Faire un don

  • R/O
  • HTTP
  • SSH

device-generic-goldfish-opengl: Listed des commits


Révision l'heure Auteur
02ffda0 multiwindow-oreo 2017-09-21 12:40:58 bohu

goldfish-opengl: update to 4153dd9ab6a14

commit 4153dd9ab6a14a263e97c40faeab237b66c9f67b
Author: bohu <>
Date: Tue Aug 1 17:05:26 2017 -0700

get emulator iamge on oc-mr1-dev work with treble

Need to remove libui dependency
otherwise, libEGL wont load because it is looking for it;
but it does not seem to need it

BUG: 64401460

Change-Id: I81a77c85307524c2b075d430a84c73556f64524e
Merged-In: I6b435afce04164308767daef3586358b04021cbd

e877c8a 2017-06-24 05:53:57 android-build-team Robot

Merge cherrypicks of [2454758, 2454859, 2454837, 2454965, 2454984, 2455001, 2454760, 2454860, 2454838, 2454899, 2454970, 2455005, 2455027, 2454973, 2455006, 2455061, 2455007, 2454914, 2454987, 2454974] into oc-release

Change-Id: I3cd8a20b7940e3541c22af7cbacb71ce6eb3d0d0

0a49f1a 2017-06-24 05:45:23 Lingfeng Yang

Fix API 26 for public emulators, take 2

Bug: 37302997
Bug: 62333501

codename changed from O to REL

So add a case for API level too

Test: Build and run sdk_google_aw_x86-userdebug emulator

Change-Id: I91f3804082bc6d61922d31ce72081a2ec88bc6b0
(cherry picked from commit 256462a4a0fb181aca670fba4241e2fe0cd295f8)

8a78b21 2017-06-24 05:45:13 Lingfeng Yang

gralloc ashmem hack to delay cb close

Tired of public emulators not working with O

See this CL for context:

This CL moves that to the guest, and detects if
each gralloc cb has been opened at least once
before allowing gralloc_free to call rcCloseColorBuffer.

Images using Treble / HIDL gralloc that have this CL
should work with older emulator versions.

Bug: 37302997

Test: Build and run sdk_google_aw_x86-userdebug emulator

Change-Id: Ice72269e4295497825050385b15dbdec2258100c
(cherry picked from commit ef6676015bfec4538a552204b51e32f02d721fb0)

dcacca1 2017-06-24 05:45:05 bohu

Emulator: update opengl to bb8d138b

Bug: 37302997

Test: Build and run sdk_google_aw_x86-userdebug emulator

To commit
Merge: 7c53152 bb1f88a
Author: Josh Gao <>
Date: Fri Apr 14 23:08:26 2017 +0000
Rename socket_loopback_server to _socket_loopback_server.

Change-Id: Id602f0c1df3be423e37af5f6c91d037d713c3b7a
(cherry picked from commit 588c0a94e384e07c8d38185710a1a968823594ba)
(cherry picked from commit 1b5d5a70fe49f2826a225878ddcd8d6ad561c886)

1ed8990 2017-03-23 16:06:30 gitbuildkicker

merge in oc-release history after reset to master

841884a 2017-03-23 13:22:40 gitbuildkicker

merge in oc-release history after reset to master

4b08be3 2017-03-23 11:40:02 Lingfeng Yang

[hwc2] Fix build

Change-Id: Ibd82639c1eb5619322e98a559f495bed36133998

936ba28 2017-03-23 09:11:11 gitbuildkicker

merge in oc-release history after reset to 18d4bde820d1b8b073c20f306f8cc1103174eaf6

596e8de 2017-03-23 03:47:28 Lingfeng Yang

[hwc2] Add surfaceInterface/goldfishHwc2

bug: 36439031
bug: 36375335

When HWC2 is used (even the HWC2on1 adapter), one of the most fundamental
changes is that acquireBuffer/releaseBuffer by BufferQueue consumers
is coarsened, delaying releaseBuffer until the app has finished
with its eglSwapBuffers operation and handling releaseBuffer as part of
a post process (mPendingRelease/releasePendingBuffer/et al).

This would be OK, except for the fact that an acquired, unreleased buffer
takes up a free buffer slot.

Emulator uses GLES composition currently, so each eglSwapBuffers by an app
directly causes another eglSwapBuffers, each of which attempts to
dequeueBuffer, taking up 2 slots right there. This was usually OK, since
releaseBuffer would be interleaved, but now, with the delayed releaseBuffer,
there are no free buffer slots and we have a deadlock situation.

Fortunately, we can set the swapped-to Surface to "async mode", which takes
exactly this situation into account. Async mode, which is for Surfaces but
really affects the BufferQueueCore queue size, makes it so that there
is an extra buffer slot so that dequeueBuffer doesn't have to block.

This CL adds a small static library for setting the swapped-to Surface
to async mode. Note that one does not simply add libgui to the shared
libraries of system/egl/, since libgui itself includes EGL/GLES2
as dependencies, putting us into DLL hell and causing all sorts of trouble!

Change-Id: I6ee4f0e6d0b668d60573887751ec6b02839df5c3

8767533 2017-03-22 08:41:22 Lingfeng Yang

Merge "Don't validate glTexSubImage2D bounds for native buffer target"

15e0295 2017-03-22 08:40:43 Lingfeng Yang

Don't validate glTexSubImage2D bounds for native buffer target

bug: 36372213

We validate glTexSubImage2D having enough width/height/depth to work
correctly, by tracking glTexImage2D / glTexStorage2D calls,
but this breaks down once we put EGL images in the mix.

