feat: checkpatch: prohibit Buglink: and warn about missing Link:
PATCH 2/2 feat: checkpatch: Warn about Reported-by: not being followed by a Link:
In reply to Kai Wasserbäch--- scripts/checkpatch.pl | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
scripts/checkpatch.pl +14 -0
@@ -3144,6 +3144,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); } } + +# check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: must be immediately followed by Link:\n" . "$here\n" . $rawline); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: must be immediately followed by Link:\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } + } } # Check Fixes: styles is correct
Signature · 1 line
2.35.1
PATCH 1/2 feat: checkpatch: error on usage of a Buglink tag in the commit log
In reply to Kai Wasserbäch--- scripts/checkpatch.pl | 6 ++++++ 1 file changed, 6 insertions(+)
scripts/checkpatch.pl +6 -0
@@ -3248,6 +3248,12 @@ sub process { } } +# check if Bug[lL]ink: or Bugzilla: is used and error + if ($in_commit_log && $line =~ /^\s*(buglink|bugzilla):\s*(.*)/i) { + ERROR("COMMIT_LOG_BAD_BUGLINK", + "Non-standard tag '" . $1 . "' - use 'Link:' instead!\n" . $herecurr); + } + # Check for git id commit length and improperly formed commit descriptions # A correctly formed commit description is: # commit <SHA-1 hash length 12+ chars> ("Complete commit subject")
Signature · 1 line
2.35.1
cover feat: checkpatch: prohibit Buglink: and warn about missing Link:
Hey,
Thorsten Leemhuis suggested the following two changes to checkpatch, which I hereby humbly submit for review. Please let me know if any changes would be required for acceptance.
NOTES:
- checkpatch is rather long, so I might have chosen the wrong section to add these two checks. Any better placement suggestion is welcome.
- checkpatch implements custum versions of Perl core modules, that might be an future area for clean-ups? Eg. right off the bat there is a uniq implementation. List::Util (core module since 5.8.0 (5.7.3 to be pedantic)) has a far better performing version in XS.
Cheers,
Kai
Kai Wasserbäch (2):
feat: checkpatch: error on usage of a Buglink tag in the commit log
feat: checkpatch: Warn about Reported-by: not being followed by a
Link:
scripts/checkpatch.pl | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+)
Signature · 1 line
2.35.1
PATCH 1/2 checkpatch: warn when unknown tags are used for links
Issue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --format=email -P --grep='^\w+:[ \t]*http' | \
grep -Poh '^\w+:[ \t]*http' | \
sort | uniq -c | sort -rn | head -n 20
103889 Link: http
415 BugLink: http
372 Patchwork: http
270 Closes: http
221 Bug: http
121 References: http
101 v1: http
77 Bugzilla: http
60 URL: http
59 v2: http
37 Datasheet: http
35 v3: http
19 v4: http
12 v5: http
9 Ref: http
9 Fixes: http
9 Buglink: http
8 v6: http
8 Reference: http
7 V1: httpSome of these non-standard tags make it harder for external tools that rely on use of proper tags. One of those tools is the regression tracking bot 'regzbot', which looks out for "Link:" tags pointing to reports of tracked regressions.
The initial idea was to use a disallow list to raise an error when encountering known unwanted tags like BugLink:; during review it was requested to use a list of allowed tags instead[2].
--- scripts/checkpatch.pl | 7 +++++++ 1 file changed, 7 insertions(+)
scripts/checkpatch.pl +7 -0
@@ -3239,6 +3239,13 @@ sub process { $commit_log_possible_stack_dump = 0; } +# Check for odd tags before a URI/URL + if ($in_commit_log && + $line =~ /^\s*(\w+):\s*http/ && $1 !~ /^Link/) { + WARN("COMMIT_LOG_USE_LINK", + "Unknown link reference '$1:', use 'Link:' instead.\n" . $herecurr); + } + # Check for lines starting with a # if ($in_commit_log && $line =~ /^#/) { if (WARN("COMMIT_COMMENT_SYMBOL",
Signature · 1 line
2.35.1
PATCH 2/2 checkpatch: warn when Reported-by: is not followed by Link:
In reply to Kai WasserbächEncourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 10 lines
Again, the commit has a link to the patch *submission*, which is almost entirely useless. There's no link to the actual problem the patch fixes.
[...]
Put another way: I can see that
Reported-by Zhangfei Gao <[email protected]>
Quoted 7 lines
in the commit, but I don't have a clue what the actual report was, and there really isn't enough information in the commit itself, except for a fairly handwavy "Device drivers might, for instance, still need to flush operations.."
I don't want to know what device drivers _might_ do. I would want to have an actual pointer to what they do and where.
Another reason why these links are wanted: the ongoing regression tracking efforts can only scale with them, as they allow the regression tracking bot 'regzbot' to automatically connect tracked reports with patches that are posted or committed to fix tracked regressions.
--- scripts/checkpatch.pl | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
scripts/checkpatch.pl +14 -0
@@ -3144,6 +3144,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); } } + + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } + } } # Check Fixes: styles is correct
Signature · 1 line
2.35.1
Joe Perches
maintainerRe: [PATCH 1/2] checkpatch: warn when unknown tags are used for links
In reply to Kai WasserbächQuoted 6 lines · Kai Wasserbäch
Issue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --format=email -P --grep='^\w+:[ \t]*http' | \
Please use the --no-merges output
Quoted 23 lines
grep -Poh '^\w+:[ \t]*http' | \
sort | uniq -c | sort -rn | head -n 20
103889 Link: http
415 BugLink: http
372 Patchwork: http
270 Closes: http
221 Bug: http
121 References: http
101 v1: http
77 Bugzilla: http
60 URL: http
59 v2: http
37 Datasheet: http
35 v3: http
19 v4: http
12 v5: http
9 Ref: http
9 Fixes: http
9 Buglink: http
8 v6: http
8 Reference: http
7 V1: http
Quoted 8 lines
Some of these non-standard tags make it harder for external tools that rely on use of proper tags. One of those tools is the regression tracking bot 'regzbot', which looks out for "Link:" tags pointing to reports of tracked regressions.
The initial idea was to use a disallow list to raise an error when encountering known unwanted tags like BugLink:; during review it was requested to use a list of allowed tags instead[2].
[]
scripts/checkpatch.pl
Quoted 1 line
scripts/checkpatch.pl +0
-0
[]
scripts/checkpatch.pl sub process { · line 3246
Quoted 9 lines
scripts/checkpatch.pl +5
-0@@ -3239,6 +3239,13 @@ sub process { $commit_log_possible_stack_dump = 0; } +# Check for odd tags before a URI/URL + if ($in_commit_log && + $line =~ /^\s*(\w+):\s*http/ && $1 !~ /^Link/) { + WARN("COMMIT_LOG_USE_LINK", + "Unknown link reference '$1:', use 'Link:' instead.\n" . $herecurr);
This would allow LinkFoo: so better would be
$line =~ /^\s*(\w+):\s*http/ && $1 ne 'Link')
or
$line =~ /^\s*(\w+):\s*http/ && $1 !~ /^Link$/) {(and checkpatch doesn't use periods after output messages)
Maybe better as well would be to use something like the below to better describe the preferred location of patch versioning info.
# Check for odd tags before a URI/URL
if ($in_commit_log &&
$line =~ /^\s*(\w+):\s*http/ && $1 ne 'Link') {
if ($1 =~ /^v(?:ersion)?\d+/i) {
WARN("COMMIT_LOG_VERSIONING",
"Patch version information should be after the --- line\n" . $herecurr);
} else {
WARN("COMMIT_LOG_USE_LINK",
"Unknown link reference '$1:', use 'Link:' instead\n" . $herecurr);
}
}Joe Perches
maintainerRe: [PATCH 2/2] checkpatch: warn when Reported-by: is not followed by Link:
In reply to Kai WasserbächQuoted 22 lines · Kai Wasserbäch
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 15 lines
Again, the commit has a link to the patch *submission*, which is almost entirely useless. There's no link to the actual problem the patch fixes.
