[FFmpeg-devel] [PATCH] CDXL demuxer and decoder

[Reimar Döffinger]

On 7 Jan 2012, at 20:52, Paul B Mahol <onemda at gmail.com> wrote:
> On 12/30/11, Reimar Doeffinger <Reimar.Doeffinger at gmx.de> wrote:
>> only partial review...
>>> +            AV_WL32(out + (x * 4), (0xFF << 24) | (r | g | b));
>> 
>> Chose a different pixfmt and use AV_WN32 (or better the variant for aligned
>> writes).
> 
> I'm little lost here, wouldn't AV_WN32() be enough?

Well WN will produce a different memory content on big-endian than WL.
So either the patch you originally submitted produced wrong results on big-endian or it will after changing it to WN.
I assumed the later, in which case that can be fixed by choosing a different PIX_FMT.
I don't know which, but there is for example both RGBA and RGB32 (I think RGB32 is it).