[ tvtime-Bugs-809153 ] v4l2 support with saa7134

Bugs item #809153, was opened at 2003-09-19 04:36
Message generated for change (Comment added) made by vektor
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=506987&aid=809153&group_id=64301

Category: Capture and tuner related
Group: Open bugs in driver/lib deps
>Status: Closed
>Resolution: Out of Date
Priority: 5
Submitted By: Jan Ondrej (ondrejj)
Assigned to: Billy Biggs (vektor)
Summary: v4l2 support with saa7134

Initial Comment:
tvtime 0.9.9 and higher are not working wirh my tuner
in v4l2 mode. Sound is always muted. Video is ok (very
good), bud I have no sound.
In log always appears: "Can't mute card.".

When I tryed to disable v4l2 support, anything works fine.
What is wrong?

tvtime log file (stripped):

tvtime: Running tvtime 0.9.10.
videoinput: Using video4linux2 driver 'saa7134', card
'LifeView FlyVIDEO3000' (bus PCI:00:0b.0).
videoinput: Version is 521, capabilities 5010015.
videoinput: Field 4, colorspace 0.
tvtime: V4L sampling 720 pixels per scanline.
videoinput: Can't mute card.  Post a bug report with your
videoinput: driver info to http://tvtime.net/
videoinput: Include this error: 'Invalid argument'

Disabling of v4l2 (quick hack :-) :

tvtime-0.9.10/src/videoinput.c
-    if( ioctl( vidin->grab_fd, VIDIOC_QUERYCAP,
&caps_v4l2 ) < 0 )
+    if( ioctl( vidin->grab_fd, VIDIOC_QUERYCAP,
&caps_v4l2 ) >= 0 ) 


----------------------------------------------------------------------

>Comment By: Billy Biggs (vektor)
Date: 2004-06-24 23:59

Message:
Logged In: YES 
user_id=153320

This long and complicated bug is now way too invalid to be
useful.

----------------------------------------------------------------------

Comment By: P Z (pzmt)
Date: 2003-11-14 06:51

Message:
Logged In: YES 
user_id=801959

Problem was me. I am using v4l snapshot tar and this doesn't 
contains mapping old ioctl -> new ioctl. I must use kernel 
patch (contains mapping).

sorry boring you

----------------------------------------------------------------------

Comment By: P Z (pzmt)
Date: 2003-11-14 06:51

Message:
Logged In: YES 
user_id=801959

Problem was me. I am using v4l snapshot tar and this doesn't 
contains mapping old ioctl -> new ioctl. I must use kernel 
patch (contains mapping).

sorry boring you

----------------------------------------------------------------------

Comment By: P Z (pzmt)
Date: 2003-11-14 04:04

Message:
Logged In: YES 
user_id=801959

tvtime is ok.
Problem is in driver, because driver doesn't implement these 
old ioctls (only new).

But I personaly still think, that including headers videodev.h 
and videodev2.h in source of tvtime is not good.

----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-10-24 08:44

Message:
Logged In: YES 
user_id=153320

pzmt: start by posting the diff of videodev.h and
videodev2.h between yours and the one in your copy of tvtime.

----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-10-24 08:43

Message:
Logged In: YES 
user_id=153320

videodev.h and vidodev2.h are included in tvtime because
that's where it makes sense to have them.  Features are
added to both APIs and so we have to keep up to date.  For
example, we used a flag from V4L1 about stereo support which
was not in videodev.h before kernel 2.4.20.  If we had
checked for /usr/include/linux/videodev.h and used that
instead of our own copy, we would not have compiled.  Also,
note that if you needed to recompile tvtime with a new API,
then this means the APIs are not binary compatible, and ALL
apps built for an old kernel would need to be recompiled! 
That does not make sense!  Kernel APIs cannot become binary
incompatible like that.

Check this out:
http://www.spinics.net/lists/vfl/msg07244.html

A few weeks ago, Chutt of MythTV told me that the
videodev2.h  API changed breaking binary compatibility. 
However, after posting to video4linux-list, Gerd confirmed
that it did NOT.  If binary compatibility has been broken,
then we need to post about it.  More than just tvtime will
break, xawtv also includes their own copy of videodev.h and
videodev2.h, and so does MythTV, and so does ... ... ...

----------------------------------------------------------------------

Comment By: P Z (pzmt)
Date: 2003-10-24 04:58

Message:
Logged In: YES 
user_id=801959

FLY3000FM (saa7134) - sound not working with tvtime 0.9.10
driver snapshot from bytesex.org version 20030924, v4l2-
20030826 version.