[...]
Put another way: I can see that
Reported-by Zhangfei Gao <[email protected]>in the commit, but I don't have a clue what the actual report was, and there really isn't enough information in the commit itself, except for a fairly handwavy "Device drivers might, for instance, still need to flush operations.."
Quoted 7 lines
Quoted 2 lines
I don't want to know what device drivers _might_ do. I would want to have an actual pointer to what they do and where.
Another reason why these links are wanted: the ongoing regression tracking efforts can only scale with them, as they allow the regression tracking bot 'regzbot' to automatically connect tracked reports with patches that are posted or committed to fix tracked regressions.
[]
scripts/checkpatch.pl
Quoted 1 line
scripts/checkpatch.pl +0
-0
[]
scripts/checkpatch.pl sub process { · line 3144
Quoted 2 lines
scripts/checkpatch.pl +0
-0@@ -3144,6 +3144,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);
I believe this use of '"$here\n" . $rawline . "\n"' to be a defect. I think this should just use $herecurr
And the unnecessary space before a newline is an abomination ;)
"Co-developed-by and Signed-off-by: name/email do not match\n" . $herecurr . $rawlines[$linenr]);
scripts/checkpatch.pl sub process { · line 3158
Quoted 12 lines
scripts/checkpatch.pl +12
-0+ + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);
Please use a space before and after a string concatenation '.'
English generally uses "a URL" and not "an URL"
https://www.techtarget.com/whatis/feature/Which-is-correct-a-URL-or-an-URL
Also the actual link line should likely be from lore
So maybe:
} elsif ($lines[$linenr] !~ m{https?://lore.kernel.org/.+}i) {
WARN("BAD_REPORTED_BY_LINK",
"Link: following Reported-by: should use a lore.kernel.org URL\n" . $herecurr . $rawlines[$linenr]);etc...
Re: [PATCH 2/2] checkpatch: warn when Reported-by: is not followed by Link:
Joe, many thx for your time and your valuable feedback. I agree with most of it and will look into addressing it tomorrow, but there is one area where I have a different option:
Quoted 16 lines · Joe Perches
Quoted 12 lines · Kai Wasserbäch
scripts/checkpatch.pl +12
-0+ + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);[...]
Quoted 6 lines
Also the actual link line should likely be from lore
So maybe:
} elsif ($lines[$linenr] !~ m{https?://lore.kernel.org/.+}i) { WARN("BAD_REPORTED_BY_LINK", "Link: following Reported-by: should use a lore.kernel.org URL\n" . $herecurr . $rawlines[$linenr]);
I'm pretty sure that's not want we want, as from regression tracking I known that we have other links that should continue to work here. Links to bugzilla.kernel.org immediately come to my mind, for example. Then there are some subsystems that use issue trackers on github or in gitlab instances (DRM for example), which must work here, too; and we occasionally even have links to bugs in bugzilla instances from RH or SUSE there, as sometimes valuable details for code archeologists can be found there.
Ciao, Thorsten
Joe Perches
maintainerRe: [PATCH 2/2] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 28 lines · Thorsten Leemhuis
Joe, many thx for your time and your valuable feedback. I agree with most of it and will look into addressing it tomorrow, but there is one area where I have a different option:
Quoted 16 lines · Joe Perches
Quoted 12 lines · Kai Wasserbäch
scripts/checkpatch.pl +12
-0+ + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);[...]
Quoted 6 lines
Also the actual link line should likely be from lore
So maybe:
} elsif ($lines[$linenr] !~ m{https?://lore.kernel.org/.+}i) { WARN("BAD_REPORTED_BY_LINK", "Link: following Reported-by: should use a lore.kernel.org URL\n" . $herecurr . $rawlines[$linenr]);
Quoted 8 lines
I'm pretty sure that's not want we want, as from regression tracking I known that we have other links that should continue to work here. Links to bugzilla.kernel.org immediately come to my mind, for example. Then there are some subsystems that use issue trackers on github or in gitlab instances (DRM for example), which must work here, too; and we occasionally even have links to bugs in bugzilla instances from RH or SUSE there, as sometimes valuable details for code archeologists can be found there.
Outside the fact there are relatively few existing uses of Link: after Reported-by:, it appears that "Fixes:" should also be allowed.
$ git log --no-merges --format=email -P -i --grep "^Reported-by:" -100000 | \
grep -A1 "^Reported-by:" | \
grep -v -P '^(Reported-by:|\-\-)' | \
cut -f1 -d":" | sort | uniq -c | sort -rn | head -20
31719 Signed-off-by
4556 Tested-by
4276 Cc
2976 Fixes
2304 Reviewed-by
1526 Acked-by
949 Suggested-by
821 Link
387
228 CC
139 Bugzilla
82 Reported-and-tested-by
69 Debugged-by
68 References
63 Bisected-by
43 Co-developed-by
34 LKML-Reference
29 Diagnosed-by
27 [ paulmck
26 Addresses-Coverity-IDand lore definitely dominates the Link: uses.
And IMO: the lkml.kernel.org entries should instead use lore.
Maybe there should be some other patterns allowed. Dunno.
$ git log --no-merges --format=email -P -i --grep "^Reported-by:" -100000 | \
grep -A1 "^Reported-by:" | \
grep '^Link:' | \
cut -f1-3 -d"/" | sort | uniq -c | sort -rn | head -20
551 Link: https://lore.kernel.org
67 Link: http://lkml.kernel.org
42 Link: https://bugzilla.kernel.org
28 Link: https://github.com
16 Link: https://lkml.kernel.org
11 Link: lore.kernel.org/r/[email protected]
10 Link: https://patchwork.freedesktop.org
10 Link: https://github.blog
9 Link: https://syzkaller.appspot.com
9 Link: https://lkml.org
6 Link: https://www.spinics.net
5 Link: https://gitlab.freedesktop.org
5 Link: https://bugs.debian.org
5 Link: http://lore.kernel.org
4 Link: https://www.openwall.com
4 Link: https://patchwork.kernel.org
4 Link: http://patchwork.freedesktop.org
3 Link: https://marc.info
3 Link: https://bugzilla.linux-nfs.org
2 Link: https://bugzilla.suse.comRe: [PATCH 2/2] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 38 lines · Joe Perches
Quoted 28 lines · Thorsten Leemhuis
Joe, many thx for your time and your valuable feedback. I agree with most of it and will look into addressing it tomorrow, but there is one area where I have a different option:
Quoted 16 lines · Joe Perches
Quoted 12 lines · Kai Wasserbäch
scripts/checkpatch.pl +12
-0+ + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);[...]
Quoted 6 lines
Also the actual link line should likely be from lore
So maybe:
} elsif ($lines[$linenr] !~ m{https?://lore.kernel.org/.+}i) { WARN("BAD_REPORTED_BY_LINK", "Link: following Reported-by: should use a lore.kernel.org URL\n" . $herecurr . $rawlines[$linenr]);Quoted 8 lines
I'm pretty sure that's not want we want, as from regression tracking I known that we have other links that should continue to work here. Links to bugzilla.kernel.org immediately come to my mind, for example. Then there are some subsystems that use issue trackers on github or in gitlab instances (DRM for example), which must work here, too; and we occasionally even have links to bugs in bugzilla instances from RH or SUSE there, as sometimes valuable details for code archeologists can be found there.
Quoted 2 lines
Outside the fact there are relatively few existing uses of Link: after Reported-by:,
Well, sure, because many people forgot to place them -- and this patch is trying to change that for reasons outlined in the commit description.
Quoted 1 line
it appears that "Fixes:" should also be allowed.
Well, that would defeat the purpose of this patch -- and there is no need to, as developers can still put Fixes: tag above or below the combo of Reported-by: and Link:.
