Hi Mike, is the bug you were talking about in crbug.com/887649?

This is separate from Issue 887649. I have never seen this and can't repro on linux-chromeos nor eve.
xiulingw@, can you check another version (latest Dev is 72.0.3611.0) and/or another device? Any repro tips?
Scott and Fady, have you seen issues like this recently with any mojo apps or window service embeddings?

I have not seen anything like this. I wonder if this happens when close to the memory limit. Does a resize fix it?

I'll see if I can track down a kevin and reproduce.

Oshima had a Kevin, and I can repro. I will take a look to see if I can find out when this started happening.

On this device when I bring up the keyboard shortcut viewer I see a bunch of errors in the log file related to gles2_cmd_decoder. Specifically:

[shortcut:3818:3818:1119/091501.164805:ERROR:layer_tree_host_impl.cc(3157)] Forcing zero-copy tile initialization as worker context is missing
[shortcut:3818:3818:1119/091501.181873:ERROR:layer_tree_host_impl.cc(3157)] Forcing zero-copy tile initialization as worker context is missing
[3316:3316:1119/091501.310502:ERROR:gles2_cmd_decoder.cc(6334)] [.BrowserCompositor-0x143d5500]GL ERROR :GL_INVALID_OPERATION : glBindTexture: texture bound to more than 1 target.
[3316:3316:1119/091501.310752:ERROR:gles2_cmd_decoder.cc(9470)] [.BrowserCompositor-0x143d5500]GL ERROR :GL_INVALID_VALUE : glTexParameteri: unknown texture
[3316:3316:1119/091501.310900:ERROR:gles2_cmd_decoder.cc(9470)] [.BrowserCompositor-0x143d5500]GL ERROR :GL_INVALID_VALUE : glTexParameteri: unknown texture
[3316:3316:1119/091501.312271:ERROR:gles2_cmd_decoder.cc(10296)] [.BrowserCompositor-0x143d5500]RENDER WARNING: there is no texture bound to the unit 0
[3316:3316:1119/091501.312578:ERROR:gles2_cmd_decoder.cc(6334)] [.BrowserCompositor-0x143d5500]GL ERROR :GL_INVALID_OPERATION : glBindTexture: texture bound to more than 1 target.
[3316:3316:1119/091501.312763:ERROR:gles2_cmd_decoder.cc(9470)] [.BrowserCompositor-0x143d5500]GL ERROR :GL_INVALID_VALUE : glTexParameteri: unknown texture
[3316:3316:1119/091501.312874:ERROR:gles2_cmd_decoder.cc(9470)] [.BrowserCompositor-0x143d5500]GL ERROR :GL_INVALID_VALUE : glTexParameteri: unknown texture
[3316:3316:1119/091501.313026:ERROR:gles2_cmd_decoder.cc(10296)] [.BrowserCompositor-0x143d5500]RENDER WARNING: there is no texture bound to the unit 0

This is on 72.0.3612.0.

I don't see the same behavior on 71.0.3578.54, so this must be recent.

Oshima sees this on another device. The other device is arm and has the same gpu as kevin (mali-t860).

This may be the same as 906171.

Does this still happen if you disable gpu-raster for ui? https://cs.chromium.org/chromium/src/ui/compositor/compositor.cc?l=216

Disabling gpu-rasterization had no impact.

I regressed this to https://chromium-review.googlesource.com/c/chromium/src/+/1316347 . Passing to Antoine.

I see mention of mash. Is this running under mash? Because there's https://bugs.chromium.org/p/chromium/issues/detail?id=906171

This isn't with mash. We moved the keyboard shortcut viewer into it's own process and using the window-service APIs, but chrome isn't using the window service. I IMd with Mohsen earlier to see if this was related and he thought 906171 was specific to multi-process mash.

"We moved the keyboard shortcut viewer into it's own process and using the window-service APIs" that's probably related to the failure if this is specific to the keyboard viewer.
Can someone explain how this works, in particular:
- how this process connects to the GPU process (establishes a GpuChannelHost)
- how this process allocates GPU memory buffers
- how/where command-line flags are passed to this process

Is there a design doc that explain these?

This was considered work on top of the window-service, so no design doc.

Fady and/or Sadrul, please correct me I'm wrong, but I believe the code that connects to the gpu from this process is here: https://chromium.googlesource.com/chromium/src/+/master/services/ws/public/cpp/gpu/gpu.cc . 

keyboard shortcut viewer lives in a utility process and is launched as a service. Here's the interesting parts for setting up command line args:
#0  ChromeContentBrowserClient::AdjustUtilityServiceProcessCommandLine (this=0x9a15e566da0, identity=..., command_line=0x9a1610be330) at ../../chrome/browser/chrome_content_browser_client.cc:2170
#1  0x00007ffff0fe000b in content::UtilityProcessHost::StartProcess (this=0x9a161cd0ca0) at ../../content/browser/utility_process_host.cc:409
#2  0x00007ffff0fe0a35 in content::UtilityProcessHost::Start (this=0x9a161cd0ca0) at ../../content/browser/utility_process_host.cc:244
#3  0x00007ffff0d0d291 in content::(anonymous namespace)::StartServiceInUtilityProcess()

