Feedback from Matroska testing

[Dandier]

I am new in Matroska, for I was previously waiting for Matroska to offer the same functionalities as OGM (i.e. subtitles, chapters,...)

When I've heard that everything that I need was now available, I have downloaded and tested the last release. Unfortunately I did not succeed in issuing a single movie. However, I have to admit that the two samples I have downloaded work okay. So I suspect that I am misusing something while creating my files.

To test Matroska I first tried to mux an OGM movie into a MKV with both VirtualDubMod 1.5.4.1 build 2066 and mkvmerge GUI packaged into MMG 0.6.0-r2.

Here are my main observations: The file muxed with VirtualDubMod can not be openned by mkvmerge (Error: Demultiplexer failed to initialize: matroska_reader: Failed to read the headers.) On the other hand the file muxed with mkvmerge is openned by VirtualDubMod.
Playback with MPC 6.4.5.9 and Matroska Splitter 1.0.1.6: OKay except that the subtitles are disabled (well, this is normal since VSFilter is not installed!)
Playback with TCMP V4 RC4 build 43 and Matroska Splitter 1.0.1.6: same as previous except that the chapter browsing does not work either.
Playback with BSPlayer 0.86 build 500 and Matroska Splitter 1.0.1.6: none of my movies can be read at all. The movie properties are correct but nothing is played (no video, no audio). Note that the downloaded samples play well with this player.
Playback with Matroska Splitter 1.0.1.6 and VSFilter 2.28: my video does not play with neither player. With BSplayer and MPC 6.4.5.9 nothing at all is read (no audio, no video); exactly the same situation as BSplayer without VSFilter... With TCMP V4 RC4: the audio plays but no video is displayed. With the downloaded samples everything works OK, except that the subtitle filter is not loaded when using BSplayer. I have then muxed a small clip directly from an AVI, and the same problems appear.

Otherwise, I have the following small minor comments about MKV: It seems significantly longer to open a MKV file than an OGM or an AVI. Anyway this delay is acceptable if it prevents the movie from crashing during the playback...
The audio track can not be selected through its language (as with OGM) but through a sequence number (Audio 1, Audio 2). It's really a detail, but I can imagine that it should not cost too much to fix.As a conclusion, I was unable to generate a fully featured movie with audio tracks, subtitles and chapters. This is pretty frustrating because I feel that other people did (the samples do feature everything but chapters); and this format really looks great.

May be I missed something in installing my environment. In that case I will greet any help. Otherwise, I hope that my report will be helping the developpers to correct some youth problems and turn this format into a new universal standard.

Cheers

[Mosu]

Originally posted by Dandier
The file muxed with VirtualDubMod can not be openned by mkvmerge (Error: Demultiplexer failed to initialize: matroska_reader: Failed to read the headers.) On the other hand the file muxed with mkvmerge is openned by VirtualDubMod.

This is a known bug that I've already fixed. The last release (0.6.2) was done yesterday evening, and it should address this issue. See also http://forum.doom9.org/showthread.php?s=&threadid=55637&perpage=&pagenumber=3

[jcsston]

Dandier: Could you run this on one of your Matroska files you created
http://www-user.tu-chemnitz.de/~noe/Video-Zeug/mkvverify.zip
and post the results?
You can run 'mkvverify -help' for help.

[Dandier]

Woo! The reactivity is really impressive on this forum!

To Mosu: I have downloaded the last Mkvmerge version, parsing of VDubMod files is now okay.

To jcsston: here are the output for some of my movies and clips. I have reported both stderr and stdout, so it's a little bit long... But basically there is distinct errors for files generated with VDubMod (file is b0rked!?) and with mkvmerge (file is obfuscated).

My large movie made with Mkvmerge 0.6.0
(mkv parser) file is obfuscated:
Segment Info present but not refer to in existing Seekhead

Duration : 01:33:00.597
File size: 667 MByte
Segment 0
Muxing App : libebml v0.5.0 + libmatroska v0.5.0
Writing App : mkvmerge v0.6.0
Segment UID : 17 DE A9 CB 9A DA A2 80 6F 24 A7 3F 79 12 10 6B
Timecode Scale: 1000000
Track 0
Number : 1
Type : video
Lacing : no
CodecID : V_MS/VFW/FOURCC
biCompression : DX50
biBitCount : 24
Default : yes
Enabled : yes
MinCache : 1
MaxCache : 1
Language : eng
Resolution :
Pixels : 640x272
Display : 640x272
Unit : pixel
Track 1
Number : 2
Type : audio
Lacing : yes
CodecID : A_VORBIS
Default : no
Enabled : yes
MinCache : 0
MaxCache : 0
Language : fre
sampling frequency: 48000 Hz
channels : 2
Track 2
Number : 3
Type : audio
Lacing : yes
CodecID : A_VORBIS
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
sampling frequency: 48000 Hz
channels : 2
Track 3
Number : 4
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : fre
Track 4
Number : 5
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : no
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
Track 5
Number : 6
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : no
Enabled : yes
MinCache : 0
MaxCache : 0
Language : por

