* 'master' of github.com:git/git:
Git 2.34-rc2
parse-options.[ch]: revert use of "enum" for parse_options()
t/lib-git.sh: fix ACL-related permissions failure
A few fixes before -rc2
async_die_is_recursing: work around GCC v11.x issue on Fedora
Document positive variant of commit and merge option "--no-verify"
pull: honor --no-verify and do not call the commit-msg hook
http-backend: remove a duplicated code branch
Fix ssh-signing test to work on a platform where the default ACL is
overly loose to upset OpenSSH (reported on an installation of Cygwin).
* ad/ssh-signing-testfix:
t/lib-git.sh: fix ACL-related permissions failure
We currently throw away any arguments given to "git jump merge". We
should instead pass them along to ls-files, since they're likely to be
pathspecs. This matches the behavior of "git jump diff", etc.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Revert the parse_options() prototype change in my recent
352e761388 (parse-options.[ch]: consistently use "enum
parse_opt_result", 2021-10-08) was incorrect. The parse_options()
function returns the number of argc elements that haven't been
processed, not "enum parse_opt_result".
Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The description that 0640 makes sure that the group members can read
the repository is correct, but calling that octal number a <umask>
is wrong. Let's call it <perm>, as the value is used to set the
permission bits.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The previous explanation was mixing the format with the identity of
the field.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Each member of the pair is explained but they are not defined
beforehand.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
URL being an acronym, it deserves to be kept uppercase.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
That's how alternative options are expressed in general.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
According to CodingGuidelines, multi-word placeholders should use
hyphens as word separators.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reviewed-by: Eli Schwartz <eschwartz@archlinux.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This discerns user inputs from verbatim options in the synopsis.
Signed-off-by: Jean-Noël Avila <jn.avila@free.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As well as checking that the relevant functionality is available, the
GPGSSH prerequisite check creates the SSH keys that are used by the test
functions it gates. If these keys are created in a directory that
has a default Access Control List, the key files can inherit those
permissions.
This can result in a scenario where the private keys are created
successfully, so the prerequisite check passes and the tests are run,
but the key files have permissions that are too permissive, meaning
OpenSSH will refuse to load them and the tests will fail.
To avoid this happening, before creating the keys, clear any default ACL
set on the directory that will contain them. This step allowed to fail;
if setfacl isn't present, that's a very likely indicator that the
filesystem in question simply doesn't support default ACLs.
Helped-by: Fabian Stelzer <fs@gigacodes.de>
Signed-off-by: Adam Dinwoodie <adam@dinwoodie.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change the "define_categories()" and "define_category_names()" functions
to take the already-parsed output of "category_list()" as an argument,
which brings our number of passes over "command-list.txt" from three
to two.
Then have "category_list()" itself take the output of "command_list()"
as an argument, bringing the number of times we parse the file to one.
Compared to the pre-image this speeds us up quite a bit:
$ git show HEAD~:generate-cmdlist.sh >generate-cmdlist.sh.old
$ hyperfine --warmup 10 -L v ,.old 'sh generate-cmdlist.sh{v} command-list.txt'
Benchmark #1: sh generate-cmdlist.sh command-list.txt
Time (mean ± σ): 22.9 ms ± 0.3 ms [User: 15.8 ms, System: 9.6 ms]
Range (min … max): 22.5 ms … 24.0 ms 125 runs
Benchmark #2: sh generate-cmdlist.sh.old command-list.txt
Time (mean ± σ): 30.1 ms ± 0.4 ms [User: 24.4 ms, System: 17.5 ms]
Range (min … max): 29.5 ms … 32.3 ms 96 runs
Summary
'sh generate-cmdlist.sh command-list.txt' ran
1.32 ± 0.02 times faster than 'sh generate-cmdlist.sh.old command-list.txt'
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Replace the "grep" we run to exclude certain programs from the
generated output with a pure-shell loop that strips out the comments,
and sees if the "cmd" we're reading is on a list of excluded
programs. This uses a trick similar to test_have_prereq() in
test-lib-functions.sh.
