Folks:

I've had very uneven experience with MythArchive. Sometimes it works, most often it fails with a message that mytharchivehelper has blown up while trying to extract the menu thumbnails. I've tried a number of things suggested in the forums here -- many of which amounted to just reinstalling the latest SVN without regard to what the problem might be. True, sometimes holding your hands over your ears and squinting real hard while the compiler compiles will fix things... but not this time.

Tonight's experimentation leads me to think that the problem changes from file to file. Below you'll see two 40-line excerpts from mythburn.log from two runs, one successful and one not. It appears that in the unsuccessful one mytharchivehelper is either unable to identify the codec for the video stream, or is misidentifying it as mp3 -- despite the fact that earlier in the run it is correctly identified.

Is there anything I can do to fix this -- meaning, is there anything I can do to find and repair whatever it is I screwed up in the installation -- or is it an honest-to-god bug in the codec detection code for mytharchivehelper

Also, it would seem to me that a thumbnail failure is not a reason to stop the entire parade, but I guess people do love their little pictures. Me, I'd be happy with a green square.

I can send more of the relevant logs if needed -- I just hate it when people clutter up forum posts with 2,000 line files of which 5 lines are relevant to the problem (gripe, gripe).

t.

=====
FIRST FILE (SUCCESSFUL):

mythburn.py (0.1.20070318-2) starting up...

........... many lines removed ...............

Input #0, mpeg, from '/video/1043_20070319033000.mpg':
Duration: 07:56:30.1, start: 0.333856, bitrate: 1651 kb/s
Stream #0.0[0x1e0]: Video: mpeg2video, yuv420p, 720x480, 8600 kb/s, 29.97 fps(r)
Stream #0.1[0x69]: Data: 0x0000
Stream #0.2[0x1c0]: Audio: mp2, 48000 Hz, stereo, 384 kb/s
0: start_time: 0.036 duration: 814.885
1: start_time: 1758.248 duration: 814.893
2: start_time: 0.030 duration: 814.856
stream: start_time: 0.334 duration: 28590.125 bitrate=1651 kb/s
2007-03-20 17:45:03.092 duration = 28590
streaminfo.xml :-
<?xml version="1.0" ?>
<!DOCTYPE FILEINFO>
<file duration="2820" filename="/video/1043_20070319033000.mpg" type="mpeg">
<streams count="3">
<video aspectratio="1.33333" bitrate="8600000" codec="mpeg2video" ffmpegindex="0" fps="29.97" height="480" id$
<data codec="Data: 0x0000" streamindex="1"/>
<audio bitrate="384000" channels="2" codec="mp2" ffmpegindex="1" id="448" language="N/A" samplerate="48000" s$
</streams>
</file>

............. many, many, many lines removed ......................

Extracting thumbnail images from: /video/tmp/work/1/stream.mv2 - at 10,964,1928,2892,3856,4820,5784,6748,
Destination file /video/tmp/work/1/chapter-%1.jpg
list.count(): 8
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
[mpeg2video @ 0x1fbb568]warning: first frame is no keyframe
Added image /video/tmp/work/1/chapter-1.jpg

........... many more lines, runs to successful completion of burn ....


SECOND FILE: UNSUCCESSFUL


mythburn.py (0.1.20070318-2) starting up...

.......... many lines removed .............

Input #0, mpeg, from '/video/1022_20070311190000.mpg':
Duration: 02:00:55.4, start: 0.333811, bitrate: 5190 kb/s
Stream #0.0[0x1e0]: Video: mpeg2video, yuv420p, 720x480, 8600 kb/s, 29.97 fps(r)
Stream #0.1[0x1c0]: Audio: mp2, 48000 Hz, stereo, 384 kb/s
0: start_time: 0.036 duration: 652.984
1: start_time: 0.030 duration: 652.957
stream: start_time: 0.334 duration: 7255.448 bitrate=5190 kb/s
2007-03-20 17:30:12.656 duration = 7255
streaminfo.xml :-
<?xml version="1.0" ?>
<!DOCTYPE FILEINFO>
<file duration="2960" filename="/video/1022_20070311190000.mpg" type="mpeg">
<streams count="2">
<video aspectratio="1.33333" bitrate="8600000" codec="mpeg2video" ffmpegindex="0" fps="29.97" height="480" id$
<audio bitrate="384000" channels="2" codec="mp2" ffmpegindex="1" id="448" language="N/A" samplerate="48000" s$
</streams>
</file>

............ many, many, many lines removed ...............


