On Fri, Sep 29, 2017 at 09:18:11AM +0200, Christian Couder wrote:
> On Fri, Sep 29, 2017 at 12:42 AM, Jeff King <peff@xxxxxxxx> wrote:
> > On Thu, Sep 28, 2017 at 08:38:39AM +0000, Olga Telezhnaya wrote:
> >
> >> diff --git a/packfile.c b/packfile.c
> >> index f69a5c8d607af..ae3b0b2e9c09a 100644
> >> --- a/packfile.c
> >> +++ b/packfile.c
> >> @@ -876,6 +876,7 @@ void prepare_packed_git(void)
> >> for (alt = alt_odb_list; alt; alt = alt->next)
> >> prepare_packed_git_one(alt->path, 0);
> >> rearrange_packed_git();
> >> + INIT_LIST_HEAD(&packed_git_mru.list);
> >> prepare_packed_git_mru();
> >> prepare_packed_git_run_once = 1;
> >> }
> >
> > I was thinking on this hunk a bit more, and I think it's not quite
> > right.
> >
> > The prepare_packed_git_mru() function will clear the mru list and then
> > re-add each item from the packed_git list. But by calling
> > INIT_LIST_HEAD() here, we're effectively clearing the packed_git_mru
> > list, and we end up leaking whatever was on the list before.
>
> The current code is:
>
> static int prepare_packed_git_run_once = 0;
> void prepare_packed_git(void)
> {
> struct alternate_object_database *alt;
>
> if (prepare_packed_git_run_once)
> return;
> prepare_packed_git_one(get_object_directory(), 1);
> prepare_alt_odb();
> for (alt = alt_odb_list; alt; alt = alt->next)
> prepare_packed_git_one(alt->path, 0);
> rearrange_packed_git();
> prepare_packed_git_mru();
> prepare_packed_git_run_once = 1;
> }
>
> As we use the "prepare_packed_git_run_once" static, this function will
> only be called only once when packed_git_mru has not yet been
> initialized, so there will be no leak.
Check reprepare_packed_git(). It unsets the run_once flag, and then
calls prepare_packed_git() again.
-Peff
