
@@ -1774,7 +1775,7 @@ int test_bitmap_hashes(struct repository *r)nth_bitmap_object_oid (bitmap_git, &oid, index_pos); -printf ("%s %" PRIu32"\n" , +printf_ln ("%s %" PRIu32"" ,oid_to_hex (&oid),get_be32 (bitmap_git->hashes + index_pos)); }
@@ -1936,7 +1937,7 @@ static off_t get_disk_usage_for_type(struct bitmap_index *bitmap_git,struct object_id oid;nth_midxed_object_oid (&oid, bitmap_git->midx, midx_pos); -die (_ ("could not find %s in pack %s at offset %" PRIuMAX), +die (_ ("could not find '%s' in pack '%s' at offset %" PRIuMAX),oid_to_hex (&oid), pack->pack_name, (uintmax_t )offset);
@@ -1972,7 +1973,7 @@ static off_t get_disk_usage_for_extended(struct bitmap_index *bitmap_git)continue ;if (oid_object_info_extended (the_repository, &obj->oid, &oi,0 ) <0 ) -die (_ ("unable to get disk usage of %s" ), +die (_ ("unable to get disk usage of '%s'" ),oid_to_hex (&obj->oid)); total += object_size;
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v8 2/6 pack-bitmap.c: mark more strings for translations
In reply to tenglong.tlFrom: Teng Long <[email protected]>
In pack-bitmap.c, some printed texts are translated, some are not. Let's support the translations of the bitmap related output.
--- pack-bitmap.c | 48 ++++++++++++++++++++++++------------------------ 1 file changed, 24 insertions(+), 24 deletions(-)
pack-bitmap.c +24 -24
@@ -139,7 +139,7 @@ static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index) index->map_size - index->map_pos);if (bitmap_size <0 ) { -error ("failed to load bitmap index (corrupted?)" ); +error (_ ("failed to load bitmap index (corrupted?)" ));ewah_pool_free (b);return NULL ; }
@@ -161,14 +161,14 @@ static int load_bitmap_header(struct bitmap_index *index)size_t header_size =sizeof (*header) -GIT_MAX_RAWSZ + the_hash_algo->rawsz;if (index->map_size < header_size + the_hash_algo->rawsz) -return error ("corrupted bitmap index (too small)" ); +return error (_ ("corrupted bitmap index (too small)" ));if (memcmp (header->magic,BITMAP_IDX_SIGNATURE ,sizeof (BITMAP_IDX_SIGNATURE )) !=0 ) -return error ("corrupted bitmap index file (wrong header)" ); +return error (_ ("corrupted bitmap index file (wrong header)" )); index->version =ntohs (header->version);if (index->version !=1 ) -return error ("unsupported version '%d' for bitmap index file" , index->version); +return error (_ ("unsupported version '%d' for bitmap index file" ), index->version);/* Parse known bitmap format options */ {
@@ -182,7 +182,7 @@ static int load_bitmap_header(struct bitmap_index *index)if (flags &BITMAP_OPT_HASH_CACHE ) {if (cache_size > index_end - index->map - header_size) -return error ("corrupted bitmap index file (too short to fit hash cache)" ); +return error (_ ("corrupted bitmap index file (too short to fit hash cache)" )); index->hashes = (void *)(index_end - cache_size); index_end -= cache_size; }
@@ -216,7 +216,7 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,* because the SHA1 already existed on the map. this is bad, there * shouldn't be duplicated commits in the index */ if (ret ==0 ) { -error ("duplicate entry in bitmap index: '%s'" ,oid_to_hex (oid)); +error (_ ("duplicate entry in bitmap index: '%s'" ),oid_to_hex (oid));return NULL ; }
@@ -260,14 +260,14 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)struct object_id oid;if (index->map_size - index->map_pos <6 ) -return error ("corrupt ewah bitmap: truncated header for entry %d" , i); +return error (_ ("corrupt ewah bitmap: truncated header for entry %d" ), i); commit_idx_pos =read_be32 (index->map, &index->map_pos); xor_offset =read_u8 (index->map, &index->map_pos); flags =read_u8 (index->map, &index->map_pos);if (nth_bitmap_object_oid (index, &oid, commit_idx_pos) <0 ) -return error ("corrupt ewah bitmap: commit index %u out of range" , +return error (_ ("corrupt ewah bitmap: commit index %u out of range" ), (unsigned )commit_idx_pos); bitmap =read_bitmap_1 (index);