Extracting thumbnail images from: /video/tmp/work/1/stream.mv2 - at 10,649,1298,1947,2596,3245,3894,4543,
Destination file /video/tmp/work/1/chapter-%1.jpg
[mp3 @ 0x844320]Could not find codec parameters (Audio: mp2, 56 kb/s)
************************************************************
ERROR: Failed while running mytharchivehelper to get thumbnails
************************************************************

Running mytharchivehelper from the command line confirms that it just can't find what it's looking for in the stream.mv2 file created by mythreplex. A look at the source code shows the work being done by ffmpeg, specifically utility functions in libavformat. So I brought ffmpeg up to date with the latest SVN feed. No change.

At this point it looks as though mythreplex is producing something mytharchivehelper can't identify. Think I saw something somewhere about dropping in projectX as a replacement; finding a way of improving or replacing mythreplex is clearly the next thing to try.

I do wish that mytharchivehelper had a better failure mode; I hate the fact that the ability to archive is held hostage by thumbnails, especially because my disk is filling.

Today I tried creating DVDs from the same file, 3 different ways:

1) ignoring the cutlist. This worked fine, with no problems at all.
2) with the original cutlist, mostly the one generated by the commercial flagger but with a few corrections. This had problems from the git-go, and did not burn
3) with a new cutlist somewhat more carefully edited. This one burned, but dropped audio after one of the commercial cuts.

Someone else theorized that "bad" cutlists were causing mythtranscode to do something stupid, and I think this confirms it. Too, I think I may know what "bad" means.

For some reason, the editor has always seemed to throw in extra cut points when you make a cut at the top of the program -- for example, I usually trim the beginning to the point where the program actually begins, about a minute into the recording usually. When I place that cut, with a "cut before this point", I usually get two cuts -- an "out" cut that I didn't ask for, as well as the "in" cut that I did. In example 2 above I had not done anything to get rid of the extra cut. In example 3 I did.

Example 3 dropped audio after the second commercial. As it happens, the commercial flagger had underestimated the extent of the commercial and so I ended up adding a later "in" cut and going back to delete the prior "in" cut. Or so I thought, anyway.

I am starting to suspect that mythtranscode spazzes out if it sees two cuts in a row that are both "in" or "out" cuts. I am going to go back and look at the edit list for 3) and see if I can see what the deal is. But it is definitely the case that cutlists are causing mythtranscode to lose its mind.

t.

Below is a dump of the cutlist for the recording I've been experimenting with, along with the query that produced it.

The list of type-0 and type-5 cuts corresponds pretty closely; these are the ends of commercials (what I was calling in-cuts in an earlier post). Likewise, type-1's and type-4's correspond pretty closely too. But not absolutely -- in particular with 1's and 4's. 0's and 5's are in a little disagreement about where the recording actually starts and ends, but that didn't seem to make any difference at the beginning of the recording that lost its audio in the last post. It lost audio at the commercial identified by
77943(type1) or 78993(type4) in the listing below.

So the next experiment to try will be forcing the type 4 cut to the same place as the type 1 in the database and seeing if that fixes it.

I wonder what mythtranscode in honorcutlist mode is actually looking at?
Any help?


mysql> select recordedmarkup.* from recordedmarkup,recorded where recorded.title LIKE '%Quick%'AND recordedmarkup.chanid=recorded.chanid AND recordedmarkup.star ttime=recorded.starttime;
+--------+---------------------+--------+--------+------+
| chanid | starttime | mark | offset | type |
+--------+---------------------+--------+--------+------+
| 1031 | 2007-03-18 20:00:00 | 0 | NULL | -3 |
| 1031 | 2007-03-18 20:00:00 | 1179 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 48806 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 86351 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 108648 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 132495 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 153386 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 181346 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 207562 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 230907 | NULL | 0 |
| 1031 | 2007-03-18 20:00:00 | 0 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 41452 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 77943 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 103985 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 126937 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 146926 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 173986 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 201095 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 223706 | NULL | 1 |
| 1031 | 2007-03-18 20:00:00 | 1 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 41452 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 78993 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 103985 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 126937 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 147826 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 173986 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 201095 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 223706 | NULL | 4 |
| 1031 | 2007-03-18 20:00:00 | 1117 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 48806 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 86351 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 108648 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 132495 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 153386 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 181346 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 207562 | NULL | 5 |
| 1031 | 2007-03-18 20:00:00 | 230905 | NULL | 5 |
+--------+---------------------+--------+--------+------+
37 rows in set (0.07 sec)
[/b]

Later... a look at the code shows that a mismatch ain't likely to be the problem; the query that gets the cutlist only looks at type 1 and type 0 cuts, and merges them correctly This makes sense, actually, because if it didn't, then lots of other things would break, not just transcoding. So we're back to the transcoder.

