[FFmpeg-devel] [PATCH] Indeo5 decoder

Maxim max_pole
Thu Apr 2 13:14:11 CEST 2009

Reimar D?ffinger schrieb:
> On Thu, Apr 02, 2009 at 11:40:49AM +0200, Maxim wrote:
>>> useless & [0]
>>> indeo5patch.txt:676:+                band->dequant_intra = &deq4x4_intra[0][0];
>>> indeo5patch.txt:677:+                band->dequant_inter = &deq4x4_inter[0][0];
>> changing it to
>> band->dequant_intra = deq4x4_intra;
>> causes the following compiler's message:
>> warning: assignment from incompatible pointer type
> One & cancels with one [0], so that would be
> band->dequant_intra = deq4x4_intra[0];

thanks, fixed...

>>> indeo5patch.txt:1035:+    av_freep(&planes[0].buf1);
>>> indeo5patch.txt:1036:+    av_freep(&planes[0].buf2);
>> Hmm, shouldn't av_freep receive a POINTER to a POINTER as it's declared
>> in its prototype? Will then the removing of the "&" lead to a wrong
>> program behavior?
> I think those a false positives, if you still haven't noticed that was
> an automated "review" by tools/patcheck

Aha, thanks! I didn't noticed it because I'm not familiar with this

> You could probably use &planes->buf1 here though, but that's probably
> worse (i.e. uglier)

I would leave it as is because there are several other statements like
so the line
looks more readable IMHO...
It would be possible to use constants instead of those numbers, i.e.


would it be better?

>>> missing const?
>>> indeo5patch.txt:1248:+    int32_t *src, *dst, tmp[64];
>>> indeo5patch.txt:1295:+    int32_t *src, *dst, tmp[16];
>> Yes, this is abit complicated! The src pointer should be declared as a
>> pointer to a const data in fact. But due to the given design of the
>> transform it cannot be made so because both IVI_IREFLECT and
>> IVI_SLANT_BFLY make changes to their operands. I need to redesign the
>> whole transform in order to make the src const. Will it worth it?
> This might just be a false positive by the script.
> I think Michael just wanted to make you look over the obvious things a
> simple tool can detect before "wasting" a human's time with a review
> (there are just too few people doing patch reviews).

Ok, I'll leave the code as is then...


More information about the ffmpeg-devel mailing list