Hello,
1. I agree that av_uninit should be removed; in fact, I also wanted to
do so and already made some commits to do so where I consider the commit
to be beneficial on its own (even without the av_uninit removal) like
348461e550f88c5e27afeafc85e7ace11fe8fae4 and its parent.
2. You fail to actu
Zhao Zhili (HE12025-04-11):
> This is only check on particular compiler implementation. It works until
> someone
> use a different compiler and UB kicks in.
No, with this suggestion there is no UB.
--
Nicolas George
___
ffmpeg-devel mailing list
ffm
> On Apr 11, 2025, at 17:32, Nicolas George wrote:
>
> Zhao Zhili (HE12025-04-11):
>> With UB, the compiler can remove branch check and assign some random
>> value to it, which cannot be detected by valgrind.
>>
>> For ab792634197e, the UB is there for decades and never detected by
>> valgrind
Zhao Zhili (HE12025-04-11):
> With UB, the compiler can remove branch check and assign some random
> value to it, which cannot be detected by valgrind.
>
> For ab792634197e, the UB is there for decades and never detected by
> valgrind, and the warning is silenced by av_uninit.
You make a valid po
> On Apr 11, 2025, at 16:36, Nicolas George wrote:
>
> Zhao Zhili (HE12025-04-11):
>> From: Zhao Zhili
>>
>> The macro is meant to suppress false uninitialized warnings. However,
>> sometimes these 'false uninitialized warnings' are really undefined
>> behavior, and leading to real issue like
> On Apr 11, 2025, at 16:36, Nicolas George wrote:
>
> Zhao Zhili (HE12025-04-11):
>> From: Zhao Zhili
>>
>> The macro is meant to suppress false uninitialized warnings. However,
>> sometimes these 'false uninitialized warnings' are really undefined
>> behavior, and leading to real issue lik
Zhao Zhili (HE12025-04-11):
> From: Zhao Zhili
>
> The macro is meant to suppress false uninitialized warnings. However,
> sometimes these 'false uninitialized warnings' are really undefined
> behavior, and leading to real issue like crash, e.g., ab792634197e.
>
> For false uninitialized warning
