Re: [meta-fsl-arm][PATCH 0/2] Fixes for current state of 1.4

[Gary Thomas <samoht.yrag@gmail.com>]

On 2012-12-19 12:16, Otavio Salvador wrote:
> On Wed, Dec 19, 2012 at 5:09 PM, Gary Thomas <samoht.yrag@gmail.com> wrote:
>> On 2012-12-19 11:39, Otavio Salvador wrote:
>>>
>>> On Wed, Dec 19, 2012 at 4:36 PM, Gary Thomas <samoht.yrag@gmail.com>
>>> wrote:
>>>>
>>>> On 2012-12-19 11:26, Gary Thomas wrote:
>>>>>
>>>>>
>>>>> On 2012-12-19 11:11, Otavio Salvador wrote:
>>>>>>
>>>>>>
>>>>>> On Wed, Dec 19, 2012 at 1:52 PM, Gary Thomas <samoht.yrag@gmail.com>
>>>>>> wrote:
>>>>>>>
>>>>>>>
>>>>>>> On 2012-12-19 08:35, Otavio Salvador wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On Wed, Dec 19, 2012 at 1:22 PM, Gary Thomas <samoht.yrag@gmail.com>
>>>>>>>> wrote:
>>>>>>>> ...
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Thanks, I'm now using the yoctoproject.org repository and can see
>>>>>>>>> the
>>>>>>>>> recent changes.
>>>>>>>>>
>>>>>>>>> Sadly though, I can't build X11 drivers - full log attached.
>>>>>>>>> | In file included from ./vivante_util/vivante_common.h:69:0,
>>>>>>>>> |                  from vivante_fbdev/vivante_fbdev_driver.c:23:
>>>>>>>>> |
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> /home/local/imx6_poky/tmp/sysroots/sabrelite/usr/include/xorg/fbdevhw.h:50:23:
>>>>>>>>> note: declared here
>>>>>>>>> | vivante_fbdev/vivante_fbdev_driver.c:780:26: warning: assignment
>>>>>>>>> from
>>>>>>>>> incompatible pointer type [enabled by default]
>>>>>>>>> | vivante_fbdev/vivante_fbdev_driver.c: In function
>>>>>>>>> 'VivCloseScreen':
>>>>>>>>> | vivante_fbdev/vivante_fbdev_driver.c:826:5: warning: passing
>>>>>>>>> argument 1
>>>>>>>>> of
>>>>>>>>> 'pScreen->CloseScreen' makes pointer from integer without a cast
>>>>>>>>> [enabled
>>>>>>>>> by
>>>>>>>>> default]
>>>>>>>>> | vivante_fbdev/vivante_fbdev_driver.c:826:5: note: expected
>>>>>>>>> 'ScreenPtr'
>>>>>>>>> but
>>>>>>>>> argument is of type 'int'
>>>>>>>>> | vivante_fbdev/vivante_fbdev_driver.c:826:5: error: too many
>>>>>>>>> arguments
>>>>>>>>> to
>>>>>>>>> function 'pScreen->CloseScreen'
>>>>>>>>> | make[2]: *** [vivante_fbdev_driver.lo] Error 1
>>>>>>>>> | make[2]: *** Waiting for unfinished jobs....
>>>>>>>>> | make[2]: Leaving directory
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> `/home/local/imx6_poky/tmp/work/sabrelite-amltd-linux-gnueabi/xf86-video-imxfb-vivante/2_12.09.01-r20.1/xserver-xorg-video-imx-viv-12.09.01/src'
>>>>>>>>> | make[1]: *** [all-recursive] Error 1
>>>>>>>>> | make[1]: Leaving directory
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> `/home/local/imx6_poky/tmp/work/sabrelite-amltd-linux-gnueabi/xf86-video-imxfb-vivante/2_12.09.01-r20.1/xserver-xorg-video-imx-viv-12.09.01'
>>>>>>>>>
>>>>>>>>> Any ideas?  Maybe I've missed something else.
>>>>>>>>>
>>>>>>>>
>>>>>>>> It seems xf86 api changes; you need to check the change in API and
>>>>>>>> prepare a patch to the driver ... otherwise I can do it ...
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> OK, I'll look at this.  I had expected this to already work, based on
>>>>>>> the earlier comments, but no problem.
>>>>>>
>>>>>>
>>>>>>
>>>>>> I've researched how to make it in a flexible way and I found