Quoted 3 lines
and lore definitely dominates the Link: uses.
And IMO: the lkml.kernel.org entries should instead use lore.
Well, that might be something nice to have (if Konstantin thinks it's worth the trouble, as lkml.kernel.org likely needs to stick around anyway to not break existing links). But that IMHO is something independent of this patch-set, because the proper solution afaics then would be to check all Link: tags in the commit message, not only those next to a Reported-by.
Ciao, Thorsten
Re: [PATCH 1/2] checkpatch: warn when unknown tags are used for links
In reply to Joe PerchesQuoted 9 lines · Joe Perches
Quoted 6 lines · Kai Wasserbäch
Issue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --format=email -P --grep='^\w+:[ \t]*http' | \Please use the --no-merges output
Done, good point.
Quoted 23 lines
[...]
scripts/checkpatch.pl sub process { · line 3246
Quoted 9 lines
scripts/checkpatch.pl +5
-0@@ -3239,6 +3239,13 @@ sub process { $commit_log_possible_stack_dump = 0; } +# Check for odd tags before a URI/URL + if ($in_commit_log && + $line =~ /^\s*(\w+):\s*http/ && $1 !~ /^Link/) { + WARN("COMMIT_LOG_USE_LINK", + "Unknown link reference '$1:', use 'Link:' instead.\n" . $herecurr);This would allow LinkFoo: so better would be
$line =~ /^\s*(\w+):\s*http/ && $1 ne 'Link') or $line =~ /^\s*(\w+):\s*http/ && $1 !~ /^Link$/) {(and checkpatch doesn't use periods after output messages)
Maybe better as well would be to use something like the below to better describe the preferred location of patch versioning info.
# Check for odd tags before a URI/URL
Quoted 11 lines
if ($in_commit_log &&
$line =~ /^\s*(\w+):\s*http/ && $1 ne 'Link') {
if ($1 =~ /^v(?:ersion)?\d+/i) {
WARN("COMMIT_LOG_VERSIONING",
"Patch version information should be after the --- line\n" . $herecurr);
} else {
WARN("COMMIT_LOG_USE_LINK",
"Unknown link reference '$1:', use 'Link:' instead\n" . $herecurr);
}
}Yeah, that looks like a really good idea. I went with that. But I'd say this is the point where this really warrants a Co-developed-by: that mentions you (and thus a Signed-off-by: from your side, too), don't you think so?
Ciao, Thorsten
Re: [PATCH 2/2] checkpatch: warn when Reported-by: is not followed by Link:
In reply to Joe PerchesQuoted 39 lines · Joe Perches
Quoted 22 lines · Kai Wasserbäch
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 15 lines
Again, the commit has a link to the patch *submission*, which is almost entirely useless. There's no link to the actual problem the patch fixes.
[...]
Put another way: I can see that
Reported-by Zhangfei Gao <[email protected]>in the commit, but I don't have a clue what the actual report was, and there really isn't enough information in the commit itself, except for a fairly handwavy "Device drivers might, for instance, still need to flush operations.."
Quoted 7 lines
Quoted 2 lines
I don't want to know what device drivers _might_ do. I would want to have an actual pointer to what they do and where.
Another reason why these links are wanted: the ongoing regression tracking efforts can only scale with them, as they allow the regression tracking bot 'regzbot' to automatically connect tracked reports with patches that are posted or committed to fix tracked regressions.
[]
scripts/checkpatch.pl
Quoted 1 line
scripts/checkpatch.pl +0
-0[]
scripts/checkpatch.pl sub process { · line 3144
Quoted 2 lines
scripts/checkpatch.pl +0
-0@@ -3144,6 +3144,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);I believe this use of '"$here\n" . $rawline . "\n"' to be a defect. I think this should just use $herecurr
Quoted 3 lines
And the unnecessary space before a newline is an abomination ;)
"Co-developed-by and Signed-off-by: name/email do not match\n" . $herecurr . $rawlines[$linenr]);
Well, that's existing code. Want me to add a separate patch that fixes both of these aspects up in that area?
Quoted 6 lines
Quoted 6 lines
scripts/checkpatch.pl +6
-0+ + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline);
Quoted 8 lines
scripts/checkpatch.pl sub process { · line 3158
Quoted 6 lines
scripts/checkpatch.pl +6
-0+ } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain an URL\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);Please use a space before and after a string concatenation '.'
Okay, went with " . $herecurr . $rawlines[$linenr]" here, the result seems to be the same afaics.
Quoted 2 lines
English generally uses "a URL" and not "an URL"
https://www.techtarget.com/whatis/feature/Which-is-correct-a-URL-or-an-URL
Thx!
/me grumbles, he should have remembered that
Ciao, Thorsten
Joe Perches
maintainerRe: [PATCH 1/2] checkpatch: warn when unknown tags are used for links
In reply to Thorsten LeemhuisQuoted 4 lines · Thorsten Leemhuis
I'd say
this is the point where this really warrants a Co-developed-by: that mentions you (and thus a Signed-off-by: from your side, too), don't you think so?
No. I think it's just being a good maintainer and offering advice.
I'll ack it when it's submitted and I don't have any other comments.
cheers, Joe
Joe Perches
maintainerRe: [PATCH 2/2] checkpatch: warn when Reported-by: is not followed by Link:
In reply to Thorsten LeemhuisQuoted 2 lines · Thorsten Leemhuis
Quoted 1 line · Joe Perches
On Thu, 2022-12-08 at 20:32 +0100, Kai Wasserbäch wrote:
[]
Quoted 8 lines
Quoted 8 lines
scripts/checkpatch.pl
Quoted 1 line
scripts/checkpatch.pl +0
-0[]
scripts/checkpatch.pl sub process { · line 3144
Quoted 2 lines
scripts/checkpatch.pl +0
-0@@ -3144,6 +3144,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]);I believe this use of '"$here\n" . $rawline . "\n"' to be a defect. I think this should just use $herecurr
Quoted 6 lines
Quoted 3 lines
And the unnecessary space before a newline is an abomination ;)
"Co-developed-by and Signed-off-by: name/email do not match\n" . $herecurr . $rawlines[$linenr]);Well, that's existing code. Want me to add a separate patch that fixes both of these aspects up in that area?
Sure, thanks.
v3 cover feat: checkpatch: warn about dicouraged link tags and missing links
In reply to Kai WasserbächHey,
Thorsten Leemhuis suggested the following two changes to checkpatch, which I hereby humbly submit for review. Please let me know if any changes would be required for acceptance.
The third patch came out of the preview rounds and fixes an odd behaviour noticed then.
The two main changes should help developers write commit messages in a way that's preferred by Linus and will greatly help Thorsten's regression tracking efforts.
Patches are against mm/mm-nonmm-stable (which apparently is the tree checkpatch changes are merged through).
Cheers,
Kai
v2->v3:
- address review feedback from Joe (grammer fixes, use of $herecurr, use of --no-merges in commit log quote) (Thorsten)
- warn when people try to add version information to the commit log using code suggested by Joe (Thorsten)
- add a patch to make existing code in a nearby area use $herecurr where it should, which Joe noticed during review
v1->v2:
- add commit messages explaining the reasoning (Thorsten)
- approach the usage of wrong tags with an allow list as suggested by Joe (Thorsten)
Kai Wasserbäch (2):
checkpatch: warn when unknown tags are used for links
checkpatch: warn when Reported-by: is not followed by Link:
Thorsten Leemhuis (1):
checkpatch: use proper way for show problematic line
scripts/checkpatch.pl | 34 ++++++++++++++++++++++++++++++---- 1 file changed, 30 insertions(+), 4 deletions(-)
Signature · 1 line
2.35.1
PATCH v3 3/3 checkpatch: use proper way for show problematic line
From: Thorsten Leemhuis <[email protected]>
Instead of using an unnecessarily complicated approach to print a line that is warned about, use $herecurr instead, just like everywhere else in checkpatch. While at it, remove a superfluous space in one of the changed lines, too.
