On Tue, May 8, 2018 at 10:04 PM, Jonathan Tan <jonathanta...@google.com> wrote:
> On Tue, 8 May 2018 12:37:36 -0700
> Stefan Beller <sbel...@google.com> wrote:
>
>> +void clear_alloc_state(struct alloc_state *s)
>> +{
>> + while (s->slab_nr > 0) {
>> + s->slab_nr--;
>> + free(s->slabs[s->slab_nr]);
>> + }
>
> I should have caught this earlier, but you need to free s->slabs itself
> too.
And nobody frees 's' either. I'm not saying cler_alloc_state() should,
but somebody should. When I tried repo_clear(the_repository) with
gitster/sb/object-store-alloc I got this
==13250== 944 (32 direct, 912 indirect) bytes in 1 blocks are
definitely lost in loss record 62 of 88
==13250== at 0x4C2CF25: calloc (vg_replace_malloc.c:718)
==13250== by 0x1AB7A8: xcalloc (wrapper.c:160)
==13250== by 0x1BF666: allocate_alloc_state (alloc.c:41)
==13250== by 0x140090: parsed_object_pool_new (object.c:462)
==13250== by 0x16CCF4: initialize_the_repository (repository.c:18)
==13250== by 0x110BD0: main (common-main.c:37)
If you want to reproduce, this is what I used to test this with.
https://gist.github.com/pclouds/86a2df6c28043f1b6fa3d4e72e7a1276
--
Duy
