Re: [PATCH] avoid -Wuse-after-free [BZ #26779]

[Jeff Law <jeffreyalaw@gmail.com>]

On 1/24/2022 5:42 PM, Martin Sebor via Libc-alpha wrote:
> On 1/21/22 17:42, Paul Eggert wrote:
>> On 1/21/22 15:14, Martin Sebor wrote:
>>>>
>>>
>>> That does look cleaner although it wasn't entirely obvious to me
>>> from looking at the code that it's the same.  Unfortunately, it
>>> doesn't help.  GCC replaces the bool variable with the equality
>>> test of the two pointers, and the warning points that out.
>>
>> That's a GCC bug.
>>
>>> Pedantically speaking it's a bug in GCC that it does that but
>>> I doubt anyone would care to do anything about it.
>>
>> You might be surprised; they do fix bugs in this area. If I could 
>> reproduce the bug I'd file a bug report, but my GCC doesn't complain 
>> so I guess it's up to you....
>
> I opened a GCC bug for the record:
>   https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104215
> GCC 12 is now in its regression-fixing stage so to get it fixed we'd
> need to make it a regression.  That's usually only done for bugs that
> trigger under conditions where they didn't before.
The problem is there's no way for the compiler to know that it can't 
move the use of q past the realloc call -- there's nothing from a 
dataflow standpoint which would prevent such movement.

In some ways the builtin_realloc call would need to say "I'm going to 
clobber this argument" and it would have to do so before translating 
into SSA form I think.

Jeff