[FFmpeg-devel] Fwd: QScale assertions cause trouble in KDEnlive with FFMPEG SVN

[Aleksander Adamowski]

Hello!

I've posted this to FFMPEG-users as a bug report, but received no reply.

I suppose I should have sent it to FFMPEG-devel, which I'm doing now.

---------- Forwarded message ----------
From: Aleksander Adamowski <aleksander.adamowski at olo.org.pl>
Date: Jul 4, 2007 7:40 PM
Subject: QScale assertions cause trouble in KDEnlive with FFMPEG SVN
To: ffmpeg-user at mplayerhq.hu


Hi!

On the KDEnlive forum there's a discussion about a problem during
export of a project to some formats:

http://www.kdenlive.org/bbforum/viewtopic.php?f=16&t=30&start=40

The problem has been tracked down to libavcodec/ratecontrol.c - more
specifically, the asserts on qscale value that are present in the
ff_rate_estimate_qscale() function.

When the following two assertions are commented out, the problem stops occuring:

Index: libavcodec/ratecontrol.c
===================================================================
--- libavcodec/ratecontrol.c    (revision 9460)
+++ libavcodec/ratecontrol.c    (working copy)
@@ -753,7 +753,7 @@
//printf("%f ", q);
         q= get_diff_limited_q(s, rce, q);
//printf("%f ", q);
-        assert(q>0.0);
+//        assert(q>0.0);

         if(pict_type==P_TYPE || s->intra_only){ //FIXME type
dependent blur like in 2-pass
             rcc->short_term_qsum*=a->qblur;
@@ -765,7 +765,7 @@
             q= short_term_q= rcc->short_term_qsum/rcc->short_term_qcount;
//printf("%f ", q);
         }
-        assert(q>0.0);
+//        assert(q>0.0);

         q= modify_qscale(s, rce, q, picture_number);



More details are available in the history of the forum thread I've
linked to on www.kdenlive.org.

Could anyone with the knowledge of FFMPEG source try to comment on that problem?

The asserts were probably placed over there for a reason and
eliminating them isn't a correct solution after all...


-- 
Best Regards,
  Aleksander Adamowski
  http://olo.org.pl