Solution - overwrite videodev.h and videode2.h in tvtime with 
includes from v4l2-20030826 package found on bytesex.org.

recompile - everythinks works ok.

Why are used videodev includes from tvtime if other exists 
(in /usr/include/linux). Should by there test for presence in 
configure ???), why include videodev.h, videodev2.h in source 
of tvtime.



----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-09-22 10:19

Message:
Logged In: YES 
user_id=153320

Glad to hear it's working. :)

If you don't mind I'm going to keep this bug open until I
document this on the website or something.  Want to make
sure nobody else gets hurt by this.

----------------------------------------------------------------------

Comment By: Jan Ondrej (ondrejj)
Date: 2003-09-22 09:59

Message:
Logged In: YES 
user_id=799485

With kernel patch it is working fine. I am sorry for bad
report. :(

I am more familiar with tarbal (kernel modul withou
patching), than kernel patches, but i must wait for final
version.

Also thanks to Gerd Knorr for help.


----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-09-22 06:39

Message:
Logged In: YES 
user_id=153320

Gerd Knorr wrote back to me on this:

"tarball or kernel patch?  The tarballs don't work currently
as noted on bytesex.org/v4l/build.html, you need a patched
kernel."

Maybe you should try one of the patches from his site instead?

----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-09-21 11:45

Message:
Logged In: YES 
user_id=153320

I have no plans to support flags that disable v4l2 to enable
v4l1 if it's just to work around a driver bug.  Let's fix
the bug, please.  It's not like the saa7134 driver is closed
source or anything.

I'll email Gerd about this bug report and see if he has any
suggestions.

----------------------------------------------------------------------

Comment By: Jan Ondrej (ondrejj)
Date: 2003-09-21 11:19

Message:
Logged In: YES 
user_id=799485

With latest saa7139-20030910 i have these unresolved modules:

[root@home saa7134-0.2.9]# depmod -ae
depmod: *** Unresolved symbols in
/lib/modules/2.4.22/v4l2/saa7134.o
depmod:         video_device_release
depmod:         video_device_alloc

IMHO, it is not a bug of my kernel or compiler. If I
understend hermann11, it is working only with kernel 2.6.0test5.

Where i can find saa7134-20030806 for download? I have all
versions from 20030813, but i don't have this. :-(

Do you plan to implement some tvtime command line options to
disable v4l2 and reenable v4l1?


----------------------------------------------------------------------

Comment By: Hermann Pitton (hermann11)
Date: 2003-09-21 09:35

Message:
Logged In: YES 
user_id=364087

The  saa7134-0.2.9 snapshot from 20030910 works with
2.6.0-test5 and the additional kraxelpatch at lastest state
(there is a new ir-common module for testing on remote
controls depending on input on 2.6x.). Due to that and 
changes in the kernel like pci names since 2.6.0-test3-bk5
and others, latest snapshots are not useable on older kernels.
The latest reported working for 2.4.22 was saa7134-20030806.
Also you can not just build the v4l2 modules at the moment.
You have to use the appropiate kernelpatch. 2.4.20 and 21
work then with saa7134-0.2.8 (there is a audio bug for
people changing between Pal B/G and Pal D/K.) 
The situation is reflected somehow at
http://bytesex.org/v4l/build.html

----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-09-21 08:48

Message:
Logged In: YES 
user_id=153320

This is really strange.  in saa7134 0.2.8, the only way
S_CTRL can return EINVAL is if it doesn't know what
V4L2_CID_AUDIO_MUTE is.  The only way that would be
possible, I think, would be if somehow you broke the module
compile or something like that.  Maybe this has to do with
the unresolved symbols you're getting with the latest
release?  Is something broken about your kernel?

----------------------------------------------------------------------

Comment By: Jan Ondrej (ondrejj)
Date: 2003-09-21 03:57

Message:
Logged In: YES 
user_id=799485

I am using kernel 2.4.22 (vanilla) with v4l2-20030826
external module (tryed also with older module),
saa7134-0.2.8 or some older 0.2.9 releases (latest 0.2.9
fails with unresolved symbols).

I think, that it is not a mute problem. Some other function
are not working too. But video is good.

----------------------------------------------------------------------

Comment By: Billy Biggs (vektor)
Date: 2003-09-20 16:55

Message:
Logged In: YES 
user_id=153320

This sucks.  What version of the saa7134 driver?  I'd like
to look through it and see under what conditions the mute
control will fail.  Pretty weird though.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=506987&aid=809153&group_id=64301


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com