ffdshow tryouts project: Discussion & Development

[clsid]

Yes, and any bug report in general.

[Matt Ashland]

Hi clsid,

I'm wondering if a simple change could get added to TffdshowDecVideo::checkAllowOutChange(...)

J. River needs (clsid==CLSID_JREncoderFilter) added to the white-list used to fill the compatibleFilter variable. This encoder supports on-the-fly format changes, just the same as a video renderer.

This is the GUID:
const CLSID CLSID_JREncoderFilter = { 0xe3eabdee, 0x2a0e, 0x4b3f, 0xa7, 0x59, 0x58, 0x5a, 0x1b, 0x62, 0xbe, 0x8a };

Thanks for all of your excellent work.

[clsid]

Added in 3971.

[Matt Ashland]

Added in 3971.

Thank you very much clsid.

[dann23]

problem with ati catalyst 11.8 and hardware acceleration.
ffdshow and mpchc don't work with dxva enabled. the same problem with wmp12 and ffdshow with dxva enabled. just a green image and audio is working (until seeking). I tried to use just wmp12 and lavsplitter. the same problem.
I tried with vlc but the same green image.
So it seems that the latest driver has gpu decoding broken.

[cca]

problem with ati catalyst 11.8 and hardware acceleration.
ffdshow and mpchc don't work with dxva enabled. the same problem with wmp12 and ffdshow with dxva enabled. just a green image and audio is working (until seeking). I tried to use just wmp12 and lavsplitter. the same problem.
I tried with vlc but the same green image.
So it seems that the latest driver has gpu decoding broken.

Can't say about ffdshow or VLC since I didn't try, but DXVA works in MPC-HC with Catalyst 11.8 on my system. I have the Radeon HD 5850 GPU, which one do you have?

[dann23]

I have a Radeon 5750.

[ikarad]

Added in 3971.
Deband filter doesn't work with ffdshow clsid 3966
bug tracker
https://sourceforge.net/tracker/?func=detail&aid=3395695&group_id=173941&atid=867360

With 3914 deband works

[ikarad]

I find a bug with Hi10p mkv video with ffdshow clsid 3966
image here
http://img838.imageshack.us/img838/9944/sansreewo.jpg (http://imageshack.us/photo/my-images/838/sansreewo.jpg/)

video here (bug at 12"37)
http://www.nyaa.eu/?page=torrentinfo&tid=236877

bug tracker
https://sourceforge.net/tracker/?func=detail&aid=3396461&group_id=173941&atid=867360

[clsid]

I can't fix that.

[clsid]

Deband filter doesn't work with ffdshow clsid 3966
bug tracker
https://sourceforge.net/tracker/?func=detail&aid=3395695&group_id=173941&atid=867360

With 3914 deband works
3943 broke it, but I don't know why yet.

[ikarad]

I can't fix that.

Who can fix it?

[ikarad]

3943 broke it, but I don't know why yet.

I hope a fix soon because since the support of hi10p, deband doesn't work.

[clsid]

The deband problem was not caused by a code bug, but it is caused by a bug in GCC. It is broken with GCC 4.6.1 and it works with GCC 4.5.3.

[ikarad]

The deband problem was not caused by a code bug, but it is caused by a bug in GCC. It is broken with GCC 4.6.1 and it works with GCC 4.5.3.

Why don't use anymore GCC4.5.3 for this moment before problem is corrected?

[Midzuki]

r3974

libavcodec's VC-1 decoder still crashes VirtualDub, MPC, WMP, whatever, when it meets a stream containing DisplayWidth and DisplayHeight flags:

( http://forum.doom9.org/showthread.php?t=162221 )

-.-

[Midzuki]

^ According to the changelogs @ xvidvideo.ru (http://www.xvidvideo.ru/chlog/ffdshow.html), it seems one did update vc1.c, BUT simply "forgot" to update vc1dec.c :rolleyes:

[LigH]

@ ikarad:

Looking at the image, this looks like a data error in the video stream; unfortunately I am not able to analyze the video stream to point at "that byte is wrong"...

If you encoded the file yourself, don't trust your harddisk or controller anymore; if you got it from elsewhere, don't trust this source anymore.

[Reimar]

^ According to the changelogs @ xvidvideo.ru (http://www.xvidvideo.ru/chlog/ffdshow.html), it seems one did update vc1.c, BUT simply "forgot" to update vc1dec.c :rolleyes:

That is a sync from Libav, not FFmpeg. Libav only contains the aspect fix, not the crash fix (so far at least).