My small clip made with VDubMod
(mkv parser) file is b0rked:
temporal inversion: chapter ends earlier than it starts

(mkv parser) file is b0rked:
end of last segment is not end of file

Duration : 00:00:44.625
File size: 6.60 MByte
Segment 0
Muxing App : libebml-0.5.0 & libmatroska-0.5.0
Writing App : VirtualDubMod 1.5.4.1 (build 2066/release)
Segment UID : A4 0C 52 06 D0 21 E8 10 B0 67 D8 33 B4 0D DA 06
Timecode Scale: 1000000
Track 0
Number : 1
Type : video
Lacing : no
CodecID : V_MS/VFW/FOURCC
biCompression : XVID
biBitCount : 16
Default : yes
Enabled : yes
MinCache : 1
MaxCache : 0
Language : eng
Resolution :
Pixels : 160x120
Display : 160x120
Unit : pixel
Track 1
Number : 2
Type : audio
Name : Original Track
Lacing : yes
CodecID : A_MS/ACM
wFormatTag : 0x0001
nAvgBytesPerSec : 88201
nBlockAlign : 2
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
sampling frequency: 44100 Hz
channels : 1
bitdepth : 16
Track 2
Number : 3
Type : audio
Name : Lepidoptera
Lacing : yes
CodecID : A_VORBIS
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
sampling frequency: 44100 Hz
channels : 2
Track 3
Number : 4
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
Track 4
Number : 5
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : fre


My small clip made with Mkvmerge 0.6.0(Note: same output with Mkvmerge 0.6.2)

(mkv parser) file is obfuscated:
Segment Info present but not refer to in existing Seekhead

Duration : 00:00:44.625
File size: 6.67 MByte
Segment 0
Muxing App : libebml v0.5.0 + libmatroska v0.5.0
Writing App : mkvmerge v0.6.0
Segment UID : 41 8F 49 11 BE 6F 56 C2 7C B5 20 3E 82 3C C8 7A
Timecode Scale: 1000000
Track 0
Number : 1
Type : video
Lacing : no
CodecID : V_MS/VFW/FOURCC
biCompression : XVID
biBitCount : 24
Default : yes
Enabled : yes
MinCache : 1
MaxCache : 1
Language : eng
Resolution :
Pixels : 160x120
Display : 160x120
Unit : pixel
Track 1
Number : 2
Type : audio
Lacing : yes
CodecID : A_PCM/INT/LIT
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
sampling frequency: 44100 Hz
channels : 1
bitdepth : 16
Track 2
Number : 3
Type : audio
Lacing : yes
CodecID : A_VORBIS
Default : no
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
sampling frequency: 44100 Hz
channels : 2
Track 3
Number : 4
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : yes
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng
Track 4
Number : 5
Type : subs
Lacing : no
CodecID : S_TEXT/UTF8
Default : no
Enabled : yes
MinCache : 0
MaxCache : 0
Language : eng

[darth rosenberg]

OK, 'obfuscated' should be reduced to 'warning'. The SegmentInfo should be indexed in the SeekHead, but a file has to be processed correctly as well if it is not indexed. It is not really an 'error'.

The VDubMod files are really not OK

[[Toff]]

I've sent a correction to Belgabor and Cyrius about the last chapter end time bug. So it should be soon in CVS. (I hope as Cyrius is in vacation :cool: )

[Dandier]

I am sad to see that Darth Rosenberg considers this only has a warning, because from my point of view it prevents me from watching my movies! ;)

More seriously I'd like to point out two important things: When playbacking my clips I get no explicit error message about parsing or decoding problems. It drives me into thinking that the problem should be more vicious that a simple format error. When VSFilter is not loaded (either because it's not installed or because the clip features no subtitle), nearly everything is OK (with a small exception concerning chapters; cf. below). I guess that there should be some kind of incompatibilities between the format of my files and VSFilter.Well! These are just assumptions from a complete dummy in video technologies...