>>>>>>
>>>>>>
>>>>>> http://cgit.freedesktop.org/xorg/driver/xf86-video-geode/commit/?id=a83fcb6f84995dc50b8adffc27e264a663fbd606
>>>>>>
>>>>>> So it's adviced to use a compat-api.h header that allows it to be
>>>>>> usable in old and new Xorg API's.
>>>>>
>>>>>
>>>>>
>>>>> Thanks.  I found much the same and have already integrated this method.
>>>>> I'll send a patch for it soon.
>>>>
>>>>
>>>>
>>>> This may take some time.  While it does now build, it immediately dies
>>>> with a segmentation violation.
>>>
>>>
>>> You might have did a wrong replace for a macro and then pass a wrong
>>> param; another thing, when adding the patch in the commit please add:
>>>
>>> Upstream-Status: Pending
>>>
>>> So we can grep for it later and propose for merge.
>>
>>
>> I think I may need to pass this on to you - it's blowing up during the very
>> initial steps.  Here's my GDB backtrace:
>> Program received signal SIGSEGV, Segmentation fault.
>> VivScreenInit (scrnIndex=scrnIndex@entry=1930440, pScreen=pScreen@entry=0x1,
>> argc=argc@entry=2130705924,
>>      argv=<optimized out>) at vivante_fbdev/vivante_fbdev_driver.c:615
>> 615     vivante_fbdev/vivante_fbdev_driver.c: No such file or directory.
>> (gdb) dir
>> tmp/work/sabrelite-amltd-linux-gnueabi/xf86-video-imxfb-vivante/2_12.09.01-r20.1/xserver-xorg-video-imx-viv-12.09.01/src/vivante_fbdev
>> Source directories searched:
>> /home/local/imx6_poky/tmp/work/sabrelite-amltd-linux-gnueabi/xf86-video-imxfb-vivante/2_12.09.01-r20.1/xserver-xorg-video-imx-viv-12.09.01/src/vivante_fbdev:$cdir:$cwd
>> (gdb) l
>> 610         TRACE_EXIT(TRUE);
>> 611     }
>> 612
>> 613     static Bool
>> 614     VivScreenInit(int scrnIndex, ScreenPtr pScreen, int argc, char
>> **argv) {
>> 615         ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
>> 616         VivPtr fPtr = GET_VIV_PTR(pScrn);
>> 617         VisualPtr visual;
>> 618         int init_picture = 0;
>> 619         int ret, flags;
>> (gdb) bt
>> #0  VivScreenInit (scrnIndex=scrnIndex@entry=1930440,
>> pScreen=pScreen@entry=0x1, argc=argc@entry=2130705924,
>>      argv=<optimized out>) at vivante_fbdev/vivante_fbdev_driver.c:615
>> #1  0x0003689c in AddScreen (pfnInit=0x2af556f0 <VivScreenInit>, argc=1,
>> argc@entry=509968, argv=0x7efffe04,
>>      argv@entry=0x6df14 <xf86SetDGAMode>) at dispatch.c:3830
>> #2  0x000755fc in InitOutput (pScreenInfo=0x1aa870 <nmajor>,
>> pScreenInfo@entry=0x1b5c88 <screenInfo>, argc=509968,
>>      argc@entry=1, argv=0x6df14 <xf86SetDGAMode>, argv@entry=0x25fec
>> <main+596>) at xf86Init.c:882
>> #3  0x00025fec in main (argc=1, argv=0x25fec <main+596>, envp=<optimized
>> out>) at main.c:204
>>
>> Looks like there are more than the [cosmetic] changes I made for
>> CloseScreen().
>>
>> My preliminary patch is attached, maybe you can help.
>>
>
> The xf86ScreenToScrn(pScrn) is the macro to use in place of
> xf86Screens[pScrn->myNum]; you need to convert all code to use those
> macros so it does the trick for you.

That helped and along with a few other changes, I now have the X server
running again.  Now I have to figure out why my mouse isn't working with
the latest kernel.

Once I get the whole X system going, I'll send an updated patch series.
Most likely tomorrow :-)