On Mar 19, 2011, at 11:47 PM, Chris Dos wrote:
> On 11/15/2010 09:40 PM, Jarod Wilson wrote:
>> On Nov 15, 2010, at 8:51 AM, Chris Dos wrote:
>>> On 11/14/2010 08:46 PM, Jarod Wilson wrote:
>>>> On Sun, Nov 14, 2010 at 1:36 AM, Chris Dos<chris@...> wrote:
>>>>> I have a Hauppauge PVR-350 card. �I've set up a stock 2.6.36 and lirc-0.9.0-pre1 source (tried cvs as well).
>>>>> When trying to load the lirc_i2c module, dmesg shows:
>>>>> Nov 13 23:17:25 leto kernel: lirc_i2c: disagrees about version of symbol lirc_register_driver
>>>>> Nov 13 23:17:25 leto kernel: lirc_i2c: Unknown symbol lirc_register_driver (err -22)
>>>>> The card and LIRC worked fine with 2.6.34 and the CVS I pulled a back then.
>>>>> I'm at a loss on how to fix this now.
>>>> I believe you've built lirc_i2c against lirc_dev in the lirc git tree,
>>>> and are trying to load it with your kernel-provided lirc_dev, thus the
>>>> symbol version mismatch.
>>> Would it be best then to disable the kernel-provided lirc_dev and using everything from the lirc source? I
>>> though the v9 of lirc was going to use what was in the kernel and compile the other modules against the
>>> kernel. I may be totally off my rocker though.
>> Hm. Building against 2.6.36 headers should have resulted in an lirc_i2c
>> that worked with the kernel-provided lirc_dev. At least, that was the
>> intention. Could certainly try replacing the kernel's lirc_dev as well,
>> and that *should* work, but I do wonder what's amiss there. Another item
>> for ye olde TODO liste... :\
> Just wanted to let you know that this bug still exists in the latest git and 2.6.38. Removing the kernels
> lirc_dev module
> allows it to work.
Finally looked at this again for a moment. I think its due to
drivers/Makefile.common having LIRC_DEVDIR added to your include
path, and the lirc_dev.h in there is getting picked over the one
from the kernel. That's my theory, anyway. Not sure yet how to
fix it either, if that's what it is...