On Sunday 18 January 2009 19:10:16 CityK wrote:
> Hans Verkuil wrote:
> > On Friday 16 January 2009 04:20:02 CityK wrote:
> >> The "hack-fix" patch applies cleanly against Hans' sources.
> >> However, the test results are negative -- the previous workaround
> >> ("modprobe tuner -r and "modprobe tuner") fails to produce the
> >> desired result.
> >
> > If you try to run 'modprobe -r tuner' when the saa7134 module build
> > from my sources is loaded, then that should not work since saa7134
> > increases the use-count of the tuner module preventing it from
> > being unloaded.
> >
> > If you can do this, then that suggests that you are perhaps not
> > using my modified driver at all.
>
> Huh? Of course I'm using your modified driver. As a recap:
> * I tried Hans' modified sources and the test result was negative.
> I also attempted (in a similar fashion as to the steps required when
> using Mike's "hack/workaround") unloading all the modules and then
> modprobing them .... as Mike later explained, the modifications Hans
> had made are not enough to correct the issue
> * Mike then asked whether:
> (a) his "hack/workaround" still applied cleanly against Hans' source,
> and stated that, if that was not the case, then he would re-spin the
> "hack" patch ... I confirmed that Mike's "hack/workaround" did indeed
> apply cleanly against Han's source.
> (b) I was successful in getting the analogue TV working after
> applying his "hack/workaround" patch against Hans' source. However,
> as I reported, the results of this test were negative ... as Mauro
> later explained, the "hack/workaround" that Mike had spun will no
> longer work given the inherent changes introduced in Hans' code
>
> As requested by Mauro, I will provide the dmesg output, both that
> from the base case and then that given when 12c_scan is employed ...
> but that will have to occur either later today or sometime during the
> week ... at present, I have switched back to an older changeset, as I
> required having a functional second TV this weekend
>
> > BTW, I've asked Mauro to pull from my tree
> > (www.linuxtv.org/hg/~hverkuil/v4l-dvb) which contains the converted
> > saa7134 and saa6752hs drivers. It's definitely something that needs
> > to be done regardless.
>
> Err, while I agree that the changes are something that need to be
> done (I don't think anyone is in disagreement with that consensus), I
> don't think that this is a case of "regardless". In fact, I stand
> behind
>
> Mike's position:
> > Anyway, if the previous workaround works after Hans' changes, then
> > I think his changes should be merged
>
> But as I have demonstrated above, and as Mauro explained, the
> previous "hack/workaround" no longer works in the case of with the
> Hans source code. The "if" case fails! Consequently, the "else"
> case should be don't merge. Why? Because we have now gone from:
> * circa pre-2.6.25, Mauro's changes that broke the boards analog TV
> support, but which could somewhat be corrected by Mike's
> "hack/workaround" * to present, where merging Hans' code eliminates
> the usability of Mike's "hack/workaround" ... in essence, analog TV
> function has now been completely killed with these boards.
I've taken a look at Mike's workaround and that will indeed no longer
work. I suspect that the core problem is related to the
SAA7134_BOARD_KWORLD_ATSC110 case in saa7134_board_init2 in
saa7134-cards.c. There the 'tuner is enabled', whatever that means. I'm
beginning to suspect that this code should perhaps be executed before
the tuner module is loaded. Does anyone know more about what is going
on here? If my analysis is correct, then this should be executed
earlier. Reading older mails in this thread suggests that this is
indeed the case.
I've made a new tree http://linuxtv.org/hg/~hverkuil/v4l-dvb-kworld/
that calls 'enables the tuner' before loading the module. See if that
works.
Anyway, I get the feeling that saa7134_board_init2 contains different
types of initializations: some open a gate or mux to a tuner and so
should be called before loading and setting the tuner, and others
should be called after loading the tuner, but before setting up the
tuner.
> Now, if it is a case that a resolution to the problem is imminently
> forthcoming, then I don't think that the merge would be much of a
> problem. However, given the breadth of the conversation so far (and
> I really do appreciate the depth of Trent's and Jean's
> discussion/consideration on the matter), it is entirely unclear to me
> that such a resolution will be found in short order (although, I
> don't discount the possibility of it either).
It should definitely go in, if only to ensure deterministic behavior.
Please test my v4l-dvb-kworld tree. I suspect that this might fix the
bug.
Regards,
Hans
--
Hans Verkuil - video4linux developer - sponsored by TANDBERG
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
