On Tue, Feb 3, 2015 at 7:38 AM, H.J. Lu <hjl.to...@gmail.com> wrote:
> On Tue, Feb 3, 2015 at 7:00 AM, Siddhesh Poyarekar <siddh...@redhat.com>
> wrote:
>> ... and I forgot to add bug-gnulib to cc before I hit send.
>>
>> Siddhesh
>>
>> On Tue, Feb 03, 2015 at 08:26:49PM +0530, Siddhesh Poyarekar wrote:
>>> Hi,
>>>
>>> obstack_init does not completely initialize the obstack structure; it
>>> leaves out the padding bits and valgrind complains about it on s390x:
>>>
>>> ==15793== Memcheck, a memory error detector
>>> ==15793== Copyright (C) 2002-2012, and GNU GPL'd, by Julian Seward et al.
>>> ==15793== Using Valgrind-3.8.1 and LibVEX; rerun with -h for copyright info
>>> ==15793== Command: /root/obstack
>>> ==15793==
>>> ==15793== Conditional jump or move depends on uninitialised value(s)
>>> ==15793== at 0x403E48CA4E: obstack_free (in /lib64/libc-2.12.so)
>>> ==15793== by 0x8000072D: main (obstack.c:12)
>>> ==15793==
>>> ==15793==
>>> ==15793== HEAP SUMMARY:
>>> ==15793== in use at exit: 0 bytes in 0 blocks
>>> ==15793== total heap usage: 1 allocs, 1 frees, 4,064 bytes allocated
>>> ==15793==
>>> ==15793== All heap blocks were freed -- no leaks are possible
>>> ==15793==
>>> ==15793== For counts of detected and suppressed errors, rerun with: -v
>>> ==15793== Use --track-origins=yes to see where uninitialised values come
>>> from
>>> ==15793== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 4 from 4)
>>>
>>>
>>> The fix below (against gnulib, but is identical for glibc) initializes
>>> all of the obstack struct at once. Verified that the valgrind warning
>>> is fixed. OK for 2.22 and gnulib?
>>>
>>> Siddhesh
>>>
>>> ChangeLog for gnulib:
>>>
>>> obstack: Initialize whole obstack structure.
>>> * lib/obstack.c (_obstack_begin): Initialize all of H.
>>>
>>> ChangeLog for glibc:
>>>
>>> [BZ #17919]
>>> * malloc/obstack.c (_obstack_begin): Initialize all of H.
>>>
>>> diff --git a/malloc/obstack.c b/malloc/obstack.c
>>> index 5bb3f0d..c1d6ded 100644
>>> --- a/lib/obstack.c
>>> +++ b/lib/obstack.c
>>> @@ -148,6 +148,8 @@ _obstack_begin (struct obstack *h,
>>> {
>>> struct _obstack_chunk *chunk; /* points to new chunk */
>>>
>>> + memset (h, 0, sizeof (struct obstack));
>>> +
>>> if (alignment == 0)
>>> alignment = DEFAULT_ALIGNMENT;
>>> if (size == 0)
>>
>>
>
> I think you should also remove
>
> h->use_extra_arg = 0;
>
And
/* The initial chunk now contains no empty object. */
h->maybe_empty_object = 0;
h->alloc_failed = 0;
--
H.J.
