On Wed, Nov 06, 2024 at 04:10:14PM +0100, Patrick Steinhardt wrote:
> Hi,
>
> this is the last part of my series of memory leak fixes. This series
> goes a bit further than past series:
>
> - Patches 1 to 16 plug remaining memory leaks exposed by our test
> suite.
>
> - Patches 17 to 21 remove the last remaining `UNLEAK()` annotations
> and ultimately remove the macro itself.
>
> - Patch 22 works around a bug in the leak sanitizer itself.
>
> - Patches 23 and 24 drop annotations where leak-free tests pass with
> the leak sanitizer.
>
> - Patches 25 and 26 unconditionally enable leak checking in all newly
> added tests and then drop the `TEST_PASSES_SANITIZE_LEAK`
> annotation.
Hi Patrick.
Thanks for working on this.
It's been a while since 956d2e4639 (tests: add a test mode for SANITIZE=leak,
run it in CI, 2021-09-23). This series marks certainly another milestone.
I have left a few comments, none of them important. The series looks
good to me. Just in case you re-roll, while reading the series I
spotted some nits:
diff --git a/bisect.c b/bisect.c
index 518be70aa3..0e804086cb 100644
--- a/bisect.c
+++ b/bisect.c
@@ -641,7 +641,7 @@ static struct commit_list *skip_away(struct commit_list *list, int count)
else if (previous)
result = previous;
else
- result = list;
+ result = list;
break;
}
previous = cur;
diff --git a/builtin/init-db.c b/builtin/init-db.c
index 9e069e2742..096f96b9c4 100644
--- a/builtin/init-db.c
+++ b/builtin/init-db.c
@@ -248,9 +248,9 @@ int cmd_init_db(int argc,
}
flags |= INIT_DB_EXIST_OK;
- ret = init_db(git_dir, real_git_dir, template_dir, hash_algo,
- ref_storage_format, initial_branch,
- init_shared_repository, flags);
+ ret = init_db(git_dir, real_git_dir, template_dir, hash_algo,
+ ref_storage_format, initial_branch,
+ init_shared_repository, flags);
free(template_dir_to_free);
free(real_git_dir_to_free);
>
> So once this series lands, the expectation is that any newly added test
> needs to be leak free by default. We still have an escape hatch in the
> form of the SANITIZE_LEAK prerequisite, but patch authors are expected
> to provide good arguments why their test cannot be made leak free.
>
> This series is built on top of 8f8d6eee53 (The seventh batch,
> 2024-11-01) with ps/leakfixes-part-9 at c810549be1
> (list-objects-filter-options: work around reported leak on error,
> 2024-11-05) merged into it.
>
> The series has some trivial conflicts with the removed test annotations
> when merging with `seen`. There are two topics in-flight that introduce
> new memory leaks (ds/path-walk and cc/promisor-remote-capability), both
> of which are fixed by the below patch. I've sent a message in response
> to the individual commits that introduce the leaks.
>
> Thanks!
>
> Patrick