I spoke with Antoine about this. There are at least two issues that need to be sorted out:
1. Make sure we copy flags to the utility processes. For the time being, I will add back the code that blindly copies all flags (ChromeContentBrowserClient::AdjustUtilityServiceProcessCommandLine).
2. WindowPortMus (by way of MusContextFactory) needs to supply a RasterContextProvider when it creates the AsyncLayerTreeFrameSink. I will send out a patch for this early next week.

https://chromium-review.googlesource.com/c/chromium/src/+/1345074 might fix this, but I haven't yet been able to verify on device.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/bd677da12372ee7d2ff4cfcb8fce4f185cc7fecc

commit bd677da12372ee7d2ff4cfcb8fce4f185cc7fecc
Author: Antoine Labour <piman@chromium.org>
Date: Wed Nov 21 16:47:00 2018

Use correct BufferUsage for picking texture target in zero-copy

We need to make sure we consistently pick the texture target, because
native GMBs may not be available for all usages.

Bug:  905569 
Change-Id: Ibc325a23494597fefab3e62da71656bae30a42df
Reviewed-on: https://chromium-review.googlesource.com/c/1345074
Reviewed-by: Jonathan Backer <backer@chromium.org>
Commit-Queue: Antoine Labour <piman@chromium.org>
Cr-Commit-Position: refs/heads/master@{#610105}
[modify] https://crrev.com/bd677da12372ee7d2ff4cfcb8fce4f185cc7fecc/cc/raster/zero_copy_raster_buffer_provider.cc

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/3a4b161705cb3fb4cb55e521bb650dd109db4b01

commit 3a4b161705cb3fb4cb55e521bb650dd109db4b01
Author: Scott Violet <sky@chromium.org>
Date: Wed Nov 21 19:35:08 2018

chromeos: copy command line switches to ash services

Ash related services, such as shortcut-viewer, create various objects that
read command line flags (such as ui::Compositor). This makes such services
copy the command line flags from chrome. Long term we shold move to a whitelist,
but shorterm not copying all flags is causing display problems for some devices.

BUG=906954, 905569 
TEST=none

Change-Id: I408798f0c0a6145b84a926f464a6d76b3a116bb9
Reviewed-on: https://chromium-review.googlesource.com/c/1346690
Reviewed-by: John Abd-El-Malek <jam@chromium.org>
Commit-Queue: Scott Violet <sky@chromium.org>
Cr-Commit-Position: refs/heads/master@{#610164}
[modify] https://crrev.com/3a4b161705cb3fb4cb55e521bb650dd109db4b01/chrome/browser/chrome_content_browser_client.cc

Antoine, I verified this is fixed on tip of tree aaabc160f11b3396fea1ab6e2a19962550483e3f 610853 .

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/ea7742fa46f63572a4223223ac88f0a8853c803e

commit ea7742fa46f63572a4223223ac88f0a8853c803e
Author: Scott Violet <sky@chromium.org>
Date: Wed Nov 28 07:29:46 2018

window-service: makes window-service clients create RasterContextProvider

Refactors code for creating RasterContextProvider out of content into
services/ws so that can be used by window-service clients as well as
content.

BUG= 905569 
TEST=covered by tests

Change-Id: Iec41d949ebd9819e597e14e194bd15ba54318527
Reviewed-on: https://chromium-review.googlesource.com/c/1347209
Commit-Queue: Scott Violet <sky@chromium.org>
Reviewed-by: Antoine Labour <piman@chromium.org>
Reviewed-by: Sadrul Chowdhury <sadrul@chromium.org>
Cr-Commit-Position: refs/heads/master@{#611588}
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/cc/raster/staging_buffer_pool.cc
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/chrome/browser/ui/ash/chrome_browser_main_extra_parts_ash.cc
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/content/browser/compositor/gpu_process_transport_factory.cc
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/content/browser/compositor/gpu_process_transport_factory.h
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/services/ws/public/cpp/gpu/BUILD.gn
[add] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/services/ws/public/cpp/gpu/shared_worker_context_provider_factory.cc
[add] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/services/ws/public/cpp/gpu/shared_worker_context_provider_factory.h
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/DEPS
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/mus_context_factory.cc
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/mus_context_factory.h
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/window_port_mus.cc
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/window_port_mus.h
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/window_tree_client.cc
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/mus/window_tree_client.h
[modify] https://crrev.com/ea7742fa46f63572a4223223ac88f0a8853c803e/ui/aura/window_unittest.cc

Verified fixed on the latest version:
Version 73.0.3635.0(platform 11377.0.0-18.12.09) dev-channel kevin test
Google_Input_Tools_Version:72.0.20181203.223901082.