[ECFT] drm/dri/mesa/xorg-server update [Part 1]

[Gustau Pérez i Querol]

Al 14/03/2011 10:06, En/na Kostik Belousov ha escrit:
> On Mon, Mar 14, 2011 at 10:25:21AM +0300, Ruslan Mahmatkhanov wrote:
>> 14.03.2011 00:03, George Liaskos пишет:
>>>> The crash is probably happening the first time something tries to access
>>>> dri, libdrm.
>>> I believe that kwin tries to use desktop effects by default. Better
>>> try a simple xinit for start to see how it goes and post the log.
>>>
>>> Ok, after some research about the patch i believe that this how it should
>>> be :
>>>
>> Tried both this patch and previous one. It compiles fine, but still doesn't
>> working. There is the messages:
>>
>> Xorg.0.log:
>> [    22.760] (II) intel(0): I830CheckAvailableMemory: 4026368 kB available
>> [    22.760] drmOpenDevice: node name is /dev/dri/card0
>> [    22.761] Failed to change owner or group for file /dev/dri! 2: No such
>> file or directory
>> [    22.761] Failed to change owner or group for file /dev/dri/card0! 2: No
>> such file or directory
>> [    22.761] drmOpenDevice: open result is -1, (No such file or directory)
>> [    22.761] Failed to change owner or group for file /dev/dri/card0! 2: No
>> such file or directory
>> [    22.761] drmOpenDevice: open result is -1, (No such file or directory)
>> [    22.761] drmOpenDevice: Open failed
> This is only an excerpt from the full log, and it is due to slightly weird
> open procedure for drm node. Either post the whole log, or look at
> the log down, where driver should say that drm opened and dri initialized.
>
>> gdm/:0.log
>> /libexec/ld-elf.so.1: /usr/local/lib/xorg/modules/drivers/intel_drv.so:
>> Undefined symbol "dixRequestPrivate"
> This looks more serious and might be the actual cause of failed startup.

   I made a simple patch which uses dixRegisterPrivateKey in place of 
dixRequestPrivate (i found somewhere the dixRequestPrivate got 
deprecated). It made it compile and X tries to start. The problem then 
is that X sigfaults because an assertion failed.

   This evening (UTC+1) it will the full log, because I think the 
assertion occurs because the way it reserves the private key is incorrect.

    However, seeing the amount of changes with xf86-video-intel (because 
of changes with xorg-server), I think it would be very difficult to get 
a working version of the old xf86-video-intel{27|29} drivers. I think it 
would be better to wait until gem/kms works (by the way, can we help 
with its development ?)

    Best regards,

    Gus

-- 
Gustau Pérez i Querol
Tècnic de sistemes.
Departament d'Enginyeria Telemàtica
Universitat Polítècnica de Catalunya
(34) 934 010 986

PGP KEY : http://www-entel.upc.edu/gus/gus.asc
Stop top-posting :	http://ca.wikipedia.org/wiki/Top-posting
			http://en.wikipedia.org/wiki/Posting_style