Re: [PATCH v2 0/4] Fix mem leaks of recent object store conversions.
- From
Stefan Beller <sbeller@google.com>
- Date
- May 11, 2018, 18:59 UTC
- Message-ID
- <CAGZ79kafGk3YPu1yuagHJRZYxpHpRUoCbSH0JCeLrdzz9ehOvQ@mail.gmail.com>
- In-Reply-To
- <20180511083723.GB22086@sigill.intra.peff.net>
On Fri, May 11, 2018 at 1:37 AM, Jeff King <peff@peff.net> wrote:
Show 13 quoted lines
> On Thu, May 10, 2018 at 12:58:45PM -0700, Stefan Beller wrote: > >> This series replaces the two commits that were queued on sb/object-store-replace, >> fixing memory leaks that were recently introduced. >> >> Compared to v1, I merged the two independent series from yesterday, >> rewrote the commit message to clear up Junios confusion and addresses Peffs >> comments for the packfiles as well. > > Mostly. :) > > My one remaining complaint is that the bitmap code may hold on to a > dangling pointer to a packed_git after this series.
Ok, I'll look into that.
Show 8 quoted lines
> > I think that is part of a larger problem, though, which is that the > bitmap code's globals need to be part of the struct raw_object_store. > I think this can already cause problems before your series if we were to > try to use bitmaps in both a superproject and a submodule in the same > process, though I think we'd at least hit the "ignoring extra bitmap > file" code path in open_pack_bitmap_1(). So right now it's an annoyance, > but after your series it becomes a potential segfault.
Ok, maybe we'll need to convert bitmaps into the object store for that.
Thanks for the pointer, Stefan