[FFmpeg-devel] FFmpeg source code is no longer C99 because of GNUism called case ranges

[Benoit Fouet]

M?ns Rullg?rd wrote:
> Benoit Fouet <benoit.fouet at purplelabs.com> writes:
>
>   
>> Roman Shaposhnik wrote:
>>     
>>> On Thu, 2007-07-05 at 08:02 +0100, M?ns Rullg?rd wrote:
>>>   
>>>       
>>>>>  static int mpeg4video_probe(AVProbeData *probe_packet)
>>>>>  {
>>>>>      uint32_t temp_buffer= -1;
>>>>> +    uint8_t b;
>>>>>      int VO=0, VOL=0, VOP = 0, VISO = 0, res=0;
>>>>>      int i;
>>>>>  
>>>>>      for(i=0; i<probe_packet->buf_size; i++){
>>>>>          temp_buffer = (temp_buffer<<8) + probe_packet->buf[i];
>>>>> -        if ((temp_buffer & 0xffffff00) == 0x100) {
>>>>> -            switch(temp_buffer){
>>>>> -            case VOP_START_CODE:             VOP++; break;
>>>>> -            case VISUAL_OBJECT_START_CODE:  VISO++; break;
>>>>> -            case 0x100 ... 0x11F:             VO++; break;
>>>>> -            case 0x120 ... 0x12F:            VOL++; break;
>>>>> -            case 0x130 ... 0x1AF:
>>>>> -            case 0x1B7 ... 0x1B9:
>>>>> -            case 0x1C4 ... 0x1FF:            res++; break;
>>>>> -            }
>>>>> -        }
>>>>> +        if ((temp_buffer & 0xffffff00) != 0x100)
>>>>> +            continue;
>>>>> +
>>>>> +        b = probe_packet->buf[i];
>>>>> +        if (b == VOP_START_CODE)
>>>>> +            VOP++;
>>>>> +        else if (b == VISUAL_OBJECT_START_CODE)
>>>>> +            VISO++;
>>>>> +        else if (b < 0x20)
>>>>> +            VO++;
>>>>> +        else if (b < 0x30)
>>>>> +            VOL++;
>>>>> +        else
>>>>> +            res += !((b > 0xAF && b < 0xB7) || (b > 0xB9 && b < 0xC4));
>>>>>      }
>>>>>  
>>>>>      if ( VOP >= VISO && VOP >= VOL && VO >= VOL && VOL > 0 && res==0)
>>>>>       
>>>>>           
>>>> Why the new variable?  The logic looks correct though.
>>>>     
>>>>         
>>>   As a precaution and cosmetics. Precaution against stupid compilers,
>>> cosmetics because:
>>>         if (b == ....)
>>> looks nicer (IMHO) than:
>>>         if ((temp_buffer & 0xffffff00) == ...)
>>>
>>>   
>>>       
>> i don't see why you would have to do the bitmask anyway...
>> when you reach the b variable, (temp_buffer & 0xffffff00) == temp_buffer
>>     
>
> Not exactly true.  You mean (temp_buffer & 0x1ff) == temp_buffer.
>
>   

indeed.
and we can replace your "not exactly true" by "completely wrong" :)

>> also, not introducing this new b variable would leave start code
>> constant definitions as is, which is their "real" values :)
>>     
>
> I'm not so sure about that.  I don't have the MPEG4 spec at hand, but
> MPEG2 calls the 24 leading bits (0x000001) the "start code prefix",
> the "start code" being the following 8 bits.
>   

i don't know and i cannot verify neither

-- 
Ben
Purple Labs S.A.
www.purplelabs.com