[Wile-E-Coyote]

Hello. I could use someone's opinion about something that bothers which I first reported on madvr's thread here (http://forum.doom9.org/showthread.php?p=1521073#post1521073), but which in the end seems to be cause by ffdshow. (I play DVDs with MPC-HC / LAV Splitter / ffdshow / madVR exclusive mode)

To sum up : I use yadif with its "double framerate" option to deinterlace PAL interlaced DVDs. However judging by the pin media type information ffdshow reports outputting 25fps (AvgTimePerFrame=400000) instead of 50 (AvgTimePerFrame=200000).

Is there any way to fix this?

[VipZ]

clsid, with the change to pthreads, x64 builds don't compile with alexins 4.6.1 gcc toolchain.

Thanks

[clsid]

Those changes are required to compile with 4.5.3.
4.6.1 gives broken deband.

[VipZ]

Those changes are required to compile with 4.5.3.
4.6.1 gives broken deband.

If I revert that change locally, and with future modifications to libav. Should I expect the compiles to be unstable in anyway or just as before with deband not working?

[Eliminateur]

is Sandy Bridge DXVA operational in ffdshow?

[clsid]

If I revert that change locally, and with future modifications to libav. Should I expect the compiles to be unstable in anyway or just as before with deband not working?You only need to revert the prefix changes in makefile_c.inc and config.h

[ikarad]

You only need to revert the prefix changes in makefile_c.inc and config.h

CLSid, where can I donwload ffdshow 3975 where you have corrected problem with deband?

On xvidvideo.ru last version is 3972 and on official ffdshow site, last version is 3974.

[Midzuki]

@ ikarad: The latest source-code is r3978 already :–/

[ikarad]

@ ikarad: The latest source-code is r3978 already :–/

Yes but the last ffdshow version downloadabled is 3974

[Midzuki]

Yes but the last ffdshow version downloadabled is 3974

Translation: your crystal ball said, r3975 will rock and r3978 will suck :)

[nevcairiel]

by the way, if a popular and common compiler like GCC can break your code, then your code is broken in the first place. :)

[ikarad]

Translation: your crystal ball said, r3975 will rock and r3978 will suck :)

No my sentence said: I have only need r3975 because it's the version that correct problem with deband. After if clsid release r3978 no problem but I have for the moment only need r3975.

[ikarad]

by the way, if a popular and common compiler like GCC can break your code, then your code is broken in the first place. :)

Maybe but it's the only filter to make deband available in ffdshow and I can't create new deband filter for ffdshow

[Midzuki]

For the impatient :rolleyes: , r3978 can be downloaded from here (http://this-file.has-been-deleted.ne.jp/404.phtml). I don't recommend the build by Dirk Paehl (http://www.paehl.com/open_source/?Welcome), because he forgot to include an apparently-necessary libpthread-2.dll in the package :rolleyes: , so that, for example, you cannot use his ffdshow for decoding VP5 or Indeo Video :rolleyes:

[ikarad]

For the impatient :rolleyes: , r3978 can be downloaded from here (https://skydrive.live.com/?cid=5ACF098E0EBAE8D5&id=5ACF098E0EBAE8D5%21147&sc=documents). I don't recommend the build by Dirk Paehl (http://www.paehl.com/open_source/?Welcome), because he forgot to include an apparently-necessary libpthread-2.dll in the package :rolleyes: , so that, for example, you cannot use his ffdshow for decoding VP5 or Indeo Video :rolleyes:
Deband is buggued with your link.

CLSID has uploaded r3978 in ffdshow site and deband works.

[Midzuki]

Deband is buggued with your link.

CLSID has uploaded r3978 in ffdshow site and deband works.

Thanks for testing and reporting, :thanks:

I will remove that suspicious :devil: build from my Skydrive folder A.S.A.P. :)

EDIT: b0rked build has been eliminated ^.^;;

[Matt Ashland]

I believe there's an ffdshow decoder bug that results in a grey screen being output with certain files. The grey screen is displayed with madVR and our own encoder, so I don't think it's a renderer or allocator problem.

A sample file is here (sorry for the large size, it's HDTV):
http://files.jriver.com/temp/jtv_ffdshow_problem.7z

It's possible this file has multiple resizes when first starting, but LAV Video, LAV CUVID, and the Microsoft MPEG decoder all work properly. The gray screen only appears when using ffdshow.

