Hi,

On Sun, Jun 26, 2011 at 10:20 PM, Kostya <[email protected]> wrote:
> On Sun, Jun 26, 2011 at 03:52:01PM -0700, Ronald S. Bultje wrote:
>> This functionality is only implemented for RGB24/32 and causes crashes
>> otherwise.
>> ---
>>  libswscale/utils.c |   13 +++++++++++++
>>  1 files changed, 13 insertions(+), 0 deletions(-)
>>
>> diff --git a/libswscale/utils.c b/libswscale/utils.c
>> index 6971418..ab28588 100644
>> --- a/libswscale/utils.c
>> +++ b/libswscale/utils.c
>> @@ -845,6 +845,19 @@ int sws_init_context(SwsContext *c, SwsFilter 
>> *srcFilter, SwsFilter *dstFilter)
>>      getSubSampleFactors(&c->chrDstHSubSample, &c->chrDstVSubSample, 
>> dstFormat);
>>
>>      // reuse chroma for 2 pixels RGB/BGR unless user wants full chroma 
>> interpolation
>> +    if (flags & SWS_FULL_CHR_H_INT &&
>> +        dstFormat != PIX_FMT_RGBA &&
>> +        dstFormat != PIX_FMT_ARGB &&
>> +        dstFormat != PIX_FMT_BGRA &&
>> +        dstFormat != PIX_FMT_ABGR &&
>> +        dstFormat != PIX_FMT_RGB24 &&
>> +        dstFormat != PIX_FMT_BGR24) {
>> +        av_log(c, AV_LOG_ERROR,
>> +               "full chroma interpolation for destination format '%s' not 
>> yet implemented\n",
>> +               sws_format_name(dstFormat));
>> +        flags &= ~SWS_FULL_CHR_H_INT;
>> +        c->flags = flags;
>> +    }
>>      if (isAnyRGB(dstFormat) && !(flags&SWS_FULL_CHR_H_INT)) 
>> c->chrDstHSubSample=1;
>>
>>      // drop some chroma lines if the user wants it
>> --
>
> Why not use condition like below? I'd like my RGB565 output with full chroma
> too and mentioning all dst formats in wall of conditions is ugly IMO.

Scaling functions use c->flags to decide whether to use a full or half
function. I.e., the value needs to match the function chosen, else
everything goes wrong...

And yes this is a design problem, I'll see what I can do about that,
but at least it's better to not just plain crash. :-).

Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to