Silence -O3 -Wall warning in malloc/hooks.c with GCC 7 [BZ #22052]
Commit Message
On Thu, Aug 31, 2017 at 10:01 AM, Florian Weimer <fweimer@redhat.com> wrote:
> On 08/31/2017 05:55 AM, H.J. Lu wrote:
>
>> One of malloc change caused:
>>
>> https://sourceware.org/bugzilla/show_bug.cgi?id=22052
>
> (uninit warning with -O3)
>
> I have not yet found a good way to suppress this. I don't know yet why
> GCC loses the information that the maybe_p variable is initialized if
> mem2chunk_check returns a non-null pointer.
>
This patch works for me. OK for master?
Comments
On Sun, Sep 3, 2017 at 8:48 AM, H.J. Lu <hjl.tools@gmail.com> wrote:
> On Thu, Aug 31, 2017 at 10:01 AM, Florian Weimer <fweimer@redhat.com> wrote:
>> On 08/31/2017 05:55 AM, H.J. Lu wrote:
>>
>>> One of malloc change caused:
>>>
>>> https://sourceware.org/bugzilla/show_bug.cgi?id=22052
>>
>> (uninit warning with -O3)
>>
>> I have not yet found a good way to suppress this. I don't know yet why
>> GCC loses the information that the maybe_p variable is initialized if
>> mem2chunk_check returns a non-null pointer.
>>
>
> This patch works for me. OK for master?
>
FYI, I opened a GCC bug:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82090
On 09/03/2017 05:48 PM, H.J. Lu wrote:
> + unsigned char *magic_p = NULL;
I think the current practice is to use <libc-diag.h> to suppress the
warning.
Is the issue that the memory clobber on the inline asm makes GCC not to
see the conditional initialization?
Thanks,
Florian
On Mon, Sep 4, 2017 at 1:33 AM, Florian Weimer <fweimer@redhat.com> wrote:
> On 09/03/2017 05:48 PM, H.J. Lu wrote:
>> + unsigned char *magic_p = NULL;
>
> I think the current practice is to use <libc-diag.h> to suppress the
> warning.
Is there a usage example for this warning?
> Is the issue that the memory clobber on the inline asm makes GCC not to
> see the conditional initialization?
Possible.
On 09/04/2017 10:29 PM, H.J. Lu wrote:
> On Mon, Sep 4, 2017 at 1:33 AM, Florian Weimer <fweimer@redhat.com> wrote:
>> On 09/03/2017 05:48 PM, H.J. Lu wrote:
>>> + unsigned char *magic_p = NULL;
>>
>> I think the current practice is to use <libc-diag.h> to suppress the
>> warning.
>
> Is there a usage example for this warning?
Search for DIAG_PUSH_NEEDS_COMMENT in the sources.
Thanks,
Florian
From 735efe5ebc3eb6aa877b00b10ac0046524cee6e3 Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.tools@gmail.com>
Date: Sun, 3 Sep 2017 08:39:55 -0700
Subject: [PATCH] Silence -O3 -Wall warning in malloc/hooks.c with GCC 7 [BZ
#22052]
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
realloc_check has
unsigned char *magic_p;
...
__libc_lock_lock (main_arena.mutex);
const mchunkptr oldp = mem2chunk_check (oldmem, &magic_p);
__libc_lock_unlock (main_arena.mutex);
if (!oldp)
malloc_printerr ("realloc(): invalid pointer");
...
if (newmem == NULL)
*magic_p ^= 0xFF;
with
static void malloc_printerr(const char *str) __attribute__ ((noreturn));
GCC 7 -O3 warns
hooks.c: In function ‘realloc_check’:
hooks.c:352:14: error: ‘magic_p’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
*magic_p ^= 0xFF;
This patch silences GCC 7 by initializing magic_p to NULL.
[BZ #22052]
* malloc/hooks.c (realloc_check): Initialize magic_p to NULL.
---
malloc/hooks.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
@@ -287,7 +287,7 @@ realloc_check (void *oldmem, size_t bytes, const void *caller)
{
INTERNAL_SIZE_T nb;
void *newmem = 0;
- unsigned char *magic_p;
+ unsigned char *magic_p = NULL;
if (bytes + 1 == 0)
{
--
2.13.5