[FFmpeg-devel] [inline assembly compliance] Issues and patches

FRÉDÉRIC RECOULES frederic.recoules at univ-grenoble-alpes.fr
Wed Apr 22 18:12:10 EEST 2020


Hi Michael, 

We are looking forward to hear news about the patches ([PATCH /5] x86 inline assembly compliance). 
Did you have time to take a look? 

We are trying to keep track of the state of the assembly in projects in which we found issues. 

Please recall that they should be now small, very local, easy to review changes and that they pass the FATE tests. 
And please contact us if you need more information about them or about the other issues we are willing to patch in the future. 

Regards, 
Frédéric Recoules 


De: "FRÉDÉRIC RECOULES" <frederic.recoules at univ-grenoble-alpes.fr> 
À: "ffmpeg-devel" <ffmpeg-devel at ffmpeg.org> 
Envoyé: Samedi 4 Avril 2020 19:55:13 
Objet: Re: [FFmpeg-devel] [inline assembly compliance] Issues and patches 

> i can retest and find a mininal command line for configure which triggers this 
> in case you cannot reproduce it ? 

Thank you, but it is a known (non) issue of Clang and is reproducible since at least 3.4.1. 
The fact is that GCC uses incomplete types in to say "block of unknown size" but Clang have chosen to not follow it on this. 

Also, I have resubmitted the patches and I hope they will be more readable and useful. 

Regards, 
Frédéric Recoules 


De: "Michael Niedermayer" <michael at niedermayer.cc> 
À: "ffmpeg-devel" <ffmpeg-devel at ffmpeg.org> 
Envoyé: Samedi 4 Avril 2020 18:55:27 
Objet: Re: [FFmpeg-devel] [inline assembly compliance] Issues and patches 

On Sat, Apr 04, 2020 at 12:28:43PM +0200, FRÉDÉRIC RECOULES wrote: 
> Thank you for your answers. 
> 
> As you have pointed out, these patches are full of unrelated changes that are not important for safety. 
> Most of them were never intended to be posted here, the diff we submitted was the one of an 
> experimental branch and we apologize to have made such a mistake. 
> We will resubmit the patch with only essential patches in a more appropriate format very soon 
> (git format-patch). 
> 
> @Michael These errors come with the Clang compiler, aren't they? 

yes, seems this was 
clang version 4.0.0 (trunk 283753) 

i can retest and find a mininal command line for configure which triggers this 
in case you cannot reproduce it ? 


> We are aware that support for inline assembly may differ from one compiler to another 
> despite the "higly-compatibility" that is stated. The safety patches we are proposing 
> do not rely on them. 
> 
> @Carl @Kieran So far, we passed the FATE tests. 
> The output is slightly different because we have merged contiguous assembly statement 
> such that the compiler can no longer insert instruction between them, but the differences 
> are only instruction swaps A.B.C -> B.A.C. 

[...] 
-- 
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB 

There will always be a question for which you do not know the correct answer. 

_______________________________________________ 
ffmpeg-devel mailing list 
ffmpeg-devel at ffmpeg.org 
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel 

To unsubscribe, visit link above, or email 
ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe". 
_______________________________________________ 
ffmpeg-devel mailing list 
ffmpeg-devel at ffmpeg.org 
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel 

To unsubscribe, visit link above, or email 
ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe". 


More information about the ffmpeg-devel mailing list