[MPlayer-dev-eng] vbr mp3 runtime

[Ingo Brückl]

Reimar Döffinger wrote on Thu, 1 Sep 2011 00:24:06 +0200:

> On Mon, Aug 29, 2011 at 04:49:37AM +0200, Ingo Brückl wrote:
>> compn wrote on Sun, 28 Aug 2011 20:31:22 -0400:
>>
>> > On Mon, 29 Aug 2011 01:52:15 +0200, Ingo Brückl wrote:
>> >>Seems to work just fine.
>>
>> > have you tried it with an mp3 file and a fuzzer (like zzuf) to make
>> > sure it doesnt crash or leak mem etc?
>>
>> Currently it's just a feasibility study. I won't put work into it until I
>> know that it will be accepted in principle.

> However exactly that is exactly the problem.
> After the last time such stuff was added it ended up with about
> one major security hole per line (and lots of lines of code)

Hmm, what can possibly become a security hole? The worst thing that can
happen is that we get a wrong number of frames leading to a wrong value
for i_bps if a vbr header exists, which is exactly what we are having at
the moment.

> I am seriously doubting that it's a good idea.

Well, if it's unwanted ... I just thought it would be nice to have vbr
support...

> It's possible to just prefer lavf demuxer for those files.
> No idea if there's any reason not to.

Yes, it is possible, but because you don't know in advance you have to use
the lavf demuxer by default. (Which leads me to the question, why do we have
own demuxers and don't just use ffmepg's? Serious question.)

Ingo