On my *nix system this makes things quite a bit slower compared to
HEAD~:
o
'sh generate-cmdlist.sh.old command-list.txt' ran
1.56 ± 0.11 times faster than 'sh generate-cmdlist.sh command-list.txt'
18.00 ± 0.19 times faster than 'sh generate-cmdlist.sh.master command-list.txt'
But when I tried running generate-cmdlist.sh 100 times in CI I found
that it helped across the board even on OSX & Linux. I tried testing
it in CI with this ad-hoc few-liner:
for i in $(seq -w 0 11 | sort -nr)
do
git show HEAD~$i:generate-cmdlist.sh >generate-cmdlist-HEAD$i.sh &&
git add generate-cmdlist* &&
cp t/t0000-generate-cmdlist.sh t/t00$i-generate-cmdlist.sh || : &&
perl -pi -e "s/HEAD0/HEAD$i/g" t/t00$i-generate-cmdlist.sh &&
git add t/t00*.sh
done && git commit -m"generated it"
Here HEAD~02 and the t0002* file refers to this change, and HEAD~03
and t0003* file to the preceding commit, the relevant results were:
linux-gcc:
[12:05:33] t0002-generate-cmdlist.sh .. ok 14 ms ( 0.00 usr 0.00 sys + 3.64 cusr 3.09 csys = 6.73 CPU)
[12:05:30] t0003-generate-cmdlist.sh .. ok 32 ms ( 0.00 usr 0.00 sys + 2.66 cusr 1.81 csys = 4.47 CPU)
osx-gcc:
[11:58:04] t0002-generate-cmdlist.sh .. ok 80081 ms ( 0.02 usr 0.02 sys + 17.80 cusr 10.07 csys = 27.91 CPU)
[11:58:16] t0003-generate-cmdlist.sh .. ok 92127 ms ( 0.02 usr 0.01 sys + 22.54 cusr 14.27 csys = 36.84 CPU)
vs-test:
[12:03:14] t0002-generate-cmdlist.sh .. ok 30 s ( 0.02 usr 0.00 sys + 13.14 cusr 26.19 csys = 39.35 CPU)
[12:03:20] t0003-generate-cmdlist.sh .. ok 32 s ( 0.00 usr 0.02 sys + 13.25 cusr 26.10 csys = 39.37 CPU)
I.e. even on *nix running 100 of these in a loop was up to ~2x faster
in absolute runtime, I suspect it's due factors that are exacerbated
in the CI, e.g. much slower process startup due to some platform
limits, or a slower FS.
The "cut -d" change here is because we're not emitting the
40-character aligned output anymore, i.e. we'll get the output from
command_list() now, not an as-is line from command-list.txt.
This also makes the parsing more reliable, as we could tweak the
whitespace alignment without breaking this parser. Let's reword a
now-inaccurate comment in "command-list.txt" describing that previous
alignment limitation. We'll still need the "### command-list [...]"
line due to the "Documentation/cmd-list.perl" logic added in
11c6659d85 (command-list: prepare machinery for upcoming "common
groups" section, 2015-05-21).
There was a proposed change subsequent to this one[3] which continued
moving more logic into the "command_list() function, i.e. replaced the
"cut | tr | grep" chain in "category_list()" with an argument to
"command_list()".
That change might have had a bit of an effect, but not as much as the
preceding commit, so I decided to drop it. The relevant performance
numbers from it were:
linux-gcc:
[12:05:33] t0001-generate-cmdlist.sh .. ok 13 ms ( 0.00 usr 0.00 sys + 3.33 cusr 2.78 csys = 6.11 CPU)
[12:05:33] t0002-generate-cmdlist.sh .. ok 14 ms ( 0.00 usr 0.00 sys + 3.64 cusr 3.09 csys = 6.73 CPU)
osx-gcc:
[11:58:03] t0001-generate-cmdlist.sh .. ok 78416 ms ( 0.02 usr 0.01 sys + 11.78 cusr 6.22 csys = 18.03 CPU)
[11:58:04] t0002-generate-cmdlist.sh .. ok 80081 ms ( 0.02 usr 0.02 sys + 17.80 cusr 10.07 csys = 27.91 CPU)
vs-test:
[12:03:20] t0001-generate-cmdlist.sh .. ok 34 s ( 0.00 usr 0.03 sys + 12.42 cusr 19.55 csys = 32.00 CPU)
[12:03:14] t0002-generate-cmdlist.sh .. ok 30 s ( 0.02 usr 0.00 sys + 13.14 cusr 26.19 csys = 39.35 CPU)
As above HEAD~2 and t0002* are testing the code in this commit (and
the line is the same), but HEAD~1 and t0001* are testing that dropped
change in [3].
