Re: [PATCH 13/16] packfile: ensure `close_pack_revindex()` frees in-memory revindex
- From
Jeff King <peff@peff.net>
- Date
- Apr 7, 2026, 06:29 UTC
- Message-ID
- <20260407062954.GG714878@coredump.intra.peff.net>
- In-Reply-To
- <e304e316ac4d4e92125e2e57b32fa445acf66e3e.1774820449.git.me@ttaylorr.com>
On Sun, Mar 29, 2026 at 05:41:23PM -0400, Taylor Blau wrote:
Show 11 quoted lines
> The following commit will introduce a case where we write a MIDX bitmap > over packs that do not themselves have on-disk *.rev files. > > This case is supported within Git, and we will simply fall back to > generating the revindex in memory. But we don't ever release that > memory, causing a leak that is exposed by a test introduced in the > following commit. > > (As far as I could find, we never free()'d memory allocated as a > byproduct of creating an in-memory revindex, likely because that code > predates the leak-checking niceties we have in the test suite now.)
Hmmm, I'm surprised we don't trigger this leak during the course of odb_free(). But I guess we do that fairly rarely. It mostly happens via repo_clear(), and we never do that for the "main" repo (just for submodules). And it's unlikely that we'd have loaded a revindex for the submodule for most operations (mostly we need them for generating packs or when using bitmaps).
So I can't see an obvious way that this would be triggered in practice with the current code. But I agree it is a leak waiting to happen, as nobody ever clears p->revindex.
Show 12 quoted lines
> diff --git a/packfile.c b/packfile.c
> index d4de9f3ffe8..97f7662e079 100644
> --- a/packfile.c
> +++ b/packfile.c
> @@ -420,6 +420,8 @@ void close_pack_index(struct packed_git *p)
>
> static void close_pack_revindex(struct packed_git *p)
> {
> + FREE_AND_NULL(p->revindex);
> +
> if (!p->revindex_map)
> return;The fact that this is called "close" made me wonder if some code might want to just relinquish access to on-disk resources, and then pay the cost when we have to unexpectedly re-compute the in-memory revindex.
But this function and its siblings are really about resetting the whole state of the packed_git struct. So the behavior you're proposing matches that.
-Peff