Any serious bugs left in 2.5.1?

[sh0dan]

Just put up a new binary in the CVS bin site - mainly a few "minor" bugfixes.

Are there any internal bugs I forgot along the way - now is the time - otherwise this is soon to be 2.5.2?

If you find bugs - or have any major bugs I have forgotten about, please post here.

[Wilbert]

With the old binary I had the following problem: When pressing F5 in the script editor of VdubMod (1.4.x and 1.5.x) it always crashes (I guess this happened after you included Belgabor's changes, but I didn't have time to test this properly). I don't know whether this is a problem of VdubMod or AviSynth, nor whether it is present using the newest binary. Maybe you can have a look at it.

The Trim bug is still present (?): "Trim(0,-1)" gives two frames. I will sent WarpEnterprises the changed docs.

[sh0dan]

Oh yeah - the trim bug - thanks!

The April 30th had an unload bug, but beside that one i also experience occasional crashes when refreshing vdubmod - the debugger also reports "Heap modified" errors in the vdubmod code - but since I am not 100% uptodate on vdubmod I figured it had been fixed - I'll dl a new version and try it out!

[FuPP]

Still a problem with addborders see here (http://forum.doom9.org/showthread.php?threadid=50534&highlight=FuPP)
FuPP

[neily]

shOdan,

A couple of minor points. As in RC1, in DirectShowSource, seek is not recognised as a parameter as it was in 2.08. Maybe just a documentation error. IsYV12 is not recognised as a script property. Trim(0,-x) returns frames 0 to x rather than 0 to x-1 as it does in 2.08. I don't know whether this is a "feature", but it should at least be consistent with 2.0x.

On the plus side, I am greatly relieved. Having spent a large amount of time trying to update AVSGenie to be 2.5 compatible, I had virtually given up. It transpired that depending on the codec, the video display component that I depended on which is based on VFW would through errors if when using YUY2 and YV12 the size was not a multiple of 16 or 32. It works fine again in RC2.

Just out of interest, what has changed such that this is the case? And any progress on a souped up ShowFrameNumber?

[PowerMacG4]

I'm not sure what version... but one of the newer builds broke eDeen. Memory access errors, etc.

[sh0dan]

@neily: I'll get the issues fixed before a release!

@PowerMacG4: Could you post a script, and the latest version it works in?

[Wilbert]

With the old binary I had the following problem: When pressing F5 in the script editor of VdubMod (1.4.x and 1.5.x) it always crashes (I guess this happened after you included Belgabor's changes, but I didn't have time to test this properly). I don't know whether this is a problem of VdubMod or AviSynth, nor whether it is present using the newest binary.
I tried the latest binaries, and it works again :)

[neily]

Hi,

There seems to be a little problem with ImageWriter. Image sequences always seem to be expected to be zero-based. If you have a series of images called image5.bmp to image9.bmp and you pass:

ImageReader("c:\image%.1d.bmp,5,9,25)

you get a clip of 10 frames duration with frames 0 to 4 blank.

In fact the length of the clip is always the value of 'end' + 1. The value of 'start' only seems to determine which images are processed and not left blank as opposed to representing the first image to be processed.

The ImageSequence plugin does handle this properly.

[Richard Berg]

I think you've described ImageReader accurately, but I didn't catch what you think should be different. In particular, what do you propose happen in your example if/when frames 0-4 are requested? A blank frame seemed appropriate enough...

[sh0dan]

Wouldn't a length of five frames be the best way?

When dealing with 3D renderings, frame numbers are often starting at arbitrary numbers, because the 3D app. spits out frame numbers based on where in the animation it is.

[PowerMacG4]

@sh0dan
Any usage of MarcFD's eDeen in any script will cause a crash in 1-300 frames of playing. I have no idea what version it worked successfully, because I have never gotten it to work since I discovered it quite a long time after Avisynth 2.5 alpha was introduced. I would imagine that a build dating to before or around the time of the release of eDeen would be more successful, since I have read old threads about how it is such a good spacial filter. Also, eDeen does successfully filter the image without artifacts during the frames in which it does not crash. I have noticed that it requires mod16 input, however.

Sorry for not being of much help, but basically any script with eDeen in it will not work.

[Richard Berg]

Thanks for the example, sh0dan. I figured in cases where start > 0 that people would want requested frame #'s to map to delivered frame #'s. If that's not the typical user scenario, the change is easy enough to make.

[RB]

Mode and Scenechange paramaters are swapped in TemporalSoften. That is, when I try to use

TemporalSoften(4,8,8,2,10)

I get an error message saying "TemporalSoften: mode must be 1 or 2".
The docs say that Scenechange is the last and Mode the last but one parameter. TemporalSoften(4,8,8,10,2) works.

[Wilbert]

I changed that, thanks!