1. https://lore.kernel.org/git/cover-v2-00.10-00000000000-20211022T193027Z-avarab@gmail.com/
2. https://lore.kernel.org/git/patch-v2-08.10-83318d6c0da-20211022T193027Z-avarab@gmail.com/
3. https://lore.kernel.org/git/patch-v2-10.10-e10a43756d1-20211022T193027Z-avarab@gmail.com/
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Replace the "sed" invocation in get_synopsis() with a pure-shell
version. This speeds up generate-cmdlist.sh significantly. Compared to
HEAD~ (old) and "master" we are, according to hyperfine(1):
'sh generate-cmdlist.sh command-list.txt' ran
12.69 ± 5.01 times faster than 'sh generate-cmdlist.sh.old command-list.txt'
18.34 ± 3.03 times faster than 'sh generate-cmdlist.sh.master command-list.txt'
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a preceding commit we changed the print_command_list() loop to use
printf's auto-repeat feature. Let's now get rid of get_category_line()
entirely by not sorting the categories.
This will change the output of the generated code from e.g.:
- { "git-apply", N_("Apply a patch to files and/or to the index"), 0 | CAT_complete | CAT_plumbingmanipulators },
To:
+ { "git-apply", N_("Apply a patch to files and/or to the index"), 0 | CAT_plumbingmanipulators | CAT_complete },
I.e. the categories are no longer sorted, but as they're OR'd together
it won't matter for the end result.
This speeds up the generate-cmdlist.sh a bit. Comparing HEAD~ (old)
and "master" to this code:
'sh generate-cmdlist.sh command-list.txt' ran
1.07 ± 0.33 times faster than 'sh generate-cmdlist.sh.old command-list.txt'
1.15 ± 0.36 times faster than 'sh generate-cmdlist.sh.master command-list.txt'
Suggested-by: Jeff King <peff@peff.net>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This is just a small code reduction. There is a small probability that
the new code breaks when the category list is empty. But that would be
noticed during the compile step.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This doesn't matter for performance, but let's not include the empty
lines in our sorting. This makes the intent of the code clearer.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This isn't for optimization as the get_categories() is a purely shell
function, but rather for ease of readability, let's just inline these
two lines. We'll be changing this code some more in subsequent commits
to make this worth it.
Rename the get_categories() function to get_category_line(), since
that's what it's doing now.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function get_categories() is invoked in a loop over all commands.
As it runs several processes, this takes an awful lot of time on
Windows. To reduce the number of processes, move the process that
filters empty lines to the other invoker of the function, where it is
needed. The invocation of get_categories() in the loop does not need
the empty line filtered away because the result is word-split by the
shell, which eliminates the empty line automatically.
Furthermore, use sort -u instead of sort | uniq to remove yet another
process.
[Ævar: on Linux this seems to speed things up a bit, although with
hyperfine(1) the results are fuzzy enough to land within the
confidence interval]:
$ git show HEAD~:generate-cmdlist.sh >generate-cmdlist.sh.old
$ hyperfine --warmup 1 -L s ,.old -p 'make clean' 'sh generate-cmdlist.sh{s} command-list.txt'
Benchmark #1: sh generate-cmdlist.sh command-list.txt
Time (mean ± σ): 371.3 ms ± 64.2 ms [User: 430.4 ms, System: 72.5 ms]
Range (min … max): 320.5 ms … 517.7 ms 10 runs
Benchmark #2: sh generate-cmdlist.sh.old command-list.txt
Time (mean ± σ): 489.9 ms ± 185.4 ms [User: 724.7 ms, System: 141.3 ms]
Range (min … max): 346.0 ms … 885.3 ms 10 runs
Summary
'sh generate-cmdlist.sh command-list.txt' ran
1.32 ± 0.55 times faster than 'sh generate-cmdlist.sh.old command-list.txt'
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add " " before a "|" at the end of a line in generate-cmdlist.sh for
consistency with other code in the file. Some of the surrounding code
will be modified in subsequent commits.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We should keep these files sorted in the C locale, e.g. in the C
locale the order is:
git-check-mailmap
git-check-ref-format
git-checkout
But under en_US.UTF-8 it's:
git-check-mailmap
git-checkout
git-check-ref-format
In a subsequent commit I'll change generate-cmdlist.sh to use C sort
order, and without this change we'd be led to believe that that change
caused a meaningful change in the output, so let's do this as a
separate step, right now the generate-cmdlist.sh script just uses the
order found in this file.
