After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 449483 - [v4l2src] doesn't work anymore with BT878
[v4l2src] doesn't work anymore with BT878
Status: RESOLVED FIXED
Product: GStreamer
Classification: Platform
Component: gst-plugins-good
0.10.6
Other Linux
: Normal critical
: NONE
Assigned To: GStreamer Maintainers
GStreamer Maintainers
Depends on:
Blocks:
 
 
Reported: 2007-06-20 14:23 UTC by Jerome Alet
Modified: 2009-01-09 22:57 UTC
See Also:
GNOME target: ---
GNOME version: ---


Attachments
Gzipped log file with GST_DEBUG=4 (73.67 KB, application/octet-stream)
2007-06-20 14:25 UTC, Jerome Alet
Details

Description Jerome Alet 2007-06-20 14:23:09 UTC
I use both Osprey 440 and Osprey 230 which contain BT878 chips.
Pipelines which used to work fine don't work anymore after I've upgraded my Debian box to plugins-good-0.10.6

This one works fine :

  $ gst-launch v4lsrc device=/dev/video3 ! video/x-raw-yuv,width=720,height=576 \
               ! queue \
               ! ffmpegcolorspace \
               ! queue \
               ! ximagesink

But if I replace "v4lsrc" with "v4l2src" above, it doesn't work anymore. See the attached trace file I'll upload in a few minutes with GST_DEBUG=4, as asked by Wingo on IRC.
Comment 1 Jerome Alet 2007-06-20 14:25:26 UTC
Created attachment 90338 [details]
Gzipped log file with GST_DEBUG=4

Gzipped trace produced with v4l2src instead of v4lsrc in the above pipeline.
Comment 2 Andy Wingo 2007-06-20 14:37:25 UTC
Thanks. It looks like that for the linux < 2.6.19 case in v4l2src_calls.c:gst_v4l2src_probe_caps_for_format, that we need to include the code from v4l2src_calls.c:gst_v4l2src_get_size_limits.

Also it seems that for the < 2.6.19 case we need to do something smarter with framerates; it's likely that supporting webcams with older kernels isn't worth it, but that we have the necessary ioctls to support tv cards.
Comment 3 Andy Wingo 2007-06-20 14:50:13 UTC
Jerome: What framerate does your v4lsrc negotiate to? What is the expected framerate?
Comment 4 Jerome Alet 2007-06-20 16:09:26 UTC
I can't check with ximagesink at the moment because I'm currently at home, but through ssh I've tried the same pipeline with v4lsrc and fakesink, and what was negotiated is :

basetransform gstbasetransform.c:787:gst_base_transform_setcaps:<ffmpegcsp0> Input caps were video/x-raw-yuv, format=(fourcc)I420, width=(int)720, height=(int)576, framerate=(fraction)25/1, and got final caps video/x-raw-yuv, format=(fourcc)I420, width=(int)720, height=(int)576, framerate=(fraction)25/1

So I suppose the default framerate when not specified is 25/1.

FYI these chips can do up to 924x576 @ 30fps, although I've never tried 30fps, and I remember having messages about incorrect buffer size problems when trying 924x576 with the older release of plugins-good. 

Personally I only need to use them at 176x144 @ 5fps currently, but I can do any testing you need, just ask.
Comment 5 Tim-Philipp Müller 2007-06-22 20:46:46 UTC
Does the patch in bug #450190 help by any chance?
Comment 6 Jerome Alet 2007-06-22 21:49:41 UTC
Unfortunately I'm not sure to be able to test a patch before the second half of July due to very high time constraints wrt availability of our hardware.
Comment 7 Tim-Philipp Müller 2007-09-24 15:58:02 UTC
Any chance you could re-try with current CVS of v4l2src and create a new debug log if it still doesn't work?
Comment 8 Tim-Philipp Müller 2007-11-13 15:37:05 UTC
> Thanks. It looks like that for the linux < 2.6.19 case in
> v4l2src_calls.c:gst_v4l2src_probe_caps_for_format, that we need to include the
> code from v4l2src_calls.c:gst_v4l2src_get_size_limits.

I think this is fixed in CVS now, which leaves this as far as I understand:
 
> Also it seems that for the < 2.6.19 case we need to do something smarter with
> framerates; it's likely that supporting webcams with older kernels isn't worth
> it, but that we have the necessary ioctls to support tv cards.
 

Comment 9 André Klapper 2008-11-15 17:20:58 UTC
No feedback - closing as per last comment.