[mythtv] [mythtv-commits] Ticket #8108: dvd playback on via epia M10k causes mythfrontend or X to segfault

[R. G. Newbury]

On 03/01/2010 02:59 PM, Michael T. Dean wrote:
> On 03/01/2010 02:24 PM, Simon Kenyon wrote:
>> MythTV wrote:
>>> #8108: dvd playback on via epia M10k causes mythfrontend or X to
>>> segfault
>>>
>>> Changes (by danielk):
>>>
>>> * status: new => closed
>>> * resolution: => invalid
>>>
>>> Comment:
>>>
>>> Stuart, that's the API. You request surfaces until the driver says no or
>>> until you have enough (32) whichever comes sooner. We already have some
>>> hacks for this driver, but at this point it doesn't make sense to add
>>> more
>>> hacks for yet another regression in a driver for a dead platform.
>>>
>> not dead for me
>
> Is this your volunteering to fix the issues with the Via XvMC and
> maintain the code for it? (Because I'm pretty sure you can't volunteer
> someone else to do so. :)
>

Mike, there is really not a lot to do to fix this. And it appears to 
affect all Via motherboards (C-3 and C-7) using XvMC. The Myth code 
requests too many buffers.

I sent some patches to Daniel Kristjansson a couple of years ago to add 
a reference variable into 'configure' for the value of 
max_num_xvmc_surfaces in ../libs/libmythtv/util-xvmc.h

The reference variable is set to 16 unless the chipset is Via, in which 
case it is set to 14.

The problem, iirc was in properly parsing and recognizing the Via 
chipsets which require the fix (although I think all of them are 
affected: it's an internal buffer thing, I think).

At any rate the processor='cat /proc/cpuinfo | grep "model name" 
test/setting, and the tests at around line 2530 in 'configure' should do 
the trick.
I can't find a copy of my email, but I can probably resurrect the content.

Geoff





-- 
         Please let me know if anything I say offends you.
          I may wish to offend you again in the future.

          Tux says: "Be regular. Eat cron flakes."