Note that this refers to the sort order of the lines in
command-list.txt, a subsequent commit will also change how we treat
the sort order of the "category" fields, but that's unrelated to this
change.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If prepare_bitmap_git() returns NULL (one easy-to-trigger cause being
that the repository does not have bitmaps at all), then we'll segfault
accessing bitmap_git->hashes:
$ t/helper/test-tool bitmap dump-hashes
Segmentation fault
We should treat this the same as a repository with bitmaps but no
name-hashes, and quietly produce an empty output. The later call to
free_bitmap_index() in the cleanup label is OK, as it treats a NULL
pointer as a noop.
This isn't a big deal in practice, as this function is intended for and
used only by test-tool. It's probably worth fixing to avoid confusion,
but not worth adding coverage for this to the test suite.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The strftime() function has a non-standard "%s" extension, which prints
the number of seconds since the epoch. But the "struct tm" we get has
already been adjusted for a particular time zone; going back to an epoch
time requires knowing that zone offset. Since strftime() doesn't take
such an argument, round-tripping to a "struct tm" and back to the "%s"
format may produce the wrong value (off by tz_offset seconds).
Since we're already passing in the zone offset courtesy of c3fbf81a85
(strbuf: let strbuf_addftime handle %z and %Z itself, 2017-06-15), we
can use that same value to adjust our epoch seconds accordingly.
Note that the description above makes it sound like strftime()'s "%s" is
useless (and really, the issue is shared by mktime(), which is what
strftime() would use under the hood). But it gets the two cases for
which it's designed correct:
- the result of gmtime() will have a zero offset, so no adjustment is
necessary
- the result of localtime() will be offset by the local zone offset,
and mktime() and strftime() are defined to assume this offset when
converting back (there's actually some magic here; some
implementations record this in the "struct tm", but we can't
portably access or manipulate it. But they somehow "know" whether a
"struct tm" is from gmtime() or localtime()).
This latter point means that "format-local:%s" actually works correctly
already, because in that case we rely on the system routines due to
6eced3ec5e (date: use localtime() for "-local" time formats,
2017-06-15). Our problem comes when trying to show times in the author's
zone, as the system routines provide no mechanism for converting in
non-local zones. So in those cases we have a "struct tm" that came from
gmtime(), but has been manipulated according to our offset.
The tests cover the broken round-trip by formatting "%s" for a time in a
non-system timezone. We use the made-up "+1234" here, which has two
advantages. One, we know it won't ever be the real system zone (and so
we're actually testing a case that would break). And two, since it has a
minute component, we're testing the full decoding of the +HHMM zone into
a number of seconds. Likewise, we test the "-1234" variant to make sure
there aren't any sign mistakes.
There's one final test, which covers "format-local:%s". As noted, this
already passes, but it's important to check that we didn't regress this
case. In particular, the caller in show_date() is relying on localtime()
to have done the zone adjustment, independent of any tz_offset we
compute ourselves. These should match up, since our local_tzoffset() is
likewise built around localtime(). But it would be easy for a caller to
forget to pass in a correct tz_offset to strbuf_addftime(). Fortunately
show_date() does this correctly (it has to because of the existing
handling of %z), and the test continues to pass. So this one is just
future-proofing against a change in our assumptions.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
One CI task based on Fedora image noticed a not-quite-kosher
consturct recently, which has been corrected.
* vd/pthread-setspecific-g11-fix:
async_die_is_recursing: work around GCC v11.x issue on Fedora
