[FFmpeg-devel] [PATCH] Issue 636 : ALAC encoding sometimes fails

[Jai Menon]

Hi,

On Thu, Dec 11, 2008 at 3:05 PM, M?ns Rullg?rd <mans at mansr.com> wrote:
> "Jai Menon" <jmenon86 at gmail.com> writes:

[...]

>>>>>              sum >>= lpc.lpc_quant;
>>>>>              sum += samples[0];
>>>>> -            residual[i] = (samples[lpc.lpc_order+1] - sum) << (32 -
>>>>> s->write_sample_size) >>
>>>>> +            residual[i] = (int32_t)(samples[lpc.lpc_order+1] - sum) << (32
>>>>> - s->write_sample_size) >>
>>>>>                            (32 - s->write_sample_size);
>>>>
>>>> id do the cast after the << because its the >> that differes from where
>>>> the sign bit is
>>>
>>> Put differently, if int is wider than 32 bits the left-hand argument
>>> to >> must be sign-extended from 32 bits to this size for the >> to
>>> work as intended.
>>>
>>
>> done.
>>
>>> It would be more efficient if 32 were replaced with CHAR_BIT*sizeof(int)
>>> instead.  This could be put in a nicely named macro, making it even more
>>> obvious what the intent of the code is.
>>
>> Also done.
>
> With this variant the cast is wrong.  Think about what happens if int
> is wider than 32 bits.

Is the attached patch okay?
Maybe this will help fix the regression failure on 64 bit systems.

-- 
Regards,

Jai
-------------- next part --------------
A non-text attachment was scrubbed...
Name: alacenc_issue636.patch
Type: text/x-patch
Size: 565 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090107/feda1a69/attachment.bin>