[Gta04-owner] v3.16-rc6

[Neil Jerram]

On 2014-07-23 12:57, Dr. H. Nikolaus Schaller wrote:
> Hi,

Hi!

> Am 23.07.2014 um 13:30 schrieb Neil Jerram:
> 
>> On 2014-07-23 11:17, Christ van Willegen wrote:
>>> On Wed, Jul 23, 2014 at 12:02 PM, Neil Jerram 
>>> <neil at ossau.homelinux.net> wrote:
>>>> On 2014-07-23 10:40, Belisko Marek wrote:
>>>>> Some time ago I did fix to avoid segfauls in X. Is this fix is
>>>>> included in your kernel:
>>>>> http://projects.goldelico.com/p/gta04-kernel/issues/586/
>>>> Well, that issue doesn't point to a specific patch or commit, so I'm 
>>>> not
>>>> sure.  But I'm using the Goldelico kernel source, so I presume I 
>>>> already
>>>> have that fix.
>>> Commit is (probably) 303e4697e762dc92a40405f4e4b8
>> 
>> That's the commit that _caused_ the problem.  What I meant is that the 
>> issue doesn't point to the commit that _fixes_ the problem.
>> 
>> But in any case, as I wrote before:
>> - I'm using the Goldelico 3.16-rc5 source, which surely should include 
>> whatever the fix is.
>> - I'm seeing a different problem from what that issue describes.
> 
> Well, the problem might be related but is not exactly the same.
> 
> Without the commit, X.org does not even start. This is because it has
> some built-in assumptions
> about the /sysfs that is to be used.
> 
> Since we have a 3.16 kernel now, these assumptions do no longer hold.
> And without recompiling
> the Xserver binary, it would not match. Therefore the patch to make it 
> start.
> 
> But there may be other problems based on the same reason. But not
> solved by the patch.
> 
> On the other hand the same Xorg installation worked fine up to 3.13 or
> 3.14 kernel if compiled
> without device tree and old style board file (since then we didn't
> test any more).
> 
> Or there is a (new) bug/incompatibility in the fbdev driver.

Many thanks for your thoughts.  The first thing I plan to do, probably 
this evening, is to install the -dbg packages for xserver-xorg-*.  
Hopefully I'll then get a meaningful backtrace in Xorg.0.log, which I 
can search for and/or report.

As another option (and especially if the backtrace indicates a driver 
problem), I could try the -omap driver instead of -fbdev, and I believe 
that would also require enabling CONFIG_DRM_OMAP in the kernel build.  
Is that worth a try (with 3.16-rc5/6), or are there reasons why that 
can't possibly work on GTA04?

Regards,
      Neil