Fix misbehavior in Git.pm that dates back to the very first version of
the library in git.git added in b1edc53d06 (Introduce Git.pm (v4),
2006-06-24). When we fail to execute a command we shouldn't ignore all
signals, those can happen e.g. if abort() is called, or if the command
segfaults.
Because of this we'd consider e.g. a command that died due to LSAN
exiting with abort() successful, as is the case with the tests listed
as running successfully with SANITIZE=leak in 9081a421a6 (checkout:
fix "branch info" memory leaks, 2021-11-16). We did run them
successfully, but only because we ignored these errors.
This was then made worse by the use of "abort_on_error=1" for LSAN
added in 85b81b35ff (test-lib: set LSAN_OPTIONS to abort by default,
2017-09-05). Doing that makes sense, but without providing that option
we'd have a "$? >> 8" of "23" on failure, with abort_on_error=1 we'll
get "0".
All of our tests pass even without the SIGPIPE exception being added
here, but as the code appears to have been trying to ignore it let's
keep ignoring it for now.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "checkout" command is one of the main sources of leaks in the test
suite, let's fix the common ones by not leaking from the "struct
branch_info".
Doing this is rather straightforward, albeit verbose, we need to
xstrdup() constant strings going into the struct, and free() the ones
we clobber as we go along.
This also means that we can delete previous partial leak fixes in this
area, i.e. the "path_to_free" accounting added by 96ec7b1e70 (Convert
resolve_ref+xstrdup to new resolve_refdup function, 2011-12-13).
There was some discussion about whether "we should retain the "const
char *" here and cast at free() time, or have it be a "char *". Since
this is not a public API with any sort of API boundary let's use
"char *", as is already being done for the "refname" member of the
same struct.
The tests to mark as passing were found with:
rm .prove; GIT_SKIP_TESTS=t0027 prove -j8 --state=save t[0-9]*.sh :: --immediate
# apply & compile this change
prove -j8 --state=failed :: --immediate
I.e. the ones that were newly passing when the --state=failed command
was run. I left out "t3040-subprojects-basic.sh" and
"t4131-apply-fake-ancestor.sh" to to optimization-level related
differences similar to the ones noted in[1], except that these would
be something the current 'linux-leaks' job would run into.
1. https://lore.kernel.org/git/cover-v3-0.6-00000000000-20211022T175227Z-avarab@gmail.com/
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These variables were added in 16805d3 (t/t91XX-svn: start
removing use of "git-" from these tests, 2008-09-08) so that
running:
git grep git-
would return fewer hits. At the time, we were transitioning
away from the use of the "dashed" git-foo form.
That transition has been over for years, and grepping for
"git-" in the test suite yields thousands of hits anyway
(all presumably false positives).
With their original purpose gone, these variables serve only
to obfuscate the tests. Let's get rid of them.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Breaks in a test assertion's && chain can potentially hide
failures from earlier commands in the chain.
Commands intended to fail should be marked with !, test_must_fail, or
test_might_fail. The examples in this patch do not require that.
Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fixed all places where it was a straightforward change from cd'ing into a
directory and back via "cd .." to a cd inside a subshell.
Found these places with "git grep -w "cd \.\.".
Signed-off-by: Jens Lehmann <Jens.Lehmann@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
I have tweaked configuration in my ~/.subversion directory, namely I am
running auto-properties and automatically adding '$Id$' expansion to
every file. This choke the last test named 'proplist' from
t9101-git-svn-props.sh, because one more property, svn:keywords is
automatically added.
I had just wrapped svn invocation with the svn_cmd that specifies empty
directory via --config-dir argument. Since the latter is the global
option, it should be recognized by all svn subcommands, so no
regressions will be introduced.
Now svn_cmd is used everywhere, not just in the failed test module: this
should guard us from the future clashes with user-defined configuration
tweaks.
Signed-off-by: Eygene Ryabinkin <rea-git@codelabs.ru>
Acked-by: Eric Wong <normalperson@yhbt.net>
This changes "git-foo" to "git foo" when message strings in tests
name git subcommands.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This replaces 'git-svn' with 'git svn' in the tests.
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Subversion tests use too many "git-foo" form, so I am converting them
in two steps.
This first step replaces literal strings "remotes/git-svn" and "git-svn-id"
by introducing $remotes_git_svn and $git_svn_id constants defined as shell
variables. This will reduce the number of false hits from "git grep".
Signed-off-by: Nanako Shiraishi <nanako3@lavabit.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Suppose someone fetches git-svn-ified commits from another repo and then
attempts to use 'git-svn init --rewrite-root=foo bar'. Using git svn rebase
after that will fail badly:
* For each commit tried by working_head_info, rebuild is called indirectly.
* rebuild will iterate over all commits and skip all of them because the
URL does not match. Because of that no rev_map file is generated at all.
* Thus, rebuild will run once for every commit. This takes ages.
* In the end there still isn't any rev_map file and thus working_head_info
fails.
Addressing this behaviour fixes an apparently not too uncommon problem with
providing git-svn mirrors of Subversion repositories. Some repositories are
accessed using different URLs depending on whether the user has push
privileges or not. In the latter case, an anonymous URL is often used that
differs from the push URL. Providing a mirror that is usable in both cases
becomes a lot more possible with this change.
Signed-off-by: Jan Krüger <jk@jk.gs>
Acked-by: Eric Wong <normalperson@yhbt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