Sigh.

I had a problem of the sound dropping out completely after a cut on a DVD. Playback of the source file (which had been manually marked then losslessly transcoded to cut out commercials) was OK, but the drop out only occurred after writing to DVD.
I marked out a few frames on either side of the original cut point then losslessly transcoded again, that seemed to solve the problem of sound disappearing when writing to DVD.

Bruce S.

_________________
Updated 2019/10/26: AthlonII X2 265 Gigabyte GA-970A-DS3P
16Gb PC 1866 DDR3, 500GB+2TB+4TB SATA HDD,
SATA DVD-RW Asus DRW-24D5MT , NVIDIA GeForce GT1080
Hauppauage Nova-T 500, Nova-T LinHes R8.6.1

Bruce:

Thanks for the confirmation. I've tried monkeying with the cutlists now, and have had some success in getting stuff that wouldn't previously record to so so. Although actually it isn't recording that's the problem, it appears to be whatever mythtranscode does when it has to honor a cutlist. I suppose it could be mythreplex, too, but it seems not to care about cutlists. Whatever it is it mythtranscode is doing seems to be OK with mythreplex though.

While it's delightful talking to myself about this, I think I need to attract the attention of a developer or two. Anybody out there?

Yep me

I think you deserve this to be fixed if only because of the amount of work you've put in to track the problem down

First to clear up a few points raised in this thread.



Myth uses its own version of libavformat which is keep in sync with the ffmpeg version fairly regularly so updating ffmpeg wouldn't have any effect in this case.



That's a good point. Maybe it could just show a generic chapter image if the thumbs cannot be created. I do wonder though if it would just fail later on in the creation process. If mytharchivehelper cannot parse through the frames its highly likely mplex or dvdauthor will also fail. You could test that by trying to recreate a DVD which fails to create the thumbs by switch to one of the other DVD menu themes that don't use thumb images.



Someone was working on doing that but it caused other problems so the idea has been dropped for the time being. Better to get mythtranscode and mythreplex working together better.




Just a thought, how are you recording these programs DVB/ATSC? If so could it be that your broadcaster is transmitting the main program in one resolution and the promo's/commercials in another and that's what confusing mythtranscode:?: You could test that theory by making sure all the cuts are inside the main program and not in a commercial.

You are setting cut points on keyframes, right? If you don't cut on keyframes you will see audio/video synchronization problems. [So I've been told, I'm not an expert in this area].

You can move around in the editor by keyframe as well as cut point/frame/time.

_________________
nMediaPC HT100SA ASUS A8N-VM CSM
AMD 3200+ AMPO DDR 3200 512MB RAM
PVR-150 (1062) WD 320 GB SATA-2 HD
NEC OEM ND-3550A DVD burner
D-Link DWL-G510 PCI wireless adapter
MS wireless optical desktop 1000
R5F27 (+ modified sa3250ch)

Bruce and Paul:

Thanks, both, for the help. Some responses:
1) I don't know if the broadcaster is encoding commercials differently, but since I've now had the problem with material from at least six different sources I'd say this ain't it.
2) Yes, I did try cutting only at keyframes. In one case it helped, but then failed on at least 5 others.

I posted more on this here

http://mysettopbox.tv/phpBB2/viewtopic. ... highlight=

The Executive Summary is "beats me". It's clearly related to cutlists because mythtranscoder works fine when the cutlists aren't there, and when they are it creates something that nothing downstream will read. I've now been through up-to-date upgrades of everything, including the underlying Fedora goodies, with no improvement. This is getting really dull.

Last week I had another problem with another DVB-T recording with no sound on the DVD (as seen before here). Fortunately I never pre-transcoded the file, so I went back and marked the cut points at keyframes, then archived to DVD, guess what, it worked.
So it does appear transcode is "Keyframe Sensitive".

Bruce S.

_________________
Updated 2019/10/26: AthlonII X2 265 Gigabyte GA-970A-DS3P
16Gb PC 1866 DDR3, 500GB+2TB+4TB SATA HDD,
SATA DVD-RW Asus DRW-24D5MT , NVIDIA GeForce GT1080
Hauppauage Nova-T 500, Nova-T LinHes R8.6.1