Both problems were found by Joe Perches.
--- scripts/checkpatch.pl | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
scripts/checkpatch.pl +4 -4
@@ -3142,17 +3142,17 @@ sub process { if ($sign_off =~ /^co-developed-by:$/i) { if ($email eq $author) { WARN("BAD_SIGN_OFF", - "Co-developed-by: should not be used to attribute nominal patch author '$author'\n" . "$here\n" . $rawline); + "Co-developed-by: should not be used to attribute nominal patch author '$author'\n" . $herecurr); } if (!defined $lines[$linenr]) { WARN("BAD_SIGN_OFF", - "Co-developed-by: must be immediately followed by Signed-off-by:\n" . "$here\n" . $rawline); + "Co-developed-by: must be immediately followed by Signed-off-by:\n" . $herecurr); } elsif ($rawlines[$linenr] !~ /^\s*signed-off-by:\s*(.*)/i) { WARN("BAD_SIGN_OFF", - "Co-developed-by: must be immediately followed by Signed-off-by:\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + "Co-developed-by: must be immediately followed by Signed-off-by:\n" . $herecurr . $rawlines[$linenr] . "\n"); } elsif ($1 ne $email) { WARN("BAD_SIGN_OFF", - "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + "Co-developed-by and Signed-off-by: name/email do not match\n" . $herecurr . $rawlines[$linenr] . "\n"); } }
Signature · 1 line
2.35.1
PATCH v3 1/3 checkpatch: warn when unknown tags are used for links
In reply to Kai WasserbächIssue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --no-merges --format=email -P \
--grep='^\w+:[ \t]*http' | grep -Poh '^\w+:[ \t]*http' | \
sort | uniq -c | sort -rn | head -n 20
103958 Link: http
418 BugLink: http
372 Patchwork: http
280 Closes: http
224 Bug: http
123 References: http
84 Bugzilla: http
61 URL: http
42 v1: http
38 Datasheet: http
20 v2: http
9 Ref: http
9 Fixes: http
9 Buglink: http
8 v3: http
8 Reference: http
7 See: http
6 1: http
5 link: http
3 Link:httpSome of these non-standard tags make it harder for external tools that rely on use of proper tags. One of those tools is the regression tracking bot 'regzbot', which looks out for "Link:" tags pointing to reports of tracked regressions.
The initial idea was to use a disallow list to raise an error when encountering known unwanted tags like BugLink:; during review it was requested to use a list of allowed tags instead[2].
--- scripts/checkpatch.pl | 12 ++++++++++++ 1 file changed, 12 insertions(+)
scripts/checkpatch.pl +12 -0
@@ -3250,6 +3250,18 @@ sub process { $commit_log_possible_stack_dump = 0; } +# Check for odd tags before a URI/URL + if ($in_commit_log && + $line =~ /^\s*(\w+):\s*http/ && $1 ne 'Link') { + if ($1 =~ /^v(?:ersion)?\d+/i) { + WARN("COMMIT_LOG_VERSIONING", + "Patch version information should be after the --- line\n" . $herecurr); + } else { + WARN("COMMIT_LOG_USE_LINK", + "Unknown link reference '$1:', use 'Link:' instead\n" . $herecurr); + } + } + # Check for lines starting with a # if ($in_commit_log && $line =~ /^#/) { if (WARN("COMMIT_COMMENT_SYMBOL",
Signature · 1 line
2.35.1
PATCH v3 2/3 checkpatch: warn when Reported-by: is not followed by Link:
In reply to Kai WasserbächEncourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 10 lines
Again, the commit has a link to the patch *submission*, which is almost entirely useless. There's no link to the actual problem the patch fixes.
[...]
Put another way: I can see that
Reported-by Zhangfei Gao <[email protected]>
Quoted 7 lines
in the commit, but I don't have a clue what the actual report was, and there really isn't enough information in the commit itself, except for a fairly handwavy "Device drivers might, for instance, still need to flush operations.."
I don't want to know what device drivers _might_ do. I would want to have an actual pointer to what they do and where.
Another reason why these links are wanted: the ongoing regression tracking efforts can only scale with them, as they allow the regression tracking bot 'regzbot' to automatically connect tracked reports with patches that are posted or committed to fix tracked regressions.
--- scripts/checkpatch.pl | 14 ++++++++++++++ 1 file changed, 14 insertions(+)
scripts/checkpatch.pl +14 -0
@@ -3155,6 +3155,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); } } + + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($lines[$linenr] !~ /https?:\/\//i) { + WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain a URL\n" . $herecurr . $rawlines[$linenr] . "\n"); + } + } } # Check Fixes: styles is correct
Signature · 1 line
2.35.1
Joe Perches
maintainerRe: [PATCH 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 2 lines · Kai Wasserbäch
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report.
Please also expand Documentation/ as appropriate to encourage this too.
scripts/checkpatch.pl
Quoted 1 line
scripts/checkpatch.pl +0
-0
[]
scripts/checkpatch.pl sub process { · line 3164
Quoted 11 lines
scripts/checkpatch.pl +7
-0@@ -3155,6 +3155,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); } } + + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) {
This capture group isn't necessary and does not guarantee there is an actual link. Also this is allowing spaces before "Link:" when the reported-by test above does not allow spaces. Please be consistent. My preference would be to not allow spaces.
scripts/checkpatch.pl sub process { · line 3167
Quoted 3 lines
scripts/checkpatch.pl +3
-0+ WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($lines[$linenr] !~ /https?:\/\//i) {
To make reading the match pattern easier, prefer m{} when the expected content uses / instead of using / as a delimiter.
Quoted 7 lines
scripts/checkpatch.pl +4
-0+ WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain a URL\n" . $herecurr . $rawlines[$linenr] . "\n"); + } + } } # Check Fixes: styles is correct
Re: [PATCH 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 5 lines · Joe Perches
Quoted 2 lines · Kai Wasserbäch
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report.
Please also expand
Documentation/as appropriate to encourage this too.
Well, I already updated it a few moths ago, but yeah, will revisit the docs and make it even clearer if needed (I assume that's the case).
Quoted 18 lines
scripts/checkpatch.pl
Quoted 1 line
scripts/checkpatch.pl +0
-0[]
scripts/checkpatch.pl sub process { · line 3164
Quoted 11 lines
scripts/checkpatch.pl +7
-0@@ -3155,6 +3155,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); } } + + # check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($rawlines[$linenr] !~ /^\s*link:\s*(.*)/i) {This capture group isn't necessary and does not guarantee there is an actual link. Also this is allowing spaces before "Link:" when the reported-by test above does not allow spaces. Please be consistent. My preference would be to not allow spaces.
Good idea. FWIW, the section patch 3 changes has a space here as well:
```
> } elsif ($rawlines[$linenr] !~ /^\s*signed-off-by:\s*(.*)/i) {
```Hence I guess I'll remove it as well, unless you for some reason tell me that's a bad idea.
Quoted 6 lines
scripts/checkpatch.pl sub process { · line 3167
Quoted 3 lines
scripts/checkpatch.pl +3
-0+ WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($lines[$linenr] !~ /https?:\/\//i) {To make reading the match pattern easier, prefer m{} when the expected content uses / instead of using / as a delimiter.
Also a good idea. That being said:
I think I'll drop this check, as your comment made me think: why should we check for "https?://" here at all? If we want to check that a "Link:" is followed by a URI, then we should do that all the time, not just in this single case.
Or am I missing something?
Quoted 7 lines
scripts/checkpatch.pl sub process { · line 3174
Quoted 7 lines
scripts/checkpatch.pl +4
-0+ WARN("BAD_REPORTED_BY_LINK", + "Link: following Reported-by: should contain a URL\n" . $herecurr . $rawlines[$linenr] . "\n"); + } + } } # Check Fixes: styles is correct
Ciao, Thorsten
v4 cover checkpatch.pl: warn about discouraged tags and missing Link: tags
In reply to Kai WasserbächHi, please consider the following checkpatch.pl patches for review.
