[MPlayer-dev-eng] [RFC] Known regressions

[Corey Hickey]

Reimar Döffinger wrote:
> Hello,
> On Sat, Sep 02, 2006 at 12:55:29PM -0400, Rich Felker wrote:
>> On Sat, Sep 02, 2006 at 09:37:25AM -0700, Corey Hickey wrote:
>>>> - [pre8 and SVN] the infamous "2pass curve failed to converge" problem
>>>>   since ffmpeg r5422.
>>> Yeah. That's been bothering me for a while. I'm going to verify the
>>> problem is still reproducible and then ask on ffmpeg-devel if reverting
>>> r5422 is at all acceptable.
>> It's not; the old code was just wrong. If you want to "revert" then
>> make an explicit option to ignore the convergence failure since thats
>> what the buggy old code essentially did. However I'm told that the
>> problem only happens when you erroneously specify a bitrate that
>> cannot be met given the qmin/etc. params.
> 
> Well, but when it aborts an encode because of that it's not great
> either. Maybe it can be made a non-fatal warning? And maybe also with a
> more helpful message.
> What I want to say with that, I think something should be done about it
> nevertheless.

Please take a look at the patch I just sent to ffmpeg-devel.

Subject: [PATCH] fix (?) 2pass convergence errors

-Corey