Concerning the chapters, I've made new testing. It appears that my previous mkvmerge video did not feature chapters (because they were not automatically imported from the OGM). I created some chapters with mkvmerge 0.6.2, and it works with both players (TCMP and MPC). I just wish I could browse with the << and >> buttons, instead of using the navigation menus; but it's a detail. My guess here is that the browsing problem is due to the bug reported by Toff and is specific to VDubMod. So everything seems under control for that peculiar aspect.

[darth rosenberg]

I am sad to see that Darth Rosenberg considers this only has a warning, because from my point of view it prevents me from watching my movies! Are you sure that exactly this one thing causes your problems? If it does, then the reader is broken...

[Dandier]

Originally posted by darth rosenberg
Are you sure that exactly this one thing causes your problems? If it does, then the reader is broken...

Of course not! And I think that the rest of my post confirms that I do suspect some more complex and hidden problems. As a matter of fact, I've just ran mkvverify on the downloaded samples and the obfuscated warning is also issued. So we can be very confident that it is not the problem source. But frankly, you must aknowledge that it does not help that much!!!

[Belgabor]

Originally posted by [Toff]
I've sent a correction to Belgabor and Cyrius about the last chapter end time bug. So it should be soon in CVS. (I hope as Cyrius is in vacation :cool: )

It now is =)

[Hiro2k]

While we are on the subject, I don't know if anyone else has had problems with MMG, but mine is kinda strange. I start the process like before and as soon as I start the process my whole system LAGS and about 5 minuets into it, windows says that it has run out of virtual memory to use, so I checked and my PF Usage was around 700MB. The CPU Usage however stays at about 20% durring the whole process. The bar in mmg doesn't show any progression and the system just keeps getting slower and slower. So I cancelled it manualy in the XP Process. But I find out that it hasn't cleared out any of the PF and forces me to reboot. I thought it might have been just mkvmerge 0.6.0, so I upgraded to 0.6.2. That gave me the same problem.

Am I doing something wrong?? Cuase I've already seen working Matroska files.

[Mosu]

Originally posted by Hiro2k
While we are on the subject, I don't know if anyone else has had problems with MMG, but mine is kinda strange. I start the process like before and as soon as I start the process my whole system LAGS and about 5 minuets into it...

Are you trying to read Matroska files? I've heavily borked the Matroska reader sometime during the last month, and I just realized it today :( The result is that mkvmerge tries to suck the complete source file into memory... Which kills system performance almost every time if the file is pretty big.

I'm fixing it now, so please be a bit patient :) Thanks :)

[Teegedeck]

BTW, do you know that the muxer has a mem leak? Memory consumption grows during operation, which is especially nasty when muxing several files subsequently.

[Mosu]

Originally posted by Teegedeck
BTW, do you know that the muxer has a mem leak? Memory consumption grows during operation, which is especially nasty when muxing several files subsequently.

Which muxer? mkvmerge, Gabest's DShow muxer, VDubMod...? What kind of source files are you using?

[Hiro2k]

Mine sources were AVI, 2 OGG, and 1 SRT plus Chapters.

[Teegedeck]

mkvmerge version 0.6.2

Sources: An AVI with VBR MP3 soundtrack. I repeatedly tried remuxing them with differing audio-interleave values when the leak became rather obvious.

[Mosu]

Originally posted by Teegedeck
mkvmerge version 0.6.2

Sources: An AVI with VBR MP3 soundtrack. I repeatedly tried remuxing them with differing audio-interleave values when the leak became rather obvious.

How can a leak become obvious by running a program several times? Does Windows not free memory a program allocates when it terminates?

[Teegedeck]

Yes, that was a perceptual error on my behalf, I guess: On the first run, mkvmerge started out at 7 MBs Ram and ended at 11 MBs; on the second I chose cue i-frames for the audio by mistake and mkvmerge ended up at something above 40 MBs of Ram. No cumulative rise in memory-consumption.

Now I'd like to experiment some more to verify - but alas mkvmerge keeps throwing a 'EBMLElement.cpp, line 497' error at me: 'S are ignored for AVI-files'. Probably just the mmg-GUI but I'm a tad too impatient to type a command-line, now. Don't get me wrong; I like matroska and all but I don't feel like alpha-testing it - at least not today.

[Mosu]

Originally posted by Teegedeck
Now I'd like to experiment some more to verify - but alas mkvmerge keeps throwing a 'EBMLElement.cpp, line 497' error at me: 'S are ignored for AVI-files'.

These two messages are unrelated. The latter is only a warning by mkvmerge itself and can be ignored safely. The first, however, is an error message by the libraries. Once you're in the mood to test again, could you paste the complete output? :)

[Teegedeck]

Will be done but don't hold your breath. ;) I should be working on something important while I type this. Should have been for days, now...