Previously we skipped validation for EGL_GL_TEXTURE_2D targets, but
it looks like for O and up, we also need to skip this validation for

Change-Id: I68e6049a00e7f0927fc4097a36564b88aea1ff00

18d4bde 2017-03-09 14:15:37 Po-Chien Hsueh

Merge "Change path to gralloc.default"

f4b9be8 2017-03-03 03:14:28 bohu

DO NOT MERGE ANYWHERE goldfish-opengl: adjust to the refactored qemu_pipe am: 28494a7ff7 -s ours am: 2e4f19c1bf -s ours
am: 0ef61727d1 -s ours

Change-Id: I37c4c0dc7055526dfaf02deef8aca3483fec431c

0ef6172 2017-03-03 03:12:29 bohu

DO NOT MERGE ANYWHERE goldfish-opengl: adjust to the refactored qemu_pipe am: 28494a7ff7 -s ours
am: 2e4f19c1bf -s ours

Change-Id: Ida2024994bd063cd5d5e7f53fd8294dab4cd6f4d

2e4f19c 2017-03-03 03:09:30 bohu

DO NOT MERGE ANYWHERE goldfish-opengl: adjust to the refactored qemu_pipe
am: 28494a7ff7 -s ours

Change-Id: I289486fcde44223c2fac33213defa5ff5ee58303

2417af4 2017-03-02 17:59:14 Po-Chien Hsueh

Change path to gralloc.default

Libraries under /hardware/libhardware/modules should go to /vendor/.
This CL changes hardcoded path to

Bug: 35907904
Test: Compiled and checked install path. Also code searched to see
if there are hardcoded paths.

Change-Id: I6c304b048047053d782a0ed91abfbc3140bdd261

efe8d93 2017-03-02 12:41:48 Bo Hu

Merge "Emulator: make qemu_pipe.h self contained in goldfish-opengl project"

28494a7 2017-03-02 09:34:24 bohu

DO NOT MERGE ANYWHERE goldfish-opengl: adjust to the refactored qemu_pipe

Qemu_pipe is refactored into a library, adjust goldfish-opengl

Change-Id: I6fe6a98a962ac1bcfeb0395a575115054464d72e

df9f29d 2017-02-28 16:23:32 Lingfeng Yang

Merge "extern "C" unconditionally"

f3462c3 2017-02-28 07:11:39 Lingfeng Yang

Have single macro for taking care of goldfish_dma_context alignment.


!defined(__aarch64__) || !defined(__x86_64__) is always true,

instead of the intended "add this field if not 64 bit."


bug: 35711775
Change-Id: I46043c4402325845524d56d8d4f910f16add25cb

eb25e2c 2017-02-28 02:44:12 Lingfeng Yang

extern "C" unconditionally

This keeps older API levels from mangling GLESv3 entry points.

Change-Id: I5e751ca9f53f510d057693d368bc15c20f66e463

104a742 2017-02-25 15:53:12 bohu

Emulator: make qemu_pipe.h self contained in goldfish-opengl project

Remove the dependency on the system/qemu_pipe.h or hardware/qemu_pipe.h;
and use the legacy qemu_pipe_open definition since goldfish-opengl is
shared accross all apis.

Change-Id: I3da10066d458a57385c408cee4d2ded0eec6e456

7dcb05d 2017-02-23 02:38:02 Lingfeng Yang

Merge "Also take into account arm64 and mips64 for goldfish_dma_context aligment."

25b9f84 2017-02-22 17:58:31 Lingfeng Yang

Also take into account arm64 and mips64 for goldfish_dma_context aligment.

bug: 35618215

Change-Id: I147aeca391f97b0bccaed0b572b2fe13e627ad00

56ac887 2017-02-22 04:35:12 Lingfeng Yang

Don't restrict GLES 3.x based on API level nor arch.

bug: 35618215

Simpler that way, and we should be able to support all archs.

Change-Id: I77f5295047b05dc9188b15501be08b95b8f41302

02cf3e8 2017-02-18 07:45:49 Lingfeng Yang

Align goldfish_dma_context to 8 byte boundaries

bug: 35465006

It looks like when Chrome is started, we are in some strange world
where sizeof(goldfish_dma_context) = 16 instead of 24,
and we get in trouble with gralloc_register buffer,
and Chrome fails to boot.

TODO: Note that youtube videos inside chrome still don't seem to work,
for some reason (this doesn't depend on whether or not goldfish_dma is

Change-Id: Ibc66885434b885993025b70d957ea8e1adf524de

324c845 2017-02-18 02:56:26 Lingfeng Yang

[deqp] Fix state_query.string

bug: 34245902

Version string needs to match the glGetIntegerv(GL_MINOR_VERSION, ...).

Change-Id: I8b51104b0347c12c93c350dda7fb03c145f8c3e2

4e5446b 2017-02-16 01:59:48 Lingfeng Yang

[deqp] Fix wrong firstIndex for non-instanced draws with nonzero divisor

bug: 34245909

If the vertex attribute divisor is nonzero,
that actually affects first index for non-instanced draws
at least.

TODO: Figure out if this affects the instanced draws too.

dEQP-GLES3.functional.draw.random.210 : Fail -> Pass

Change-Id: Ida45d4689e0418ad8679091556a9fcaa64bbe25e

c92685e 2017-01-31 13:22:55 Lingfeng Yang

Fix regression in fbo.completeness after GL_EXT_color_buffer_float

Change-Id: Ie17da0442c80b972e5a66353b3998160fe04f032

Afficher sur ancien navigateur de dépôt.