@@ -275,13 +275,13 @@ static int load_bitmap_entries_v1(struct bitmap_index *index)return -1 ;if (xor_offset >MAX_XOR_OFFSET || xor_offset > i) -return error ("corrupted bitmap pack index" ); +return error (_ ("corrupted bitmap pack index" ));if (xor_offset >0 ) { xor_bitmap = recent_bitmaps[(i - xor_offset) %MAX_XOR_OFFSET ];if (xor_bitmap ==NULL ) -return error ("invalid XOR offset in bitmap pack index" ); +return error (_ ("invalid XOR offset in bitmap pack index" )); } recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
@@ -331,7 +331,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,struct strbuf buf =STRBUF_INIT ;get_midx_filename (&buf, midx->object_dir);/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: '%s'" , buf.buf); +warning (_ ("ignoring extra bitmap file: '%s'" ), buf.buf);close (fd);strbuf_release (&buf);return -1 ;
@@ -388,7 +388,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitif (bitmap_git->pack || bitmap_git->midx) {/* ignore extra bitmap file; we can only handle one */ -warning ("ignoring extra bitmap file: '%s'" , packfile->pack_name); +warning (_ ("ignoring extra bitmap file: '%s'" ), packfile->pack_name);close (fd);return -1 ; }
@@ -819,7 +819,7 @@ static struct bitmap *find_objects(struct bitmap_index *bitmap_git, revs->include_check_data = &incdata;if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); show_data.bitmap_git = bitmap_git; show_data.base = base;
@@ -1629,15 +1629,15 @@ static void test_bitmap_type(struct bitmap_test_data *tdata, }if (bitmap_type ==OBJ_NONE ) -die ("object '%s' not found in type bitmaps" , +die (_ ("object '%s' not found in type bitmaps" ),oid_to_hex (&obj->oid));if (bitmaps_nr >1 ) -die ("object '%s' does not have a unique type" , +die (_ ("object '%s' does not have a unique type" ),oid_to_hex (&obj->oid));if (bitmap_type != obj->type) -die ("object '%s': real type '%s', expected: '%s'" , +die (_ ("object '%s': real type '%s', expected: '%s'" ),oid_to_hex (&obj->oid),type_name (obj->type),type_name (bitmap_type));
@@ -1651,7 +1651,7 @@ static void test_show_object(struct object *object, const char *name, bitmap_pos =bitmap_position (tdata->bitmap_git, &object->oid);if (bitmap_pos <0 ) -die ("object not in bitmap: '%s'" ,oid_to_hex (&object->oid)); +die (_ ("object not in bitmap: '%s'" ),oid_to_hex (&object->oid));test_bitmap_type (tdata, object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1666,7 +1666,7 @@ static void test_show_commit(struct commit *commit, void *data) bitmap_pos =bitmap_position (tdata->bitmap_git, &commit->object.oid);if (bitmap_pos <0 ) -die ("object not in bitmap: '%s'" ,oid_to_hex (&commit->object.oid)); +die (_ ("object not in bitmap: '%s'" ),oid_to_hex (&commit->object.oid));test_bitmap_type (tdata, &commit->object, bitmap_pos);bitmap_set (tdata->base, bitmap_pos);
@@ -1683,10 +1683,10 @@ void test_bitmap_walk(struct rev_info *revs)struct ewah_bitmap *bm;if (!(bitmap_git =prepare_bitmap_git (revs->repo))) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" ));if (revs->pending.nr !=1 ) -die ("you must specify exactly one commit to test" ); +die (_ ("you must specify exactly one commit to test" ));fprintf_ln (stderr,"Bitmap v%d test (%d entries loaded)" , bitmap_git->version, bitmap_git->entry_count);
@@ -1702,7 +1702,7 @@ void test_bitmap_walk(struct rev_info *revs) }if (result ==NULL ) -die ("commit '%s' doesn't have an indexed bitmap" ,oid_to_hex (&root->oid)); +die (_ ("commit '%s' doesn't have an indexed bitmap" ),oid_to_hex (&root->oid)); revs->tag_objects =1 ; revs->tree_objects =1 ;