The first two changes make checkpatch.pl check for a few mistakes wrt to links to bug reports Linus recently complained about a few times. Avoiding those is also important for my regression tracking efforts a lot, as the automated tracking performed by regzbot relies on the proper usage of the Link: tag.
The third patch fixes a few small oddities noticed in existing code during review of the two changes.
Ciao, Thorsten
---v3->v4:
- address review feedback from Joe (do not allow leading spaces in matches, check if Link: is actually followed by a URL in one go, and use m{} for matching https://) (Thorsten)
- catch Reported-and-tested-by as well (Thorsten)
v2->v3:
- address review feedback from Joe (grammer fixes, use of $herecurr, use of --no-merges in commit log quote) (Thorsten)
- warn when people try to add version information to the commit log using code suggested by Joe (Thorsten)
- add a patch to make existing code in a nearby area use $herecurr where it should, which Joe noticed during review
v1->v2:
- add commit messages explaining the reasoning (Thorsten)
- approach the usage of wrong tags with an allow list as suggested by Joe (Thorsten)
Kai Wasserbäch (2):
checkpatch: warn when unknown tags are used for links
checkpatch: warn when Reported-by: is not followed by Link:
Thorsten Leemhuis (1):
checkpatch: use proper way for show problematic line
scripts/checkpatch.pl | 34 +++++++++++++++++++++++++++++----- 1 file changed, 29 insertions(+), 5 deletions(-)
base-commit: 5dc4c995db9eb45f6373a956eb1f69460e69e6d4
Signature · 1 line
2.39.0
PATCH v4 2/3 checkpatch: warn when Reported-by: is not followed by Link:
From: Kai Wasserbäch <[email protected]>
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 10 lines
Again, the commit has a link to the patch *submission*, which is almost entirely useless. There's no link to the actual problem the patch fixes.
[...]
Put another way: I can see that
Reported-by Zhangfei Gao <[email protected]>
Quoted 7 lines
in the commit, but I don't have a clue what the actual report was, and there really isn't enough information in the commit itself, except for a fairly handwavy "Device drivers might, for instance, still need to flush operations.."
I don't want to know what device drivers _might_ do. I would want to have an actual pointer to what they do and where.
Another reason why these links are wanted: the ongoing regression tracking efforts can only scale with them, as they allow the regression tracking bot 'regzbot' to automatically connect tracked reports with patches that are posted or committed to fix tracked regressions.
--- scripts/checkpatch.pl | 12 ++++++++++++ 1 file changed, 12 insertions(+)
scripts/checkpatch.pl +12 -0
@@ -3155,8 +3155,20 @@ sub process { "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); } } + +# check if Reported-by: is followed by a Link: + if ($sign_off =~ /^reported(?:|-and-tested)-by:$/i) { + if (!defined $lines[$linenr]) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } elsif ($rawlines[$linenr] !~ m{^link:\s*https?://}i) { + WARN("BAD_REPORTED_BY_LINK", + "Reported-by: should be immediately followed by Link: with a URL to the report\n" . $herecurr . $rawlines[$linenr] . "\n"); + } + } } + # Check Fixes: styles is correct if (!$in_header_lines && $line =~ /^\s*fixes:?\s*(?:commit\s*)?[0-9a-f]{5,}\b/i) {
Signature · 1 line
2.39.0
PATCH v4 3/3 checkpatch: use proper way for show problematic line
In reply to Thorsten LeemhuisInstead of using an unnecessarily complicated approach to print a line that is warned about, use $herecurr instead, just like everywhere else in checkpatch.
While at it, remove a superfluous space in one of the changed lines, too. In a unmodified line also remove a superfluous check for a space before a signed-off-by tag, to me consistent with the check at the start of the section.
All three problems were found by Joe Perches during review of new code inspired by the code modified here.
--- scripts/checkpatch.pl | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
scripts/checkpatch.pl +5 -5
@@ -3142,17 +3142,17 @@ sub process { if ($sign_off =~ /^co-developed-by:$/i) { if ($email eq $author) { WARN("BAD_SIGN_OFF", - "Co-developed-by: should not be used to attribute nominal patch author '$author'\n" . "$here\n" . $rawline); + "Co-developed-by: should not be used to attribute nominal patch author '$author'\n" . $herecurr); } if (!defined $lines[$linenr]) { WARN("BAD_SIGN_OFF", - "Co-developed-by: must be immediately followed by Signed-off-by:\n" . "$here\n" . $rawline); - } elsif ($rawlines[$linenr] !~ /^\s*signed-off-by:\s*(.*)/i) { + "Co-developed-by: must be immediately followed by Signed-off-by:\n" . $herecurr); + } elsif ($rawlines[$linenr] !~ /^signed-off-by:\s*(.*)/i) { WARN("BAD_SIGN_OFF", - "Co-developed-by: must be immediately followed by Signed-off-by:\n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + "Co-developed-by: must be immediately followed by Signed-off-by:\n" . $herecurr . $rawlines[$linenr] . "\n"); } elsif ($1 ne $email) { WARN("BAD_SIGN_OFF", - "Co-developed-by and Signed-off-by: name/email do not match \n" . "$here\n" . $rawline . "\n" .$rawlines[$linenr]); + "Co-developed-by and Signed-off-by: name/email do not match\n" . $herecurr . $rawlines[$linenr] . "\n"); } }
Signature · 1 line
2.39.0
PATCH v4 1/3 checkpatch: warn when unknown tags are used for links
In reply to Thorsten LeemhuisFrom: Kai Wasserbäch <[email protected]>
Issue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --no-merges --format=email -P \
--grep='^\w+:[ \t]*http' | grep -Poh '^\w+:[ \t]*http' | \
sort | uniq -c | sort -rn | head -n 20
103958 Link: http
418 BugLink: http
372 Patchwork: http
280 Closes: http
224 Bug: http
123 References: http
84 Bugzilla: http
61 URL: http
42 v1: http
38 Datasheet: http
20 v2: http
9 Ref: http
9 Fixes: http
9 Buglink: http
8 v3: http
8 Reference: http
7 See: http
6 1: http
5 link: http
3 Link:httpSome of these non-standard tags make it harder for external tools that rely on use of proper tags. One of those tools is the regression tracking bot 'regzbot', which looks out for "Link:" tags pointing to reports of tracked regressions.
The initial idea was to use a disallow list to raise an error when encountering known unwanted tags like BugLink:; during review it was requested to use a list of allowed tags instead[2].
--- scripts/checkpatch.pl | 12 ++++++++++++ 1 file changed, 12 insertions(+)
scripts/checkpatch.pl +12 -0
@@ -3250,6 +3250,18 @@ sub process { $commit_log_possible_stack_dump = 0; } +# Check for odd tags before a URI/URL + if ($in_commit_log && + $line =~ /^\s*(\w+):\s*http/ && $1 ne 'Link') { + if ($1 =~ /^v(?:ersion)?\d+/i) { + WARN("COMMIT_LOG_VERSIONING", + "Patch version information should be after the --- line\n" . $herecurr); + } else { + WARN("COMMIT_LOG_USE_LINK", + "Unknown link reference '$1:', use 'Link:' instead\n" . $herecurr); + } + } + # Check for lines starting with a # if ($in_commit_log && $line =~ /^#/) { if (WARN("COMMIT_COMMENT_SYMBOL",
Signature · 1 line
2.39.0
Joe Perches
maintainerRe: [PATCH v4 0/3] checkpatch.pl: warn about discouraged tags and missing Link: tags
In reply to Thorsten LeemhuisQuoted 10 lines · Thorsten Leemhuis
Hi, please consider the following checkpatch.pl patches for review.
