On Wed, Jul 10, 2024 at 11:30:13AM +0900, Rubén Justo wrote:
> As we currently describe in t/README, it can happen that:
>
> Some tests run "git" (or "test-tool" etc.) without properly checking
> the exit code, or git will invoke itself and fail to ferry the
> abort() exit code to the original caller.
>
> Therefore, GIT_TEST_SANITIZE_LEAK_LOG=true is needed to be set to
> capture all memory leaks triggered by our tests.
>
> It seems unnecessary to force users to remember this option, as
> forgetting it could lead to missed memory leaks.
>
> We could solve the problem by setting GIT_TEST_SANITIZE_LEAK_LOG to
> "true" by default, but that might suggest we think "false" makes sense,
> which isn't the case.
>
> Therefore, the best approach is to remove the option entirely while
> maintaining the capability to detect memory leaks in blind spots of our
> tests.
Yeah, I think that reasoning makes sense.
> diff --git a/ci/lib.sh b/ci/lib.sh
> index 814578ffc6..51f8f59a29 100755
> --- a/ci/lib.sh
> +++ b/ci/lib.sh
> @@ -370,7 +370,6 @@ linux-musl)
> linux-leaks|linux-reftable-leaks)
> export SANITIZE=leak
> export GIT_TEST_PASSING_SANITIZE_LEAK=true
> - export GIT_TEST_SANITIZE_LEAK_LOG=true
> ;;
OK, we can drop this line snice it's now the default. Good.
> diff --git a/t/README b/t/README
> index d9e0e07506..c2a732d59e 100644
> --- a/t/README
> +++ b/t/README
> @@ -382,33 +382,9 @@ mapping between "TEST_PASSES_SANITIZE_LEAK=true" and
> those tests that
> pass under "SANITIZE=leak". This is especially useful when testing a
> series that fixes various memory leaks with "git rebase -x".
>
> -GIT_TEST_SANITIZE_LEAK_LOG=true will log memory leaks to
> -"test-results/$TEST_NAME.leak/trace.*" files. The logs include a
> -"dedup_token" (see +"ASAN_OPTIONS=help=1 ./git") and other options to
> -make logs +machine-readable.
> -
> -With GIT_TEST_SANITIZE_LEAK_LOG=true we'll look at the leak logs
> -before exiting and exit on failure if the logs showed that we had a
> -memory leak, even if the test itself would have otherwise passed. This
> -allows us to catch e.g. missing &&-chaining. This is especially useful
> -when combined with "GIT_TEST_PASSING_SANITIZE_LEAK", see below.
> -
> GIT_TEST_PASSING_SANITIZE_LEAK=check when combined with "--immediate"
> will run to completion faster, and result in the same failing
> tests. The only practical reason to run
> -GIT_TEST_PASSING_SANITIZE_LEAK=check without "--immediate" is to
> -combine it with "GIT_TEST_SANITIZE_LEAK_LOG=true". If we stop at the
> -first failing test case our leak logs won't show subsequent leaks we
> -might have run into.
> -
> -GIT_TEST_PASSING_SANITIZE_LEAK=(true|check) will not catch all memory
> -leaks unless combined with GIT_TEST_SANITIZE_LEAK_LOG=true. Some tests
> -run "git" (or "test-tool" etc.) without properly checking the exit
> -code, or git will invoke itself and fail to ferry the abort() exit
> -code to the original caller. When the two modes are combined we'll
> -look at the "test-results/$TEST_NAME.leak/trace.*" files at the end of
> -the test run to see if had memory leaks which the test itself didn't
> -catch.
After this patch, the documentation seems to end abruptly with "The only
practical reason to run". I think we need to either delete those lines,
too, or complete the thought.
I do think they are saying something useful, which is: in "check" mode,
you should always use "--immediate" since the point is just to find
scripts which aren't labeled correctly. But I think that is true whether
you are using the leak log or not. Your log will be incomplete, of
course, if you used "--immediate", but the point is to see whether we
find even one.
> GIT_TEST_PROTOCOL_VERSION=<n>, when set, makes 'protocol.version'
> default to n.
> diff --git a/t/test-lib.sh b/t/test-lib.sh
> index 79d3e0e7d9..942828c55d 100644
> --- a/t/test-lib.sh
> +++ b/t/test-lib.sh
> @@ -1270,8 +1270,8 @@ check_test_results_san_file_ () {
> say "As TEST_PASSES_SANITIZE_LEAK=true isn't set the above leak is 'ok'
> with GIT_TEST_PASSING_SANITIZE_LEAK=check" &&
> invert_exit_code=t
> else
> - say "With GIT_TEST_SANITIZE_LEAK_LOG=true our logs revealed a memory
> leak, exit non-zero!" &&
> - invert_exit_code=t
> + say "Our logs revealed a leak!" &&
> + test "$test_failure" != 0 || invert_exit_code=t
> fi
> }
This adds back in the test_failure fix from 47c6d4dad2 (test-lib: fix
GIT_TEST_SANITIZE_LEAK_LOG, 2024-06-30), but in a different way. I think
we'd want to build on top, and then you just need to update the messages
on either side of that final elif/else.
> @@ -1555,28 +1555,28 @@ then
> passes_sanitize_leak=t
> fi
>
> - if test "$GIT_TEST_PASSING_SANITIZE_LEAK" = "check"
> + if test -z "$passes_sanitize_leak" &&
> + ! test "$GIT_TEST_PASSING_SANITIZE_LEAK" = "check" &&
> + test_bool_env GIT_TEST_PASSING_SANITIZE_LEAK false
> then
> - sanitize_leak_check=t
> - if test -n "$invert_exit_code"
> + skip_all="skipping $this_test under GIT_TEST_PASSING_SANITIZE_LEAK=true"
> + test_done
> + else
> + if test "$GIT_TEST_PASSING_SANITIZE_LEAK" = "check"
> then
> [...]
I'm not sure why we need to touch this block. The "if
GIT_TEST_SANITIZE_LEAK_LOG" just below it I assumed would go away. But
all of this has to do with "check" versus "true", etc? There might be
new refactoring / simplification opportunities opened up by getting rid
of the LEAK_LOG variable, but we should do those on top.
I guess what's happening is that you've rearranged it so that:
> - if test_bool_env GIT_TEST_SANITIZE_LEAK_LOG false
> - then
> if ! mkdir -p "$TEST_RESULTS_SAN_DIR"
> then
> BAIL_OUT "cannot create $TEST_RESULTS_SAN_DIR"
...when this conditional goes away, the existing body is still in the
"else". But even though it would make the diff noisy to reindent, I
think we are better off doing so to make it clear what the actual change
is.
> @@ -1599,9 +1599,6 @@ elif test "$GIT_TEST_PASSING_SANITIZE_LEAK" = "check"
> ||
> test_bool_env GIT_TEST_PASSING_SANITIZE_LEAK false
> then
> BAIL_OUT_ENV_NEEDS_SANITIZE_LEAK "GIT_TEST_PASSING_SANITIZE_LEAK=true"
> -elif test_bool_env GIT_TEST_SANITIZE_LEAK_LOG false
> -then
> - BAIL_OUT_ENV_NEEDS_SANITIZE_LEAK "GIT_TEST_SANITIZE_LEAK_LOG=true"
> fi
OK, this final elif is responsible for complaining when you set LEAK_LOG
but don't have an actual leak-checking build. But once it goes away,
there's no need to complain. Makes sense.
-Peff
