[FFmpeg-devel] Strange issue debugging AAC

[Alex Converse]

On Thu, Sep 18, 2008 at 11:03 AM, Robert Swain <robert.swain at gmail.com> wrote:
> 2008/9/18 Alex Converse <alex.converse at gmail.com>:
>> On Thu, Sep 18, 2008 at 3:02 AM, Siarhei Siamashka
>> <siarhei.siamashka at gmail.com> wrote:
>>> On Thursday 18 September 2008, Alex Converse wrote:
>>>> Hey all,
>>>>
>>>> I'm running into a weird issue with get_bits() in the aac decoder that
>>>> I was hoping some one could help me with so that I can continue
>>>> hunting down bugs.
>>>>
>>>> Consider the attached patch:
>>>>
>>>> and the test stream at
>>>> http://standards.iso.org/ittf/PubliclyAvailableStandards/ISO_IEC_14496-4_20
>>>>04_Conformance_Testing/audio_conformance/mpeg4audio-conformance/compressedMp
>>>>4/al07_44.mp4
>>>>
>>>> I'm getting different results from the #if and #else sections of the
>>>> code, and it's boggling my mind. Any assistance would be greatly
>>>> appreciated. I've tried both gcc-3.4 and gcc-4.3.
>>>
>>> +    int scale_bits;
>>> ...
>>> +#if 1
>>>     scale = pow(2., pow(2., get_bits(gb, 2) - 3));
>>> +#else
>>> +    scale_bits = get_bits(gb, 2);
>>> +    scale = pow(2., pow(2., scale_bits - 3));
>>> +#endif
>>> +    av_log(ac->avccontext, AV_LOG_ERROR, "scale = %f\n", scale);
>>>
>>>
>>> get_bits returns unsigned int, so "(double)(get_bits(gb, 2) - 3)" and
>>> "(double)((int)get_bits(gb, 2) - 3)" are not identical.
>>>
>>
>> Thanks. Trivial patch attached.
>
> Did you intend this to be applied? I'll apply the cast, of course, but
> the av_log() is undesirable for normal use. :)
>

Whoops, I didn't intend for the av_log to stay in there. Sorry, about that.

--Alex