Re: [PATCH 02/13] commit-graph: stop depending on `struct odb_source`

> On Mon, Oct 05, 2026 at 03:43:44PM -0400, Karthik Nayak wrote:
>> Patrick Steinhardt <ps@pks.im> writes:
>>
>> [snip]
>
> Thanks for trimming! One more ask though: it's helpful to retain the
> diff header itself so that one knows which file this is that you are
> commenting on :)
>

>> > @@ -28,7 +29,7 @@
>> >  #include "tree.h"
>> >  #include "chunk-format.h"
>> >
>> > -void git_test_write_commit_graph_or_die(struct odb_source *source)
>> > +void git_test_write_commit_graph_or_die(struct repository *repo)
>> >  {
>> >  	int flags = 0;
>> >  	if (!git_env_bool(GIT_TEST_COMMIT_GRAPH, 0))
>> > @@ -37,7 +38,7 @@ void git_test_write_commit_graph_or_die(struct odb_source *source)
>> >  	if (git_env_bool(GIT_TEST_COMMIT_GRAPH_CHANGED_PATHS, 0))
>> >  		flags = COMMIT_GRAPH_WRITE_BLOOM_FILTERS;
>> >
>> > -	if (write_commit_graph_reachable(source, flags, NULL))
>> > +	if (write_commit_graph_reachable(repo, repo->objects->sources->path, flags, NULL))
>> >  		die("failed to write commit-graph under GIT_TEST_COMMIT_GRAPH");
>> >  }
>> >
>>
>> Shouldn't the caller of `git_test_write_commit_graph_or_die()` send in
>> (repo, path) and we forward that path, instead of using the path from
>> `repo->objects->sources->path`?
>
> I'd agree if this were a properly designed function. But it's basically
> just a hack for our test suite, so I was aiming for the easiest fix
> possible to make this work.
>
> Patrick