Encoder warning with B-Frame: what does it mean?

[Eric B]

I've encoded movies (E.T. & T2 Z1 converted in 23.976fps with dvd2avi) with divx 5.0.5 (home theater profile @ 1300k) under last VDMod version. I've used also robot4rip & Gordianknot.
I have "done(warnings)" status on some encodes. It is 3 following times at the second pass. I 've re-encode one but the warning are still here.

I've successed to recopy the warning (occured one time when looking at the log tab during encoding):
"Video codec produced delay frame while trying to flush codec B-frame delay! Ignoring to avert possible infinite loop."

What does it mean? Sould I recompress my movie without B-Frame ?

[Shark]

Well, if it can make you less alone, i get the same pb here :D

Except i get this warning with 5.03.

I think it's just a bebugger things but there's still a pb with that...

Shark

[jonny]

There is a special "return code" while encoding a frame using bframes, this return code indicates that the frame must be skipped (i don't go into details, but is needed to handle bframes).

Now, DivX 5.0.5 have a bug: it returns, sometimes, an infinite number of those "skipping codes" at the end of the second (or multipass) encode.

VirtualDub detect this and avoid an infinite loop (the result is usually an Avi file with the last frame missing).

[SeeMoreDigital]

I think the problem also, unfortunately occurs more frequently for lower processor speed PC users. Especially when visiting the web or just having other programs running in the background, while encoding!

Cheers

[Shark]

Well, i'm quite sure the cpu does not have anything to do with this particular error message. The reason is that i have a p4 2.8 and 1024 Mo DDR400 and i get this error as i don't do anything at the same time. More strange, i get this error on a given movie but not on another, which also mean that the codec has nothing to do neither.

Jonny, i would really appreciate if you could point me to somewhere where i could get some more informations about what you suggest as a reason. I like to understand exactly what happens ;)

Last but not least, i just discovered something!
I do a a sample of encoding with a vob sample and i don't have this error. I take the same vob sample but this time i use trim in order to do a sample of this sample (ie instead of encoding 1 min (the vob content) i just encode 30 sec. and then i get this error! Is it possible that the trim function would have cut in between 2 keyf and generate an error?

Shark

[jonny]

You can find more inside the VirtualDub sourcecode :D (the problem is mentioned in the VDub's change log - but detailed infos are only in the sourcecode).

From Dub.cpp (1.5.4 sources):

DivX 5.0.5 seems to have a bug where in the second pass of a multipass operation it outputs an endless number of delay frames at the end! This causes us to loop infinitely trying to flush a codec delay that never ends. Unfortunately, there is one case where such a string of delay frames is valid: when the length of video being compressed is shorter than the B-frame delay. We attempt to detect when this situation occurs and avert the loop.

If you are a programmer or you understand in some way programming stuffs i can explain you the details, let me know.

Cheers
jonny

[jonny]

Last but not least, i just discovered something!
I do a a sample of encoding with a vob sample and i don't have this error. I take the same vob sample but this time i use trim in order to do a sample of this sample (ie instead of encoding 1 min (the vob content) i just encode 30 sec. and then i get this error! Is it possible that the trim function would have cut in between 2 keyf and generate an error?

This have nothing to do with trim, it also happens using a simple "version()" avs script.
If i tell you is DivX related... is DivX related, period. :)
( The error is systematic, not random like SeeMoreDigital says, there is probably a rule to make it happen or not... but i'm really not interested to find it :D )

[SeeMoreDigital]

Strange

Only last week I generated 3No 2pass encodes of the NTSC version of T2 Extreme at 'film speed'. Using a video setting of 1292kbps and an audio setting of 128kbps

The first was with DivX 5.0.2, the second was with DivX 5.0.5 and the third was with DivX 5.1.1 beta2!

There were no errors reported with any of them!

Sorry I can't offer anymore more help

And in case you were wondering, DivX 5.1.1 beta2 did look the best!

Cheers