The first two changes make checkpatch.pl check for a few mistakes wrt to links to bug reports Linus recently complained about a few times. Avoiding those is also important for my regression tracking efforts a lot, as the automated tracking performed by regzbot relies on the proper usage of the Link: tag.
The third patch fixes a few small oddities noticed in existing code during review of the two changes.
Hey Andrew. Please forward this patch series upstream.
Thanks, Joe
Quoted 34 lines
Ciao, Thorsten
---v3->v4:
- address review feedback from Joe (do not allow leading spaces in matches, check if Link: is actually followed by a URL in one go, and use m{} for matching https://) (Thorsten)
- catch Reported-and-tested-by as well (Thorsten)
v2->v3:
- address review feedback from Joe (grammer fixes, use of $herecurr, use of --no-merges in commit log quote) (Thorsten)
- warn when people try to add version information to the commit log using code suggested by Joe (Thorsten)
- add a patch to make existing code in a nearby area use $herecurr where it should, which Joe noticed during review
v1->v2:
- add commit messages explaining the reasoning (Thorsten)
- approach the usage of wrong tags with an allow list as suggested by Joe (Thorsten)Kai Wasserbäch (2):
checkpatch: warn when unknown tags are used for links
checkpatch: warn when Reported-by: is not followed by Link:Thorsten Leemhuis (1):
checkpatch: use proper way for show problematic linescripts/checkpatch.pl | 34 +++++++++++++++++++++++++++++----- 1 file changed, 29 insertions(+), 5 deletions(-)base-commit:
5dc4c995db9eb45f6373a956eb1f69460e69e6d4
Re: [PATCH v4 1/3] checkpatch: warn when unknown tags are used for links
In reply to Thorsten LeemhuisHello,
Quoted 31 lines · Thorsten Leemhuis
From: Kai Wasserbäch <[email protected]>
Issue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --no-merges --format=email -P \ --grep='^\w+:[ \t]*http' | grep -Poh '^\w+:[ \t]*http' | \ sort | uniq -c | sort -rn | head -n 20 103958 Link: http 418 BugLink: http 372 Patchwork: http 280 Closes: http 224 Bug: http 123 References: http 84 Bugzilla: http 61 URL: http 42 v1: http 38 Datasheet: http 20 v2: http 9 Ref: http 9 Fixes: http 9 Buglink: http 8 v3: http 8 Reference: http 7 See: http 6 1: http 5 link: http 3 Link:http
Quoted 4 lines
Some of these non-standard tags make it harder for external tools that rely on use of proper tags. One of those tools is the regression tracking bot 'regzbot', which looks out for "Link:" tags pointing to reports of tracked regressions.
I'm sorry for the late feedback but would it be possible to add an exception for the "Closes" tag followed by a URL?
This tag is useful -- at least for us when maintaining the MPTCP subtree -- to have tickets being automatically closed when a patch is accepted. I don't think this "Closes" tag is a "random one that makes no sense" but I agree it is not an "official" one described in the documentation.
On our side, we are using GitHub to manage issues but this also works with GitLab and probably others. Other keywords are also accepted [1][2] but I guess it is best to stick with one, especially when it is already used according to the list provided above.
Would it then be OK to allow this "Closes" tag in checkpatch.pl and mention it in the documentation (Submitting patches)?
Or should we switch to the "Link" tag instead (and re-do the tracking manually)?
Cheers,
Matt
[1]
https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/using-keywords-in-issues-and-pull-requests
[2]
https://docs.gitlab.com/ee/user/project/issues/managing_issues.html#default-closing-pattern
Signature · 2 lines
Tessares | Belgium | Hybrid Access Solutions www.tessares.net
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
In reply to Thorsten LeemhuisQuoted 8 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
From: Kai Wasserbäch <[email protected]>
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link. And regression tracking doesn't seem to happen much on syzbot reports either.
I like the addition otherwise, it's already catching missing links in netdev land!
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 10 lines · Jakub Kicinski
Quoted 8 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
From: Kai Wasserbäch <[email protected]>
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 2 lines
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link.
Hmmm. Not sure. Every special case makes things harder for humans and software that looks at a commits downstream. Clicking on a link also makes things easy for code archaeologists that might look into the issue months or years later (which might not even know how to find the report and potential discussions on lore from the syzbot report ID).
Hence, wouldn't it be better to ask the syzbot folks to change their reporting slightly and suggest something like this instead in their reports (the last line is the new one):
``` IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: [email protected] Link: https://lore.kernel.org/r/[email protected]/ ```
This might not be to hard if they known the message-id in advance. Maybe they could even use the syzbot report ID as msg-id to make things even easier. And for developers not much would change afaics, they just need to copy and paste two lines instead of one.
Quoted 2 lines
And regression tracking doesn't seem to happen much on syzbot reports either.
Yeah, right now I most of the time stay away from CI reports and leave the tracking to the people that run the CI (unless it's something I consider worth tracking), but I hope that might change over time to have things in one place.
Quoted 2 lines
I like the addition otherwise, it's already catching missing links in netdev land!
Thx for saying this!
Ciao, Thorsten
Re: [PATCH v4 1/3] checkpatch: warn when unknown tags are used for links
In reply to Matthieu BaertsQuoted 19 lines · Matthieu Baerts
Quoted 18 lines · Thorsten Leemhuis
From: Kai Wasserbäch <[email protected]>
Issue a warning when encountering URLs behind unknown tags, as Linus recently stated ```please stop making up random tags that make no sense. Just use "Link:"```[1]. That statement was triggered by an use of 'BugLink', but that's not the only tag people invented:
$ git log -100000 --no-merges --format=email -P \ --grep='^\w+:[ \t]*http' | grep -Poh '^\w+:[ \t]*http' | \ sort | uniq -c | sort -rn | head -n 20 103958 Link: http 418 BugLink: http 372 Patchwork: http 280 Closes: http 224 Bug: http 123 References: http [...]
Quoted 7 lines
Quoted 4 lines
Some of these non-standard tags make it harder for external tools that rely on use of proper tags. One of those tools is the regression tracking bot 'regzbot', which looks out for "Link:" tags pointing to reports of tracked regressions.
I'm sorry for the late feedback but would it be possible to add an exception for the "Closes" tag followed by a URL?
As I just wrote in a reply to Jakub: Not sure. Every special case makes things harder for humans and software that looks at a commits downstream.
Quoted 13 lines
This tag is useful -- at least for us when maintaining the MPTCP subtree -- to have tickets being automatically closed when a patch is accepted. I don't think this "Closes" tag is a "random one that makes no sense" but I agree it is not an "official" one described in the documentation.
On our side, we are using GitHub to manage issues but this also works with GitLab and probably others. Other keywords are also accepted [1][2] but I guess it is best to stick with one, especially when it is already used according to the list provided above.
Would it then be OK to allow this "Closes" tag in checkpatch.pl and mention it in the documentation (Submitting patches)?
Quoted 7 lines
Or should we switch to the "Link" tag instead (and re-do the tracking manually)?
[1]
https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/using-keywords-in-issues-and-pull-requests
[2]
https://docs.gitlab.com/ee/user/project/issues/managing_issues.html#default-closing-pattern
For the record, let me repeat and further elaborate what I already said on social media before you wrote your mail:
* I'm not mostly neutral here, but it was Linus who wrote "please stop
making up random tags that make no sense." in [1]. This was triggered by a use of "BugLink:"; maybe there are tools out there that rely on that tag, hence their users might ask for a exception as well. That's why I think it's Linus call to grant any exceptions.
[1]
https://lore.kernel.org/all/CAHk-=wgs38ZrfPvy=nOwVkVzjpM3VFU1zobP37Fwd_h9iAD5JQ@mail.gmail.com/
* if such an exception is made, it IMHO must be documented in our
documentation, so any software and humans that rely on these tags are aware of it.
