Teach git-fetch to exploit server side automatic tag following
PATCH v2 3/3 Teach git-fetch to exploit server side automatic tag following
If the remote peer upload-pack process supports the auto-include-tag protocol extension then we can avoid running a second fetch cycle on the client side by letting the server send us the annotated tags along with the objects it is packing for us. In the following graph we can now fetch both "tag1" and "tag2" on the same connection that we fetched "master" from the remote when we only have L available on the local side:
T - tag1 S - tag2
/ /
L - o ------ o ------ B
\ \
\ \
origin/master masterThe objects for "tag1" are implicitly downloaded without our direct knowledge. The existing "quickfetch" optimization within git-fetch discovers that tag1 is complete after the first connection and does not open a second connection.
--- builtin-fetch.c | 3 +++ t/t5503-tagfollow.sh | 26 ++++++++++++++++++++++++++ transport.c | 5 +++++ transport.h | 3 +++ 4 files changed, 37 insertions(+), 0 deletions(-)
builtin-fetch.c +3 -0
@@ -555,6 +555,8 @@ static int do_fetch(struct transport *transport,read_ref (rm->peer_ref->name, rm->peer_ref->old_sha1); } +if (tags ==TAGS_DEFAULT && autotags) +transport_set_option (transport,TRANS_OPT_FOLLOWTAGS ,"1" );if (fetch_refs (transport, ref_map)) {free_refs (ref_map);return 1 ;
@@ -568,6 +570,7 @@ static int do_fetch(struct transport *transport, ref_map =NULL ;find_non_local_tags (transport, &ref_map, &tail);if (ref_map) { +transport_set_option (transport,TRANS_OPT_FOLLOWTAGS ,NULL );transport_set_option (transport,TRANS_OPT_DEPTH ,"0" );fetch_refs (transport, ref_map); }
t/t5503-tagfollow.sh +26 -0
@@ -121,4 +121,30 @@ test_expect_success 'fetch B, S (commit and tag : 1 connection)' ' git diff expect actual ' +cat - <<EOF >expect +#S +want $B +want $S +#E +EOF +test_expect_success 'new clone fetch master and tags' ' + git branch -D cat + rm -f $U + ( + mkdir clone2 && + cd clone2 && + git init && + git remote add origin .. && + GIT_DEBUG_SEND_PACK=3 git fetch 3>../$U && + test $B = $(git rev-parse --verify origin/master) && + test $S = $(git rev-parse --verify tag2) && + test $B = $(git rev-parse --verify tag2^0) && + test $T = $(git rev-parse --verify tag1) && + test $A = $(git rev-parse --verify tag1^0) + ) && + test -s $U && + cut -d" " -f1,2 $U >actual && + git diff expect actual +' + test_done
transport.c +5 -0
@@ -560,6 +560,7 @@ static int close_bundle(struct transport *transport)struct git_transport_data {unsigned thin :1 ;unsigned keep :1 ; +unsigned followtags :1 ;int depth;struct child_process *conn;int fd[2 ];
@@ -580,6 +581,9 @@ static int set_git_option(struct transport *connection, }else if (!strcmp (name,TRANS_OPT_THIN )) { data->thin = !!value;return 0 ; + }else if (!strcmp (name,TRANS_OPT_FOLLOWTAGS )) { + data->followtags = !!value; +return 0 ; }else if (!strcmp (name,TRANS_OPT_KEEP )) { data->keep = !!value;return 0 ;
@@ -628,6 +632,7 @@ static int fetch_refs_via_pack(struct transport *transport, args.keep_pack = data->keep; args.lock_pack =1 ; args.use_thin_pack = data->thin; + args.auto_include_tag = data->followtags; args.verbose = transport->verbose >0 ; args.depth = data->depth;
transport.h +3 -0
@@ -53,6 +53,9 @@ struct transport *transport_get(struct remote *, const char *);/* Limit the depth of the fetch if not null */ #define TRANS_OPT_DEPTH "depth" +/* Aggressively fetch annotated tags if possible */ +#define TRANS_OPT_FOLLOWTAGS "followtags" +/** * Returns 0 if the option was used, non-zero otherwise. Prints a * message to stderr if the option is not used.
Signature · 1 line
1.5.4.3.529.gb25fb
Hi,
just two comments from me for this patch series:
- it's a beaut, and
- should followtags not be the default?
Ciao,
Dscho
Quoted 3 lines · Johannes Schindelin
just two comments from me for this patch series:
- it's a beaut, and
Thanks.
Quoted 1 line
- should followtags not be the default?
No. Absolutely not.
The client may not want tag objects from this remote. It might not want those tags for all sorts of reasons. Maybe they are doing a one-shot pull. Maybe they don't trust this remote. Maybe they trust this remote but this remote is famous for 192M tags containing PDFs of photo images of build logs printed out on paper, then photographed on a wood table and finally scanned in at 600 dpi.
So it should only be enabled *if* we were willing to open a second connection to this remote to followtags from it. And that's not a change from prior behavior either, its just faster.
Signature · 1 line
Shawn.
Hi,
Quoted 4 lines · Shawn O. Pearce
Quoted 1 line · Johannes Schindelin
- should followtags not be the default?
Quoted 8 lines
No. Absolutely not.
The client may not want tag objects from this remote. It might not want those tags for all sorts of reasons. Maybe they are doing a one-shot pull. Maybe they don't trust this remote. Maybe they trust this remote but this remote is famous for 192M tags containing PDFs of photo images of build logs printed out on paper, then photographed on a wood table and finally scanned in at 600 dpi.
Sorry, I meant to say: "should the followtags feature not be on by default in the circumstances where we would follow tags anyway"...
Maybe you do that, but I did not see it.
Quoted 3 lines
So it should only be enabled *if* we were willing to open a second connection to this remote to followtags from it. And that's not a change from prior behavior either, its just faster.
Thanks,
Dscho
Quoted 14 lines · Johannes Schindelin
Quoted 4 lines · Shawn O. Pearce
Quoted 1 line · Johannes Schindelin
- should followtags not be the default?
Quoted 8 lines
No. Absolutely not.
The client may not want tag objects from this remote. It might not want those tags for all sorts of reasons. Maybe they are doing a one-shot pull. Maybe they don't trust this remote. Maybe they trust this remote but this remote is famous for 192M tags containing PDFs of photo images of build logs printed out on paper, then photographed on a wood table and finally scanned in at 600 dpi.
Quoted 4 lines
Sorry, I meant to say: "should the followtags feature not be on by default in the circumstances where we would follow tags anyway"...
Maybe you do that, but I did not see it.
Nope. followtags should be on if the client wants tags, and off if the client does not want tags. Its that simple. There's no reason for the client to disable followtags if the server will actually support it, as disabling it when you do have tags to follow will only cost you extra net latency.
Signature · 1 line
Shawn.