@@ -1711,7 +1711,7 @@ void test_bitmap_walk(struct rev_info *revs) result_popcnt =bitmap_popcount (result);if (prepare_revision_walk (revs)) -die ("revision walk setup failed" ); +die (_ ("revision walk setup failed" )); tdata.bitmap_git = bitmap_git; tdata.base =bitmap_new ();
@@ -1729,7 +1729,7 @@ void test_bitmap_walk(struct rev_info *revs)if (bitmap_equals (result, tdata.base))fprintf_ln (stderr,"OK!" );else -die ("mismatch in bitmap results" ); +die (_ ("mismatch in bitmap results" ));bitmap_free (result);bitmap_free (tdata.base);
@@ -1747,7 +1747,7 @@ int test_bitmap_commits(struct repository *r) MAYBE_UNUSED void *value;if (!bitmap_git) -die ("failed to load bitmap indexes" ); +die (_ ("failed to load bitmap indexes" )); kh_foreach(bitmap_git->bitmaps, oid, value, { printf_ln("%s", oid_to_hex(&oid));
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v8 3/6 pack-bitmap.c: rename "idx_name" to "bitmap_name"
In reply to tenglong.tlFrom: Teng Long <[email protected]>
In "open_pack_bitmap_1()" and "open_midx_bitmap_1()" we use a var named "idx_name" to represent the bitmap filename which is computed by "midx_bitmap_filename()" or "pack_bitmap_filename()" before we open it.
There may bring some confusion in this "idx_name" naming, which might lead us to think of ".idx "or" multi-pack-index" files, although bitmap is essentially can be understood as a kind of index, let's define this name a little more accurate here.
--- pack-bitmap.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-)
pack-bitmap.c +7 -7
@@ -314,10 +314,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git, struct multi_pack_index *midx) {struct stat st; -char *idx_name =midx_bitmap_filename (midx); -int fd =git_open (idx_name); +char *bitmap_name =midx_bitmap_filename (midx); +int fd =git_open (bitmap_name); -free (idx_name); +free (bitmap_name);if (fd <0 )return -1 ;
@@ -369,14 +369,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git {int fd;struct stat st; -char *idx_name; +char *bitmap_name;if (open_pack_index (packfile))return -1 ; - idx_name =pack_bitmap_filename (packfile); - fd =git_open (idx_name); -free (idx_name); + bitmap_name =pack_bitmap_filename (packfile); + fd =git_open (bitmap_name); +free (bitmap_name);if (fd <0 )return -1 ;
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v8 4/6 pack-bitmap.c: do not ignore error when opening a bitmap file
In reply to tenglong.tlFrom: Teng Long <[email protected]>
Calls to git_open() to open the pack bitmap file and multi-pack bitmap file do not report any error when they fail. These files are optional and it is not an error if open failed due to ENOENT, but we shouldn't be ignoring other kinds of errors.
--- pack-bitmap.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-)
pack-bitmap.c +12 -5
@@ -317,10 +317,13 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,char *bitmap_name =midx_bitmap_filename (midx);int fd =git_open (bitmap_name); -free (bitmap_name); - -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning_errno ("cannot open '%s'" , bitmap_name); +free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
@@ -376,10 +379,14 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_git bitmap_name =pack_bitmap_filename (packfile); fd =git_open (bitmap_name); -free (bitmap_name); -if (fd <0 ) +if (fd <0 ) { +if (errno !=ENOENT ) +warning_errno ("cannot open '%s'" , bitmap_name); +free (bitmap_name);return -1 ; + } +free (bitmap_name);if (fstat (fd, &st)) {close (fd);
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v8 5/6 pack-bitmap.c: using error() instead of silently returning -1
In reply to tenglong.tlFrom: Teng Long <[email protected]>
In "open_pack_bitmap_1()" and "open_midx_bitmap_1()", it's better to return error() instead of "-1" when some unexpected error occurs like "stat bitmap file failed", "bitmap header is invalid" or "checksum mismatch", etc.