To test with our software, install Media Center (jriver.com), pick 'Red October HQ' to use madVR in Options > Video, and play the .jtv file from the package above.

Thanks for any help.

[Mike89]

This may sound like a stupid question, but is ffdshow ever going to get to the point of "it's done"? Still almost daily builds for years now and doesn't seem like there is an end in sight. I know the saying, "it's done when it's done" but man, how many more years of daily builds is it going to take? I also know the saying, "a work in progress" but seems there would be a point where the progress would just start going around in a circle.

[Mike89]

I agree with that but for example if using a codec pack (like shark 007), it updates every time a new build comes out, like a couple of times a week sometimes. Seems a bit excessive to "always be up to the latest" especially when this frequency goes on for years.

[Snowknight26]

I must be missing something. Are you replying to yourself?

Whatever the answer, why would you want ffdshow to stop being updated?

[Mike89]

There was a post in between my two. Must have got deleted for some reason. I didn't say I wanted it to stop being updated. I was saying it seems kind of excessive being updated almost daily or weekly for the last few years. I just wondered how many more years this is going to continue with this frequency as there seems to be no end in sight.

[SamuriHL]

They typically work towards a major release and I thought a while back they had one planned. But daily development isn't going to stop on an open source project. :) The key is to get more "milestone" releases out there that are considered stable. That hasn't been happening.

[Midzuki]

I must be missing something. Are you replying to yourself?

There was a post in between my two. Must have got deleted for some reason.

Yes, that was a reply from Stephen R. Savage, who, as usual, deletes his posts without any apparent good-reason :devil: :D

http://forum.doom9.org/showthread.php?p=1379292#post1379292

[Mike89]

Well I wish the makers of the codec packs would take notice (or understand) of what a "milestone" release is instead of them updating almost as fast as these builds come out. I don't think I'm the only one confused as to why there are so many releases and when one should update. The codec pack makers sure don't seem to know.

[SamuriHL]

Everyone typically wants the "latest" when dealing with software which is not always a good thing. When dealing with open source it's often difficult because there isn't a solid "release cycle". (At least, typically....being open source doesn't preclude a project from having a solid release cycle, just that typically there isn't a defined process). This makes it difficult to know when to upgrade and when not to. So called "nightlies" are just the accumulation of changes that devs commit that day. Often they aren't fully tested until end users get ahold of them and find bugs. It's not like the devs submit their changes to a review process and have a full qa test suite to run against them. :) An end user has almost no chance of finding the right balance between stable and "new features". I understand the frustration, but, that's the nature of this type of software development. Buyer beware. :D

[ryrynz]

Yeah there is no AFAIK true "stable" release when it comes to ffdshow, if you have an issue you downgrade (and submit a big report) or upgrade and hopefully that fixes it, not terribly hard. There is no real alternative to this software so just accept the state of development for what it is, as it is something for nothing and we can't really complain about that now can we? :D

[Thunderbolt8]

Yes, that was a reply from Stephen R. Savage, who, as usual, deletes his posts without any apparent good-reason :devil: :D

http://forum.doom9.org/showthread.php?p=1379292#post1379292last name says it all

[Wile-E-Coyote]

Did no one see this?

Hello. I could use someone's opinion about something that bothers which I first reported on madvr's thread here (http://forum.doom9.org/showthread.php?p=1521073#post1521073), but which in the end seems to be cause by ffdshow. (I play DVDs with MPC-HC / LAV Splitter / ffdshow / madVR exclusive mode)

To sum up : I use yadif with its "double framerate" option to deinterlace PAL interlaced DVDs. However judging by the pin media type information ffdshow reports outputting 25fps (AvgTimePerFrame=400000) instead of 50 (AvgTimePerFrame=200000).

Is there any way to fix this?

[kirakami]

ffdshow r3978 20110825

bugs:
VobSub Hangs.

[mindbomb]

hi, im using r3978.

This version, in raw video mode, and it appears every version before, has a buggy RGB conversion when used with the cyberlink powerdvd 10 or 11 decoder in HAM mode.
By buggy, i mean green lines at the edges. Usually just the bottom edge, the right edge, or both.

[SledgeHammer_999]

Are there any plans to improve ASS/SSA rendering or use libass for that instead?