[neily]

Richard,

I think I agree with shOdan about ImageReader and that it should work like ImageSequence. The first frame rendered should be denoted by the start value. Sure, it is no great hassle to trim the unwanted frames, but this shouldn't be neccessary.

One little additional parameter may be useful, and that is a boolean to indicate whether frames missing from the sequence are rendered or not.

[Richard Berg]

Good points.

My remaining concern would be keeping compatibility with ImageWriter. That is, if the client app requests frame 6-10 when writing, it will probably be set up to do the same when reading, which would fail unless reconfigured to request 0-4. This stems from the fact that currently both filters treat the start/end parameters as helpers only -- it's the client app that's in real control. This style seems to me more in tune with the way the rest of Avisynth functions.

One little additional parameter may be useful, and that is a boolean to indicate whether frames missing from the sequence are rendered or not.
How would this work? When a missing file is encountered, would it return a blank frame, the previous frame, or what? I do agree that the current method of throwing an exception isn't optimal.

[WarpEnterprises]

(as I saw in another thread) there is no script function IsYV12 so far.

in script.h it's there but not in script.c

[kilg0r3]

Another TemporalSoften thingy
This combonation gives an acces violation error
Movie = TemporalSoften(Movie,x,y,z)
Movie = TemporalSoften(Movie,x,y,z)

All da best

[sh0dan]

May 14 released.

- Fixed some compiler optimization errors with temporalsoften (with a speed penalty) :(
- Build date is now (semi)automatically updated in version.
- Trim now returns (x-1) frames as supposed.
- Added script functions: IsYV12(clip) IsPlanar(clip) IsInterleaved(clip)


Are the worst issues fixed?

[neily]

Hi,

Temporal soften fixed. Trim count fixed.

IsYV12 and IsInterleaved not recognised.

Is "seek" no longer a parameter of DirectShowSource?

I have myself put up a new version of AVSGenie which you may wish to try. See my earlier thread.

[sh0dan]

Embarrasing - try redownloading ;)

[mf]

When using mfToon (http://forum.doom9.org/showthread.php?s=&threadid=53364) and Convolution3DYV12 together, AVS catches an access violation. This doesn't happen with the sourceforge version. Confirmed by hikke.

[sh0dan]

I give you one little bugfix, and you expect me to help you with ALL your problems?

Just kidding. :)

I'll try it out - just have to set up a Visual Studio at home first. Thanks for the reports!

[mf]

Originally posted by sh0dan
I give you one little bugfix, and you expect me to help you with ALL your problems?
lol
Well--- I could use some money too... :p

[JuanC]

I would like to know if any one else is having problems to work with MPEG1 clips using the latest avisynth builds and any VFAPI plugin.

Or could I better ask: Has any1 been able to work with MPEG1 clips using latest avisynth builds and any VFAPI plugin without problems?

For details on my problems take a look at my posts on this thread: http://forum.doom9.org/showthread.php?s=&threadid=52075

:J

[sh0dan]

@neily: You should try out the latest version - I experimentally made it possible to enable seeking - I do however not have any files that support seeking.

@JuanC: Could you post an example script?? I use a VFAPI AVI wrapper when including tmpg projects - and then I only use AVISource, without any problems.

[DDogg]

Did you ever take a look at the problem (I think) with Tex style quotes? I did a bug report in SF on it.

BTW, Juan's examples are in that link he gave.

Edit: Darn! I see now where you declared it not a bug. I still do not understand that, but I certainly accept your viewpoint.

[neily]

shOdan

IsYV12 and IsInterleaved now recognised.

Seek now seems in order too. I do agree that DirectShowSource never seems ideal. I always tend to use DVD2AVI for MPEG2 and frameserving from VDub for MPEG1.

ImageReader now also starts at the "begin" frame, though I can still see occasions where the ability to leave out absent frames from the sequence rather than rendering them as blank frames might be handy.

Good work!

[JuanC]

Originally posted by sh0dan
@JuanC: Could you post an example script?? I use a VFAPI AVI wrapper when including tmpg projects - and then I only use AVISource, without any problems. Sample scripts from my posts in this thread: MPEG1DEC does it exist? (http://forum.doom9.org/showthread.php?s=&threadid=52075). (Please see for additional details)LoadVFAPIPlugin("x:\video\TMPGEnc\TMPGEnc.vfp", "TMPGEnc")
TMPGEnc("x:\VIDEOS\Coldplay - The Scientist.tpr")
FlipVertical()LoadVFAPIPlugin("x:\video\m2v_vfp\m2v.vfp", "M2VVFP")
M2VVFP("F:\My shared folder\Coldplay - The Scientist.MPEG")
FlipVertical()
Reverse()Anyway, I will try the VFAPI wrapper... Thanks, :J