[FFmpeg-devel] [PATCH] WMA Voice decoder

[Ronald S. Bultje]

Hi,

On Thu, Feb 11, 2010 at 1:53 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
> On Wed, Feb 10, 2010 at 10:34 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
>> On Tue, Feb 2, 2010 at 2:58 PM, Ronald S. Bultje <rsbultje at gmail.com> wrote:
>>> I have something without the crazy loops and using av_log2(), maybe
>>> this is better?
>>
>> And here's V6. Thanks to Reimar & Vitor for all the review!
>>
>> (Quant and intercoeff tables omitted from this patch to ease the
>> mailinglist software.)
>
> v7 here, bikeshed issue fixed and some optimizations for
> aw_pulse_set2() and aw_parse_coords() (almost all loops gone!) from
> Vitor and Kostya.

As requested on IRC, I ran a couple of sample files (alltogether using
all features of the decoder) through zzuf (with -b X-Y to skip messing
up the header, since then the file can't be recognized and nothing
happens) with various -r levels. It never crashed, but output
increasingly became noisy (with -r levels around .0001) or started to
skip (around -r 0.001). At -r 0.01, it barely generated any frames
anymore, mostly because the ASF demuxer fails to read complete frames.

So I conclude that there's no obvious security issues either. I'll
probably apply tomorrow unless someone complains loudly.

Ronald