View Full Version : ColourLike problem
gwendolien
10th February 2013, 19:46
Using Colourlike->WriteHistogram with a dvd source runs normal.
Using it with an mkv source gives this exception:
Traceback (most recent call last):
File "avsp.pyo", line 8080, in OnMenuVideoRefresh
File "avsp.pyo", line 12397, in ShowVideoFrame
File "avsp.pyo", line 12903, in UpdateScriptAVI
File "pyavs.pyo", line 677, in __init__
File "pyavs.pyo", line 132, in __init__
File "avisynth.pyo", line 227, in Invoke
WindowsError: [Error -529697949] Windows Error 0xE06D7363
Using WritePolarHistogram with the same mkv source runs OK.
What possibly is going wrong here?
using this version:
http://wilbertdijkhof.com/mg262/Colourlike_v10.zip
IanB
10th February 2013, 22:30
0xE06D7363 is an unhandled C++ exception (throw without a matching catch, probably some std::{something} library fault).
If you run your script in VirtualDub what does it say about the error?
gwendolien
10th February 2013, 22:56
Thanks IanB.
I usually run AVsPMod with avisynth scripts.
Have loaded VDubMod and thsi reported an error
"cannot find Entrypoint CuMemcpy3DAsync of nvcuda.dll,
but I am "used" to see that...VDubMod usually runs perfect here after clicking the error message away.
Then, running the script, VDubMod(version 1.5.10.3) reports:
MS Visual C++ Runtime Library-->"This application has requested ...to end in an unusual way"
Running the script in VDub itself(version 1.9.11) I managed to save a crash report with amidst other info:
Thread call stack:
74ea812f: KERNELBASE!RaiseException [74ea0000+80db+54]
74ea812f: KERNELBASE!RaiseException [74ea0000+80db+54]
76929d45: msvcrt!malloc [76920000+9cee+57]
05962b1f: ColourLike_x!_AvisynthPluginInit2@4 [05950000+7460+b6bf]
0595fa13: ColourLike_x!_AvisynthPluginInit2@4 [05950000+7460+85b3]
05954d0d: ColourLike_x!00004d0d
04b11301: AviSynth!00011301
04b036d6: AviSynth!000036d6
04b03a91: AviSynth!00003a91
04b01166: AviSynth!00001166
To me this is abacadabra, I'm afraid...
IanB
10th February 2013, 23:11
76929d45: msvcrt!malloc [76920000+9cee+57] will throw std::bad_alloc (http://www.cplusplus.com/reference/new/bad_alloc/) when there is no more memory available.
You probably have a memory leak in the tools you use to read your .mkv file (ffmpegsource ???), try a less buggy version (something less than r683).
gwendolien
11th February 2013, 09:50
Thanks, both IanB and rean!
Now you both mention a memory leak, it strikes me that I succesfully run the WriteHistogram earlier with this clip, but probably(bad memory here...!) on an avi-format version of this clip...
So I'll try this approach first before indulging in solving/preventing this memory leak problem as such.
I'll report as soon as results allow.
Thanks again!
gwendolien
11th February 2013, 13:53
Helas, the same error pops up with the use of an avi-formt for the clip.
I noticed the follwing behaviour:
very soon after launching the script, the error pops up.
Then nothing seems to be going on w.r.t. avisynth activity in taskmanager.
however, when I hold the mouse over the frame-window or the corresponding frame slider
below the avisynth script window-part, the "turning circle" of "wait for results" shows up
AND
when I click in either of these two window sections, a short burst of Avisynt-activity is noticed in taskmanager
and immediately the error pops up again...
How do I know what "background" resource are used by Avisynth when running the script?
How to tackle the memory leak problem, is unclear to me.
gwendolien
11th February 2013, 16:43
MOST PROBABLY SOMETHING ELSE:
Tracing back all changes applied since the last succesfull run of WriteHistogram on dvd as well as mkv clips,
I realised I changed the format from 720x576 to some small portion,
using Crop(34,0,74,28) and that's the culprit!
For some reason or other, using WriteHistogram with the full image is OK on both dvd and mkv.
But both fail when I revert to the small size cut of those clips...
Very peculiar but the error is consistently popping up,
whether I use dvd(mpeg2), mkv or xvid-avi clipformats, combined with the small Cropped version, not with the full frame version.
Anything with a smaller width than 100 invokes the error...
For the moment I can live with 100 as width, but it remains a mystery to me what is actually the problem behind this all...
Gavino
11th February 2013, 17:43
using Crop(34,0,74,28) and that's the culprit!
For some reason or other, using WriteHistogram with the full image is OK on both dvd and mkv.
But both fail when I revert to the small size cut of those clips...
Some (badly-written) plugins fail when there is a Crop() in the input filter chain, and it's possible that is what is happening here.
Try using Crop(..., align=true) and see if it makes any difference.
gwendolien
11th February 2013, 21:07
THANKS Gavino, I think your suggestion brought me to something:
when I use align = true, everything runs fine, at least sometimes...
But: when I use crop values to multiples of four without the align=true always....
So it looks like multiples of 4/no multiples of 4 are the real culprit in the WriteHistogram routine
THANKS again!
-----------------------------------------------------
There is nothing in such perfect harmony with itself as a "contradictio in terminis".
vBulletin® v3.8.11, Copyright ©2000-2025, vBulletin Solutions Inc.