mesa 13.0.0 latest rebase

[Mauro Rossi]

Hi, for people interested in latest mesa 13.0.0 status.


I had not much time, but here you can find latest rebase to mesa 13.0 branch.

https://github.com/maurossi/mesa/commits/13.0.0_nougat-x86

Here follows the list of the small changes compared to current nougat-x86
(most important ones cherry-picked or inspired by android-ia github):

"android: st/mesa: fix building error in libmesa_st_mesa" (similar
patch found in android-ia)

"st/dri: remove trailing whitespace"

"Add the RGBA8888 and RGBX8888 back to Intel screen config"
(cherry-picked from android-ia, fixes some 3d black rendering issues
in Basemark ES2 and OpenGL ES Examples )

i915: add mock implementation of GL_OES_EGL_image_external (v2)  -
this was changed as per corresponding i965, now extension
GL_OES_EGL_image_external is not enabled by default in this commit...

...but it is then enabled in "HACK: i915: enable image external
sampling" - which is not intended for upstreaming to mesa-dev/stable

Last change is "android: intel/genxml: add rules to generate xml
headers (v2)" where commit message has changed according to the
purpose of adding buildingrules for future uses.


Residual issues:

1) GUI restart of swrast happening due to buffer re-registration by
SurfaceFlinger of a buffer originally created by SurfaceFlinger
itself: now avoided with Paulo's workaround in hardware_libhardware.

As a minor side effect the screen capture, which also relies on
SurfaceFlinger re-registered buffers, is not working:

11-01 21:05:48.770  1238  1238 D gralloc : Registering a buffer in the
process that created it. This may cause memory ordering problems.
11-01 21:05:48.770  1238  1238 E gralloc : Returning 1 to avoid memory
ordering problems / crashes...
11-01 21:05:48.770  1238  1238 W GraphicBufferMapper:
registerBuffer(0x705331a10510) failed 1 (Unknown error -1)
11-01 21:05:48.770  1238  1238 E GraphicBuffer: unflatten:
registerBuffer failed: Unknown error -1 (1)
11-01 21:05:48.770  1238  1238 E Surface : dequeueBuffer:
IGraphicBufferProducer::

requestBuffer failed: 1
11-01 21:05:48.770  1508  2851 W WindowManager: Screenshot failure
taking screenshot for (800x600) to layer 21025

I have no idea on how to solve this differently, are dma-buf/handles
supposed to help here or some changes in EGL swrast code?

2) some apps/demos still render black on i965, like for example most
of Rajawali demos are fine, but Sphere Mapping demo shows black
picture

3) RGBA/RBGX formats for non Intel gpus are still missing, thus
causing black rendering of images with those pixel formats

That being said, no further regressions observed in mesa 13.0.0

Mauro

[youling 257]

32 bit 6.0 r1 ，vlc player and mx player 's YUV not work，video black screen . z3735f .

no android-6.0.1_r74 mesa 13 to test .

在 2016年11月2日星期三 UTC+8上午4:29:38，Mauro Rossi写道：

[youling 257]

RBG 32，many color block，just 800x480 video 

在 2016年12月5日星期一 UTC+8上午3:08:25，youling 257写道：

[youling 257]

i don't know should use what English word，i don't even know use what chinese word . 
http://tieba.baidu.com/p/2171481723 ，http://imgsrc.baidu.com/forum/pic/item/1fc257a7d933c895f93c3058d01373f0830200bc.jpg ，it is 色块/color block ?  video background color block ?

在 2016年12月5日星期一 UTC+8上午3:14:08，youling 257写道：

[allenho...@gmail.com]

Thanks for the updates! I have an app on Android-x86 that is doing screen capture using mediaprojection. This app is not working right now due to the issue you mentioned above. Do you guys have any plan to make screen capture work again? Or is there any workaround to do screen capture? Thanks in advance!

Allen

[Mauro Rossi]

Hi,

You mean that the workaround to avoid MAPERR with swrast has the side-effect that screen capture does not work for any You driver,
correct?

Currently we have no solution yet,
But if you don't use swrast, ATM you could revert workaround commit in frameworks/native

Mauro