Android webview hardware rendering broken |
|||||
Issue descriptionBisected to https://codereview.chromium.org/1745903002 All textures are just being renderered as black. Loads of logs error logs like this: 03-25 16:25:18.095 21574 21899 E chromium: [ERROR:gles2_cmd_decoder.cc(7835)] [GroupMarkerNotSet( crbug.com/242999 )!:54271E90]RENDER WARNING: texture bound to texture unit 0 is not renderable. It maybe non-power-of-2 and have incompatible texture filtering. looking more..
,
Mar 25 2016
Chrome on android not unaffected
,
Mar 25 2016
,
Mar 25 2016
more relevant error message: 03-25 16:43:37.398 23319 23349 E chromium: [ERROR:gles2_cmd_decoder.cc(14640)] [GroupMarkerNotSet( crbug.com/242999 )!:54220A98]GL ERROR :GL_INVALID_OPERATION : glCreateAndConsumeTextureCHROMIUM: invalid mailbox name This would be on the parent compositor, android render thread end
,
Mar 25 2016
Broken only for component build I think o_O My regular local build with gn: is_component_build=true is_clang=true use_goma=false I flipped these three, and the resulting apk works. I'm suspecting that the globals in MailboxManagerSync somehow isn't global anymore o_O
,
Mar 25 2016
Yeah, the static MailboxManagerSync::TextureGroup::mailbox_to_group_ variable has different address on the producer and receiver ends. How this can actually happen and the build is still successful is beyond me..
,
Mar 25 2016
For webview, two MailboxManagerSyncs are created from two different components: content/common/gpu/gpu_channel_manager.cc gpu/command_buffer/service/in_process_command_buffer.cc So then my theory is that gpu sources (including MailboxManagerSyncs) is now built twoce, into the content shared lib, and whatever shared lib in_process_command_buffer.cc happens to be in.
,
Mar 25 2016
,
Mar 25 2016
Yeah, afaict gpu/command_buffer/service and gpu/config was built twice in component build. Once into the content shared lib, and once into the gpu shared lib. I'm not sure why original CL had to depend on the source targets rather than the targets themselves. Latter seem to work for me fine: https://codereview.chromium.org/1838493002
,
Mar 25 2016
Original CL revert by brettw |
|||||
►
Sign in to add a comment |
|||||
Comment 1 by boliu@chromium.org
, Mar 25 2016