Ciao, Thorsten
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
In reply to Thorsten LeemhuisQuoted 9 lines · On Thu, 2 Mar 2023 06:17:22 +0100 Thorsten Leemhuis wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 8 lines
Quoted 2 lines
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link.
Hmmm. Not sure. Every special case makes things harder for humans and software that looks at a commits downstream. Clicking on a link also makes things easy for code archaeologists that might look into the issue months or years later (which might not even know how to find the report and potential discussions on lore from the syzbot report ID).
No other system comes close to syzbot in terms of reporting meaningful bugs, IMHO special casing it doesn't risk creep.
Interestingly other bots attach links which are 100% pointless noise:
Oh, eh. Let's see how noisy this check is once the merge window is over.
Quoted 10 lines
Hence, wouldn't it be better to ask the syzbot folks to change their reporting slightly and suggest something like this instead in their reports (the last line is the new one):
``` IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: [email protected] Link: https://lore.kernel.org/r/[email protected]/ ```
Quoted 4 lines
This might not be to hard if they known the message-id in advance. Maybe they could even use the syzbot report ID as msg-id to make things even easier. And for developers not much would change afaics, they just need to copy and paste two lines instead of one.
Dmitry, WDYT?
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 40 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 2 Mar 2023 06:17:22 +0100 Thorsten Leemhuis wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 8 lines
Quoted 2 lines
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link.
Hmmm. Not sure. Every special case makes things harder for humans and software that looks at a commits downstream. Clicking on a link also makes things easy for code archaeologists that might look into the issue months or years later (which might not even know how to find the report and potential discussions on lore from the syzbot report ID).
No other system comes close to syzbot in terms of reporting meaningful bugs, IMHO special casing it doesn't risk creep.
Interestingly other bots attach links which are 100% pointless noise:
Oh, eh. Let's see how noisy this check is once the merge window is over.
Quoted 10 lines
Hence, wouldn't it be better to ask the syzbot folks to change their reporting slightly and suggest something like this instead in their reports (the last line is the new one):
``` IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: [email protected] Link: https://lore.kernel.org/r/[email protected]/ ```
Quoted 6 lines
Quoted 4 lines
This might not be to hard if they known the message-id in advance. Maybe they could even use the syzbot report ID as msg-id to make things even easier. And for developers not much would change afaics, they just need to copy and paste two lines instead of one.
Dmitry, WDYT?
Hi Jakub, Thorsten,
Adding a Link to syzbot reports should be relatively trivial.
Ted proposed to use Link _instead_ of Reported-by:
https://github.com/google/syzkaller/issues/3596
Quoted 5 lines
in fact, it might be nice if we could encourage upstream developers put in the commit trailer:
in addition to, or better yet, instead of:
Reported-by [email protected]
We could also use a link in the Reported-by tag, e.g.:
Some folks parse Reported-by to collect stats.
What is better?
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 40 lines · Dmitry Vyukov
Quoted 39 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 2 Mar 2023 06:17:22 +0100 Thorsten Leemhuis wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 8 lines
Quoted 2 lines
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link.
Hmmm. Not sure. Every special case makes things harder for humans and software that looks at a commits downstream. Clicking on a link also makes things easy for code archaeologists that might look into the issue months or years later (which might not even know how to find the report and potential discussions on lore from the syzbot report ID).
No other system comes close to syzbot in terms of reporting meaningful bugs, IMHO special casing it doesn't risk creep.
Interestingly other bots attach links which are 100% pointless noise:
Oh, eh. Let's see how noisy this check is once the merge window is over.
Quoted 10 lines
Hence, wouldn't it be better to ask the syzbot folks to change their reporting slightly and suggest something like this instead in their reports (the last line is the new one):
``` IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: [email protected] Link: https://lore.kernel.org/r/[email protected]/ ```
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
This might not be to hard if they known the message-id in advance. Maybe they could even use the syzbot report ID as msg-id to make things even easier. And for developers not much would change afaics, they just need to copy and paste two lines instead of one.
Dmitry, WDYT?
Adding a Link to syzbot reports should be relatively trivial.
Sounds good.
Quoted 8 lines
Ted proposed to use Link _instead_ of Reported-by:
https://github.com/google/syzkaller/issues/3596Quoted 5 lines
in fact, it might be nice if we could encourage upstream developers put in the commit trailer:
in addition to, or better yet, instead of:
Reported-by [email protected]
Quoted 7 lines
We could also use a link in the Reported-by tag, e.g.:
Some folks parse Reported-by to collect stats.
What is better?
Here are my thoughts:
* we should definitely have a "Link:" to the report in lore, as that's the long-term archive under our own control and also where discussions happen after the report was posted; but I'm biased here, as such a tag would make tracking with regzbot a no-brainer ;)
* "Reported-by:" IMHO should stay for the hat tip and stats aspects; I don't care if it includes the syzbot report ID or not (omitting it might be good for the stats aspects and is more friendly to the eyes, but those are just details)
* a Link: to the syzkaller web ui might be nice, too -- and likely is the easiest thing to look out for on the syzbot server side
IOW something like this maybe:
Something like the following would look more normal, but of course is only possible if syzbot starts out to look for such Link: tags (not sure if the msgid is valid here, but you get the idea):
Link:
https://lore.kernel.org/r/syzbot+cafecafecaca0cafecafe-syzkaller-appspotmail-com@google.com/
Ciao, Thorsten
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 90 lines · Thorsten Leemhuis
Quoted 40 lines · Dmitry Vyukov
Quoted 39 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 2 Mar 2023 06:17:22 +0100 Thorsten Leemhuis wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 8 lines
Quoted 2 lines
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link.
Hmmm. Not sure. Every special case makes things harder for humans and software that looks at a commits downstream. Clicking on a link also makes things easy for code archaeologists that might look into the issue months or years later (which might not even know how to find the report and potential discussions on lore from the syzbot report ID).
No other system comes close to syzbot in terms of reporting meaningful bugs, IMHO special casing it doesn't risk creep.
Interestingly other bots attach links which are 100% pointless noise:
Oh, eh. Let's see how noisy this check is once the merge window is over.
Quoted 10 lines
Hence, wouldn't it be better to ask the syzbot folks to change their reporting slightly and suggest something like this instead in their reports (the last line is the new one):
``` IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: [email protected] Link: https://lore.kernel.org/r/[email protected]/ ```Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
This might not be to hard if they known the message-id in advance. Maybe they could even use the syzbot report ID as msg-id to make things even easier. And for developers not much would change afaics, they just need to copy and paste two lines instead of one.
Dmitry, WDYT?
Adding a Link to syzbot reports should be relatively trivial.
Sounds good.
Quoted 8 lines
Ted proposed to use Link _instead_ of Reported-by:
https://github.com/google/syzkaller/issues/3596Quoted 5 lines
in fact, it might be nice if we could encourage upstream developers put in the commit trailer:
in addition to, or better yet, instead of:
Reported-by [email protected]Quoted 7 lines
We could also use a link in the Reported-by tag, e.g.:
Some folks parse Reported-by to collect stats.
What is better?
Here are my thoughts:
* we should definitely have a "Link:" to the report in lore, as that's the long-term archive under our own control and also where discussions happen after the report was posted; but I'm biased here, as such a tag would make tracking with regzbot a no-brainer ;)
* "Reported-by:" IMHO should stay for the hat tip and stats aspects; I don't care if it includes the syzbot report ID or not (omitting it might be good for the stats aspects and is more friendly to the eyes, but those are just details)
* a Link: to the syzkaller web ui might be nice, too -- and likely is the easiest thing to look out for on the syzbot server side
IOW something like this maybe:
Quoted 7 lines
Something like the following would look more normal, but of course is only possible if syzbot starts out to look for such Link: tags (not sure if the msgid is valid here, but you get the idea):
Reported-by [email protected]Link:
https://lore.kernel.org/r/syzbot+cafecafecaca0cafecafe-syzkaller-appspotmail-com@google.com/
Oh, you mean lore link.
