Comskip 0.77 build 61: Sound processing bug solved

[erik]

Frame 938 - Black frame with brightness of 22,uniform of 186 and volume of -1
Frame 939 - Black frame with brightness of 22,uniform of 160 and volume of -1
Frame 940 - Black frame with brightness of 22,uniform of 151 and volume of -1
Frame 941 - Black frame with brightness of 22,uniform of 206 and volume of -1

The above lines are from the log where it is still decoding. The -1 shows that sound is running at least 1 frame late but that should not be a problem
Look in the log file after the lines

Code:

Black Frame List
---------------------------

There you will find the list of detected black frames. There should be no -1 in that list.
If there are -1's then I need to know your build number, its in the readme.txt of the distribution zip file.

You can set the volume level to something very high (50000) this will effectively disable sound processing.

If you don't mind I would like to receive by email a .log and a .csv file, zip them together, I can receive 10MByte emails, it will help me improve comskip.

How is the memory/page file in your system? Comskip requires a LOT of memory for a 4 hour show and it will exit when not enough memory is available

[pBS]

well,no csv is being created as it crashes before it makes any file and stops immediately,all i have left are the log and .txt....if i control-c it it will make the file,but i notice nothing shows up until after all processing is done..[csv]
attaching files from control-c stoppage before crash happens:

[erik]

pBS,

We are discussing two different problems:

1) Does the sound detection work because you get the -1 for the volume level.
To check this I need the csv file from a short show that successfully was processed by comskip and that shows the -1 in the black frame volumes.

2) Comskip does crash on your system when you try to process a large mpg file.
Your log file ends with "Resizing frame array to accomodate 22500 frames." This implies that the memory allocation for this resize failed. Please do check the size of your physical memory and the size of your page file.
GBPVR and its plugins do need an awfull amount of memory (C#) and you may simply be running out of memory.

[pBS]

ahh,yes the sound *looks like it's working...
yep,checked the physical mem and it doesn't use particularly large amounts,and changed swap file size and still crashes on same frames...i'm thinking a whacked timeline is probably the culprit..but it does seem to happen near the start of the second episode of a 2 episode show...2 hrs + 2 hrs..actually 2 seperate shows in same file...[recorded straight through tho]
sent you a prelim csv,cuz i can't seem to get any csv output when it crashes..
tried it from command prompt and what you see in the log is the last thing i saw in the window...nothing different...so guess not a buffer prob..
may be just a weird mpeg that had stutters,so i'll try some more long ones and see if i can find a common cause...
most of the other ones i had problems with i long since deleted but i'm sure there were more,have a whole drive with oldies on it and it crashed when trying to do the whole directory...
i don't think the lengths really have anything to do with it...
[just seems to happen on the long ones,probably performance related when recording]
hmm,maybe the livetv setting?
will try it with that off...
thanx for all the help..

[pBS]

I turned livetv off and it dosn't crash...
am using the newest one 77.75..
[and yes it does still crash with 77.75 with livetv on]
not a big deal cuz not many situations where it's available...
[only when recording and watching same time,not while watching live tv buffer]
anyone using DirMon to kick it off on live tv buffer?
thanx for the help..
Paul

[erik]

Could you mail me a .csv file of a show that crashes with liveTV on but comskipped without live_tv?

[pBS]

on it's way..
i'm glad i thought of the livetv thing,i it always was kinda buggy,but i enjoy using it for the neat factor..
hope that helps quash the bug..it really is an amazing piece of code..

[erik]

Bug found and solved, see anouncements