There are places where we do not replace, such as when the bitmap does not exist (no bitmap in repository is allowed) or when another bitmap has already been opened (in which case it should be a warning rather than an error).
--- pack-bitmap.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
pack-bitmap.c +5 -1
@@ -326,6 +326,7 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,free (bitmap_name);if (fstat (fd, &st)) { +error_errno (_ ("cannot fstat bitmap file" ));close (fd);return -1 ; }
@@ -350,8 +351,10 @@ static int open_midx_bitmap_1(struct bitmap_index *bitmap_git,if (load_bitmap_header (bitmap_git) <0 ) goto cleanup; -if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) +if (!hasheq (get_midx_checksum (bitmap_git->midx), bitmap_git->checksum)) { +error (_ ("checksum doesn't match in MIDX and bitmap" )); goto cleanup; + }if (load_midx_revindex (bitmap_git->midx) <0 ) {warning (_ ("multi-pack bitmap is missing required reverse index" ));
@@ -389,6 +392,7 @@ static int open_pack_bitmap_1(struct bitmap_index *bitmap_git, struct packed_gitfree (bitmap_name);if (fstat (fd, &st)) { +error_errno (_ ("cannot fstat bitmap file" ));close (fd);return -1 ; }
Signature · 1 line
2.35.0.rc0.679.gc613175da2
PATCH v8 6/6 pack-bitmap.c: continue looping when first MIDX bitmap is found
In reply to tenglong.tlFrom: Teng Long <[email protected]>
In "open_midx_bitmap()", we do a loop with the MIDX(es) in repo, when the first one has been found, then will break out by a "return" directly.
But actually, it's better to continue the loop until we have visited both the MIDX in our repository, as well as any alternates (along with _their_ alternates, recursively).
The reason for this is, there may exist more than one MIDX file in a repo. The "multi_pack_index" struct is actually designed as a singly linked list, and if a MIDX file has been already opened successfully, then the other MIDX files will be skipped and left with a warning "ignoring extra bitmap file." to the output.
The discussion link of community:
https://public-inbox.org/git/[email protected]/
--- pack-bitmap.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)
pack-bitmap.c +3 -2
@@ -506,15 +506,16 @@ static int open_pack_bitmap(struct repository *r,static int open_midx_bitmap (struct repository *r,struct bitmap_index *bitmap_git) { +int ret =-1 ;struct multi_pack_index *midx;assert (!bitmap_git->map);for (midx =get_multi_pack_index (r); midx; midx = midx->next) {if (!open_midx_bitmap_1 (bitmap_git, midx)) -return 0 ; + ret =0 ; } -return -1 ; +return ret; } static int open_bitmap(struct repository *r,
Signature · 1 line
2.35.0.rc0.679.gc613175da2
Re: [PATCH v8 0/6] pack-bitmap.c: optimize error messages
In reply to tenglong.tlQuoted 5 lines · tenglong.tl
From: Teng Long <[email protected]>
Changes since v7:
Remove [7/7] and will move to another separate topic.
If that is the only change, then I am relieved, as I can safely ignore this round ;-)
The previous round (without the last step) has been in 'next' since Tuesday and cannot be replaced.
It seems that both of the topics are in good shape. Let's plan to merge the other one about showing the scope of the configuration variable in trace2 output down to 'next' soonish.
Thanks.
Re: [PATCH v8 0/6] pack-bitmap.c: optimize error messages
Quoted 2 lines · Junio C Hamano
If that is the only change, then I am relieved, as I can safely ignore this round ;-)
Yes, only removed 7/7, the others do not change.
Quoted 6 lines
The previous round (without the last step) has been in 'next' since Tuesday and cannot be replaced.
It seems that both of the topics are in good shape. Let's plan to merge the other one about showing the scope of the configuration variable in trace2 output down to 'next' soonish.
Oh, I see.
Thanks.