[FFmpeg-devel] [PATCH] avcodec/texturedspenc: Fix indexing in color distribution determination

[Marton Balint]

On Tue, 24 May 2022, Michael Niedermayer wrote:

> On Wed, Jan 04, 2017 at 11:52:50PM +0100, Vittorio Giovara wrote:
>> On Wed, Jan 4, 2017 at 1:06 AM, Michael Niedermayer
>> <michael at niedermayer.cc> wrote:
>>> On Tue, Jan 03, 2017 at 02:15:20PM +0100, Vittorio Giovara wrote:
>>>> On Mon, Jan 2, 2017 at 2:00 AM, James Almer <jamrial at gmail.com> wrote:
>>>>> On 1/1/2017 8:28 PM, Michael Niedermayer wrote:
>>>>>> Fixes CID1396405
>>>>
>>>> What is the CID about?
>>>
>>> sorry for not quoting this prviously, it isnt very interresting.
>>> my patch is based on just this and guesswork, which is why it should
>>> be tested.
>>>
>>> 257        for (y = 0; y < 4; y++) {
>>>    assignment: Assigning: x = 4.
>>>    const: At condition x < 4, the value of x must be equal to 4.
>>>    dead_error_condition: The condition x < 4 cannot be true.
>>> 258            for (x = 4; x < 4; x += 4) {
>>>    CID 1396405 (#1 of 1): Logically dead code (DEADCODE)dead_error_begin: Execution cannot reach this statement: muv += bp[x * 4 + y * stride];.
>>
>> Ah, yes that seems wrong, probably a leftover of the code conversion
>> (from https://github.com/nothings/stb/blob/master/stb_dxt.h#L293-L298)
>> I believe it should be sufficient to do `for (x = 0; x < 4; x++)` but
>> I won't be able to test that for a while I'm afraid
>
> This issue seems stil open, i just stumbled accross the CID again
> when looking at coverity
> anyone has means to test this ?

Actually fate now covers this in vbn-dxt5 and vbn-dxt1 tests. Average 
MSE/PSNR is slightly improved with the patch, so I have applied it.

Regards,
Marton