Bug 1516298 Comment 26 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

Original comment by
on 
As far as I can tell the test always times out waiting for a paint to happen, and shortly before that, there's this in the log:

```
07-09 16:30:26.368 11118 11175 E libEGL  : eglCreateWindowSurface: native_window_api_connect (win=0x7f48463d8010) failed (0xffffffed) (already connected to another API?)
07-09 16:30:26.368 11118 11175 E libEGL  : eglCreateWindowSurface:464 error 3003 (EGL_BAD_ALLOC)
07-09 16:30:26.368 11118 11175 I CompositorBridgeParent: Unable to renew compositor surface; remaining in paused state
```

which would explain why the paint never happens. It's not really clear to me why this happens, possibly a GeckoView bug, but it always seems to happen right after starting a new subtest. So maybe the code to launch a new subtest in a fresh window is cycling windows so fast that EGL or some other part of the pipeline gets confused about which surface is attached where. I'll see if inserting a small delay helps.
Revision 1 by
on 
As far as I can tell the test always times out waiting for a paint to happen, and shortly before that, there's this in the log:

```
07-09 16:30:26.366 11118 11175 I GeckoConsole: Renderer: Android Emulator OpenGL ES Translator (llvmpipe (LLVM 6.0, 256 bits))
07-09 16:30:26.366 11118 11175 I GeckoConsole: FBO Texture Target: TEXTURE_2D
07-09 16:30:26.368 11118 11175 E BufferQueueProducer: [SurfaceTexture-0-11118-18] connect: BufferQueue has been abandoned
07-09 16:30:26.368 11118 11175 E libEGL  : eglCreateWindowSurface: native_window_api_connect (win=0x7f48463d8010) failed (0xffffffed) (already connected to another API?)
07-09 16:30:26.368 11118 11175 E libEGL  : eglCreateWindowSurface:464 error 3003 (EGL_BAD_ALLOC)
07-09 16:30:26.368 11118 11175 I CompositorBridgeParent: Unable to renew compositor surface; remaining in paused state
```

which would explain why the paint never happens. It's not really clear to me why this happens, possibly a GeckoView bug, but it always seems to happen right after starting a new subtest. So maybe the code to launch a new subtest in a fresh window is cycling windows so fast that EGL or some other part of the pipeline gets confused about which surface is attached where. I'll see if inserting a small delay helps.

Back to Bug 1516298 Comment 26