Re: Gene's pcHDTV 3000 analog problem

From: Michael Krufky
Date: Wed Nov 30 2005 - 22:59:41 EST


Don Koch wrote:

On Wed, 30 Nov 2005 00:29:23 -0500
Michael Krufky wrote:



Gene Heskett wrote:



On Tuesday 29 November 2005 20:26, Michael Krufky wrote:

[...]



ll I can think of doing next is to have Gene, Don or Perry do a
bisection test on our cvs repo.... checking out different cvs revisions
until we can narrow it down to the day the problem patch was applied.

::sigh::




A sigh? More like an 'oh fudge' or whatever your fav expletive deleted
is...



Who wants to do it? I'll give you detailed instructions if you're
willing.




Can you farm it out, one set of patches to each of us? Or do you want
to setup a seperate cvs tree based on the v4l code in 2.6.14.3, and
incrementally patch it as we each report its still ok, until it breaks
again? I think I'd prefer the latter so we are all near the same
page even if it takes 3x longer to arrive at the answer. How many
actual patches in terms of dependency groups are there? I know, I
coulda went all week without asking that :(



Actually, cvs has a parameter that lets you specify cutoff dates...

This is what I am suggesting that you do... Base this on my previous cvs instructions....

reminder: http://linuxtv.org/v4lwiki/index.php/How_to_build_from_CVS

so....

1st:

cvs -d :pserver:anonymous@xxxxxxxxxxxxxxx:/cvs/video4linux login
cvs -d :pserver:anonymous@xxxxxxxxxxxxxxx:/cvs/video4linux co v4l-dvb
cd v4l-dvb
make clean
make
make install

test

(you already did this - you said doesnt work)


[...]


cvs up -D 2005-10-15
make clean
make
make install

doesnt work? 1 week earlier:

cvs up -D 2005-10-07
make clean
make
make install



Let's put it this way: for me, 2005-10-10 doesn't work and anything earlier doesn't build. I've tried building against 2.6.15-rc3, 2.6.14 and 2.6.13. The card doesn't
work against the built-in 2.6.13 code, but the tuner is sending bizarre stuff
to it (channel 2 is *not* at 97.25 MHz). 2.6.14 kept spewing:

CORE IOCTL: 0xc054561d
cx88[0]: ioctl 0xc054561d (v4l2, rw, VIDIOC_G_TUNER)

all over the syslog.

Build issues include broken Makefiles (around 10-08) and missing header files.



Cheers,

Mike


hmm... would those be unresolved symbols in relation to dvb frontends?

This would be because we only recently merged the cvs trees, and the older builds dont include the dvb frontends in the v4l build. You can fix it by doing the following:

Be sure to "make clean" before and after each checkout -- it's necessary because of what the scripts below do to the build configuration.

Before running make (EACH TIME) do this:

1- edit v4l-dvb/v4l/scripts/merge-trees.sh ... delete everything BETWEEN these two lines... being sure to leave both of them intact.:

#!/bin/sh

...

patch -p2 <<'DIFF'




...do the same thing to v4l-dvb/v4l/scripts/unmerge-trees.sh

You will have to do it for each checkout, unfortunately. If you are using a more recent checkout (past few days) then you do not have to edit these files, as they have been deleted when the build was updated for the merger of v4l+dvb cvs.


2) make merge-trees

3) make

4) make install

...because of the "merge-trees" command, you'll have to run "make clean" each time before you cvs check out (cvs co) again.

I hope this works........


-Mike
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/