[Buildroot] [PATCH v4 00/14] Add Xorg support for iMX6

[Gary Bisson]

Jerome, All,

On 12/04/2014 05:14 AM, Jérôme Pouiller wrote:
> [snip]
>> It seems I can't build Qt5 properly now... I get the following errors;
>> In file included from ../../../../include/QtCore/qtextstream.h:1:0,
>>                    from ../../../../include/QtCore/QTextStream:1,
>>                    from qeglfsscreen.h:47,
>>                    from qeglfswindow.h:46,
>>                    from qeglfsintegration.cpp:44:
>> ../../../../include/QtCore/../../src/corelib/io/qtextstream.h:54:2:
>> error: #error qtextstream.h must be included before any header file that
>> defines Status
>>    #error qtextstream.h must be included before any header file that
>> defines Status
>> ...
>> Makefile:1860: recipe for target '.obj/qeglfsscreen.o' failed
>> make[6]: *** [.obj/qeglfsscreen.o] Error 1
>> Makefile:117: recipe for target 'sub-eglfs-make_first' failed
>> make[5]: *** [sub-eglfs-make_first] Error 2
>>
>> Could you send your entire config file? That might be easier to track
>> down the issue this way.
> hmmm... weird, I cannot reproduce it. Did you applied patch named "qt5base:
> fix Vivante x11/EGL configuration"?

Yes, here is my git log:
$ git log --oneline -15
89f26bf qmx6_defconfig: update
137294b freescale_imx6*_defconfig: bump kernel to 3.10.17_1.0.1_ga
fc6af14 nitrogen6x_defconfig: bump kernel to 3.10.17_1.0.1_ga
1ccd90a Vivante drivers: bump to version 3.10.17-1.0.1
3f47b07 qt5base: fix Vivante x11/EGL configuration
f15dbd5 qt5base: apply new patch naming policy
169008f libdrm: fix support for xserver with Vivante drivers
c88d616 xdriver_xf86-video-imx-viv: new package
c04135b gpu-viv-bin-mx6q: change output selection mechanism
f160a0f gpu-viv-bin-mx6q: make fb/x11 choice explicit
75d8516 gpu-viv-bin-mx6q: fix GLESv2 support
9f07513 gpu-viv-bin-mx6q: fix compiling issues with EGL_API_FB
f3288a5 gpu-viv-bin-mx6q: use self-extractible helper
85bc45d mesa3d: Give possibility to external backends to enable DRI/Gallium
820bd03 libz160: new package

> Just in case it would be useful, you will find my .config here:
>     https://gist.github.com/jerome-pouiller/3c5a0c5c6f3a06cd5f75

Is it normal that your configuration is using the FB version of the 
vivante libs? In my case I wanted to test everything at once (X11 libs & 
Qt5). Have you tried with the X libs?

>> Also, I'm sorry not to have said something earlier but I'm not sure
>> about your qmx6_defconfig as it includes graphics and Qt5 by default. I
>> thought the config files were meant to be as minimalistic as possible
>> for the user to decide what packages he wants to use. I'm not sure that
>> adding Qt5 by default is suitable especially since this is the only
>> defconfig doing so.
> Well, if we want to keep it as minimalist as possible, we should not enable
> gpu-viv-bin-mx6q. But, it does not make sense to enable gpu-viv-bin-mx6q if
> there no library to use it.

Yes by minimalistic I mean CPU + bootloader + kernel without any custom 
libs.

> In add, I read in Buildroot Developer Days from ELCE 2013 report[1] that BR
> will begin to adding 'demo' config.
>
> Finally I enabled Qt.
>
> I confess I have hesitated to send this patch as is, exactly for reasons you
> gave (and the more time passes the less I am convinced by my arguments).
>
>
> [1] http://elinux.org/Buildroot:DeveloperDaysELCE2013#Buildroot_Developers_Meeting.2C_26-27_October_2013.2C_Edinburgh.2C_UK

My understanding of it is to create a demo config like 
qmx6_qt5_demo_defconfig whereas a qmx6_defconfig would stay minimal but 
I might be wrong.

Thanks,
Gary