We can parse out our hash from any tag, but the problem is that the current email api we use, does not allow to specify Message-ID before sending, so we don't know it when generating the text.
We don't even know it after sending, the API is super simple:
https://pkg.go.dev/google.golang.org/appengine/mail
So we don't know what the lore link will be...
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 98 lines · Dmitry Vyukov
Quoted 89 lines · Thorsten Leemhuis
Quoted 40 lines · Dmitry Vyukov
Quoted 39 lines · Jakub Kicinski
Quoted 9 lines · On Thu, 2 Mar 2023 06:17:22 +0100 Thorsten Leemhuis wrote:
Quoted 8 lines · Jakub Kicinski
Quoted 6 lines · On Fri, 20 Jan 2023 13:35:19 +0100 Thorsten Leemhuis wrote:
Encourage patch authors to link to reports by issuing a warning, if a Reported-by: is not accompanied by a link to the report. Those links are often extremely useful for any code archaeologist that wants to know more about the backstory of a change than the commit message provides. That includes maintainers higher up in the patch-flow hierarchy, which is why Linus asks developers to add such links [1, 2, 3]. To quote [1]:
Quoted 8 lines
Quoted 2 lines
Is it okay if we exclude syzbot reports from this rule? If full syzbot report ID is provided - it's as good as a link.
Hmmm. Not sure. Every special case makes things harder for humans and software that looks at a commits downstream. Clicking on a link also makes things easy for code archaeologists that might look into the issue months or years later (which might not even know how to find the report and potential discussions on lore from the syzbot report ID).
No other system comes close to syzbot in terms of reporting meaningful bugs, IMHO special casing it doesn't risk creep.
Interestingly other bots attach links which are 100% pointless noise:
Oh, eh. Let's see how noisy this check is once the merge window is over.
Quoted 10 lines
Hence, wouldn't it be better to ask the syzbot folks to change their reporting slightly and suggest something like this instead in their reports (the last line is the new one):
``` IMPORTANT: if you fix the issue, please add the following tag to the commit: Reported-by: [email protected] Link: https://lore.kernel.org/r/[email protected]/ ```Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
This might not be to hard if they known the message-id in advance. Maybe they could even use the syzbot report ID as msg-id to make things even easier. And for developers not much would change afaics, they just need to copy and paste two lines instead of one.
Dmitry, WDYT?
Adding a Link to syzbot reports should be relatively trivial.
Sounds good.
Quoted 8 lines
Ted proposed to use Link _instead_ of Reported-by:
https://github.com/google/syzkaller/issues/3596Quoted 5 lines
in fact, it might be nice if we could encourage upstream developers put in the commit trailer:
in addition to, or better yet, instead of:
Reported-by [email protected]Quoted 7 lines
We could also use a link in the Reported-by tag, e.g.:
Some folks parse Reported-by to collect stats.
What is better?
Here are my thoughts:
* we should definitely have a "Link:" to the report in lore, as that's the long-term archive under our own control and also where discussions happen after the report was posted; but I'm biased here, as such a tag would make tracking with regzbot a no-brainer ;)
* "Reported-by:" IMHO should stay for the hat tip and stats aspects; I don't care if it includes the syzbot report ID or not (omitting it might be good for the stats aspects and is more friendly to the eyes, but those are just details)
* a Link: to the syzkaller web ui might be nice, too -- and likely is the easiest thing to look out for on the syzbot server side
IOW something like this maybe:
Quoted 7 lines
Something like the following would look more normal, but of course is only possible if syzbot starts out to look for such Link: tags (not sure if the msgid is valid here, but you get the idea):
Reported-by [email protected]Link:
https://lore.kernel.org/r/syzbot+cafecafecaca0cafecafe-syzkaller-appspotmail-com@google.com/
Quoted 8 lines
Oh, you mean lore link.
We can parse out our hash from any tag, but the problem is that the current email api we use, does not allow to specify Message-ID before sending, so we don't know it when generating the text.
We don't even know it after sending, the API is super simple:
https://pkg.go.dev/google.golang.org/appengine/mail
So we don't know what the lore link will be...
That's... unfortunate, as from my understanding of things that would be the most important "Link:" to have in any patches that fix issues report by syzbot. But well, that's how it is for now. In that case I'd vote for this:
Regzbot can handle this, as long as somebody tells it about that URL.
IOW: it creates a little extra work for some human. But that is not much of a problem, especially as of now, as I only track syzbot reports that for one reason or another make me go "I should better track this".
Ciao, Thorsten
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 7 lines · On Thu, 2 Mar 2023 10:48:22 +0100 Thorsten Leemhuis <[email protected]> wrote:
Quoted 6 lines
We can parse out our hash from any tag, but the problem is that the current email api we use, does not allow to specify Message-ID before sending, so we don't know it when generating the text.
We don't even know it after sending, the API is super simple:
https://pkg.go.dev/google.golang.org/appengine/mail
So we don't know what the lore link will be...
Quoted 7 lines
That's... unfortunate, as from my understanding of things that would be the most important "Link:" to have in any patches that fix issues report by syzbot. But well, that's how it is for now. In that case I'd vote for this:
As you previously mentioned, patch preparers should also include the lore link so any followup discussion is easily located.
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 12 lines · Andrew Morton
Quoted 12 lines
Quoted 6 lines
We can parse out our hash from any tag, but the problem is that the current email api we use, does not allow to specify Message-ID before sending, so we don't know it when generating the text.
We don't even know it after sending, the API is super simple:
https://pkg.go.dev/google.golang.org/appengine/mail
So we don't know what the lore link will be...That's... unfortunate, as from my understanding of things that would be the most important "Link:" to have in any patches that fix issues report by syzbot. But well, that's how it is for now. In that case I'd vote for this:
Quoted 5 lines
Quoted 2 lines
As you previously mentioned, patch preparers should also include the lore link so any followup discussion is easily located.
If the link we need to include is to lore, then we don't need to change the current syzbot Reported-by, right? Instead of asking 3 tags, we can ask only for:
But as I mentioned we can't provide the lore link at the moment, we can only add a text to ask to include it.
This also means that checkpatch does not need special casing for syzbot.
Re: [PATCH v4 2/3] checkpatch: warn when Reported-by: is not followed by Link:
Quoted 19 lines · Dmitry Vyukov
Quoted 12 lines · Andrew Morton
Quoted 12 lines
Quoted 6 lines
We can parse out our hash from any tag, but the problem is that the current email api we use, does not allow to specify Message-ID before sending, so we don't know it when generating the text.
We don't even know it after sending, the API is super simple:
https://pkg.go.dev/google.golang.org/appengine/mail
So we don't know what the lore link will be...That's... unfortunate, as from my understanding of things that would be the most important "Link:" to have in any patches that fix issues report by syzbot. But well, that's how it is for now. In that case I'd vote for this:
Quoted 5 lines
Quoted 2 lines
As you previously mentioned, patch preparers should also include the lore link so any followup discussion is easily located.
Quoted 6 lines
If the link we need to include is to lore, then we don't need to change the current syzbot Reported-by, right? Instead of asking 3 tags, we can ask only for:
Reported-by [email protected]
Yeah, that's not perfect (see below), but at least better. As mentioned
earlier: if the Reported-by: includes the sysbot-id (e.g. the df61b36319e045c00a08) is up to you.
Quoted 2 lines
But as I mentioned we can't provide the lore link at the moment, we can only add a text to ask to include it.
Yeah, that would be good. Normally it's the oblation of the developer anyway to add Link: tags to any report (which most of the time means: in lore) when fixing things. Obviously the chance that they actually do it is a lot bigger when syzbot would suggest it.
Quoted 1 line
This also means that checkpatch does not need special casing for syzbot.
Yup
Ciao, Thorsten