mytharchive create DVD fails using projectx wth subtitles

[Larry Johnson <keepitsimpleengineer@…>]

When creating a DVD iso from two recordings with cutlists and subtitles selected, mytharchive failed. From mythburn.log

renameProjectXFiles start -----------------------------------------
chmod: changing permissions of `/home/mytharchive/temp/': Operation not permitted
------------------------------------------------------------

[Larry Johnson <keepitsimpleengineer@…>]

mythburn.log compressed

[Larry Johnson <keepitsimpleengineer@…>]

contents of temp/2 directory compressed

[paulh]

Do you still have the file newfile2_log.txt available? There appears to be something wrong with the parsing of ProjectX log file to get the file names.

[Larry Johnson <keepitsimpleengineer@…>]

Replying to paulh:

Do you still have the file newfile2_log.txt available? There appears to be something wrong with the parsing of ProjectX log file to get the file names.

Attached.

[J.Pilk@…]

For starters:

2010-12-14 11:27:32 Didn't find any subtitle elements in stream info file.

[Larry Johnson <keepitsimpleengineer@…>]

Replying to J.Pilk@…:

For starters:

2010-12-14 11:27:32 Didn't find any subtitle elements in stream info file.

Subtitles are in the original recording and show up in cutlist viewing, see attached screenshot Screenshot-MythTV Frontend-1062_20101104040000.jpg

[Larry Johnson <keepitsimpleengineer@…>]

Replying to Larry Johnson <keepitsimpleengineer@…>:

Replying to J.Pilk@…:

For starters:

2010-12-14 11:27:32 Didn't find any subtitle elements in stream info file.

Subtitles are in the original recording and show up in cutlist viewing, see attached screenshot Screenshot-MythTV Frontend-1062_20101104040000.jpg

So I uploaded the 1062_20101104040000.mpg file to my workstation and ran ProjectX(0.90.4.00b32⋯2009-08-23). ProjectX reported "Teletext: n/a" which matched what the htpc/ProjectX reported.

When I view the file on Mythtv htpc or the remote client on the workstation, the subtitles are present. I have no idea why this is.

Should this be another bug report?

[J.Pilk@…]

I'm afraid I don't know enough about how subtitles are encoded to say what's happening, but my advice for MythArchive is to take to heart the philosophy of your email address and do everything you possibly can _outside_ MythArchive. Otherwise you'll process for hours and quite often crash out with an unhelpful error report and no obvious way of picking up the pieces. Usually the report just notes a failure. I tried to alter this situation in the patch I submitted some time ago, but that wasn't created with a version control tool and adapting it for 0.24 looks as if it will take some effort.

[Larry Johnson <keepitsimpleengineer@…>]

I ran ffmpeg -i on the original recording after uploading it to another computer (mythtv client, ubuntu 10.10). The result is attached.

[J.Pilk@…]

I don't recognise all the ffmpeg preamble but it does look as if it's only finding two streams. Are your subtitles really separately encoded, so that you can enable/disable them, or are they just part of the picture - ie, overlaid at source? I wouldn't expect MythArchive to crash just because it didn't find a subtitle stream, anyway. I do wonder if ProjectX is applying a 0.24 cutlist as if it was 0.23; I'm finding some confusion about the initial cut-out and cut-in.

[Larry Johnson <keepitsimpleengineer@…>]

Looking at the progress log:

2010-12-14 10:48:24 ffmpeg -threads 2 -v 1 -i "/pvr/mythtv/recordings/1062_20101104040000.mpg" -r ntsc -target dvd -b 4771k -s 720x480 -acodec copy -copyts -aspect 4:3 "/home/mytharchive/temp/work/1/newfile2.mpg" -map 0:0 -map 0:1

Which processes the recording. Shouldn't it have the option "-scodec copy" to force subtitle copy stream?

[J.Pilk@…]

As I suggested in Comment 10, the ProjectX output is from a few GOPs at the start of the recording and a few at the end - presumably not what was intended; so the translation from Myth cutlist to ProjectX cutlist needs to be revised in Mythburn.py, too. mythcutprojectx uses bytecount positions, mythburn.py uses frame counts. I had one pathological recording where the bytecount approach worked and the framecount didn't, but it may be that framecount is usually more precise. It seems to me that the bytecount algorithm ought to cut only at keyframes, but in practice I get the impression that it's finer-grained than that. Could be, of course, that the commercial breaks are also tied to keyframes, but the Myth editor doesn't suggest that. It probably varies from one edit desk to another, anyway - not to mention international differences.

This looks to be a separate issue from the elusive subtitle stream.

[Larry Johnson <keepitsimpleengineer@…>]

Replying to J.Pilk@…:

As I suggested in Comment 10, the ProjectX output is from a few GOPs at the start of the recording and a few at the end - presumably not what was intended; so the translation from Myth cutlist to ProjectX cutlist needs to be revised in Mythburn.py, too. mythcutprojectx uses bytecount positions, mythburn.py uses frame counts. I had one pathological recording where the bytecount approach worked and the framecount didn't, but it may be that framecount is usually more precise. It seems to me that the bytecount algorithm ought to cut only at keyframes, but in practice I get the impression that it's finer-grained than that. Could be, of course, that the commercial breaks are also tied to keyframes, but the Myth editor doesn't suggest that. It probably varies from one edit desk to another, anyway - not to mention international differences.

This looks to be a separate issue from the elusive subtitle stream.

Alas. A check of the recordedprogram record for this recorded program indicates that the program has closedcaption but not subtitles. Apparently mytharchive does not support closedcaption but does support subtitles (teletext). I'm looking at ccextractor.

[J.Pilk@…]

I've opened a new ticket, #9389, to deal with the cutlist translation change; and looking at the bytecount script again has reminded me that it works with offsets from the keyframe, not the keyframe itself, so my misgivings above are probably unfounded. Bytecount is probably more robust, and just as precise.

[paulh]

MythArchive doesn't support closed captions. Changes to the way the cut list is saved in the DB has broken removing the cut list but that has a separate ticket.