New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 737235 link

Starred by 2 users

Issue metadata

Status: Verified
Owner: ----
Closed: Jun 2017
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: Chrome
Pri: 1
Type: Bug



Sign in to add a comment

Black screen at login screen

Project Member Reported by hansberry@chromium.org, Jun 27 2017

Issue description

Chrome Version: ToT (sorry, can't access chrome:version)
OS: veyron-minnie, very recent test image (again, can't access chrome:version)

What steps will reproduce the problem?
(1) Deploy chrome (most recent build from master) to veyron-minnie.
(2) Observe.

What is the expected result?
To see the login screen.

What happens instead?
A black screen. Weirdly, the mouse cursor still seems to work fine -- it even seems to correctly recognize clickable elements because it transitions to the "pointer finger" cursor on some regions of the black screen.

More information:
This bug does not occur when I launch chrome virtually on my workstation -- only on a real device. I've only confirmed this bug on a veyron-minnie. I'm not sure if this affects other devices, but I would be surprised if the minnie is the only affected device. I'll followup once I test on other devices.

Unfortunately I cannot access about:gpu because I cannot login / see anything. However, I retrieved logs from /var/log/ui/ui.LATEST. I've attached the entire log to this bug, but here are the highlights:
[17063:17063:0627/100553.346159:ERROR:gles2_cmd_decoder.cc(2523)] [GroupMarkerNotSet( crbug.com/242999 )!:68087E35]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : BackFramebuffer::Create: <- error from previous GL command
[17063:17063:0627/100553.357759:ERROR:texture_manager.cc(3343)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : glTexImage2D: <- error from previous GL command
...
[17063:17063:0627/100553.446358:ERROR:gpu_memory_buffer_factory_native_pixmap.cc(123)] Failed to create GLImage 1280x800 format 12
[17063:17063:0627/100553.446875:ERROR:gles2_cmd_decoder.cc(17871)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_OPERATION : glBindTexImage2DCHROMIUM: no image found with the given ID
[17063:17063:0627/100553.498561:ERROR:gles2_cmd_decoder.cc(4417)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : glDrawElements: framebuffer incomplete
[17063:17063:0627/100553.498765:ERROR:gles2_cmd_decoder.cc(4417)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : glDrawElements: framebuffer incomplete
(The same error from gles2_cmd_decoder.cc repeats...)
[17063:17063:0627/100553.500268:ERROR:gles2_cmd_decoder.cc(4417)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : glDrawElements: framebuffer incomplete
[17063:17063:0627/100553.500342:ERROR:gles2_cmd_decoder.cc(4417)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : glDrawElements: framebuffer incomplete
[17063:17063:0627/100553.500540:ERROR:gles2_cmd_decoder.cc(12010)] [.DisplayCompositor-0x364cad80]GL ERROR :GL_INVALID_VALUE : glScheduleOverlayPlaneCHROMIUM: unsupported texture format
[17063:17063:0627/100553.506290:ERROR:gles2_cmd_decoder.cc(2523)] [GroupMarkerNotSet( crbug.com/242999 )!:68F47D35]GL ERROR :GL_INVALID_FRAMEBUFFER_OPERATION : BackFramebuffer::Create: <- error from previous GL command
[16865:16865:0627/100553.786889:ERROR:CONSOLE(1)] "Uncaught ReferenceError: login is not defined", source:  (1)
[17063:17063:0627/100553.827650:ERROR:gpu_memory_buffer_factory_native_pixmap.cc(123)] Failed to create GLImage 1280x800 format 12


Our team began seeing this issue yesterday, 6/26/2017. I went back through the history of src/gpu/ (checking out commits and rebuilding chrome) because that's where the errors where coming from to see if I could find the last time this bug did not occur:
* Jun 20 23:22:33: https://chromium.googlesource.com/chromium/src/+/96a04c90505930f286a37b8beb1c6852d0644fcd: Issue is not present 
* Jun 23 18:52:19: https://chromium.googlesource.com/chromium/src/+/279ae138790e2ba909d0229c1e429ffffda8578f: Issue is not present
* Jun 23 18:59:36: https://chromium.googlesource.com/chromium/src/+/64bddcd3d17a15e211a9b92f7e0a998d2713ecbb: Issue is not present
* [no other commits in src/gpu/ between these two commits]
* Jun 24 04:57:28: https://chromium.googlesource.com/chromium/src/+/aaa32d0ed6c515bd8e92c311a27d198b803369e6: Issue is present
* Jun 26 18:47:40: https://chromium.googlesource.com/chromium/src/+/12fd478c980c2af7ea532d7470a350acc19c3a9d: Issue is present
* Jun 26 22:26:35: https://chromium.googlesource.com/chromium/src/+/3d66a60a5666729cb88e23c0361b4acc09c45b88: Issue is present

https://chromium.googlesource.com/chromium/src/+/aaa32d0ed6c515bd8e92c311a27d198b803369e6 didn't seem to introduce this issue so I'm not sure where this bug began. I'll try to bisect more.







 
ui.LATEST.txt
49.3 KB View Download
Confirmed this issue on 9690.0.0 (nearly the most recent test-image available) so this isn't an problem with the image being out of date.
Cc: reve...@chromium.org zmo@chromium.org sadrul@chromium.org
Adding some owners of gpu and ui on CC for further visibility.

Comment 3 by wutao@chromium.org, Jun 27 2017

R61-9647.0.0 has the same issue.
Cc: r...@chromium.org
Components: UI>Shell>StartScreen
Silly question: Are you sure you can't log in? e.g., if you hover the mouse over where the clickable user pod should be, click it, type your password and hit Enter... does anything happen?

How did you retrieve logs? if your device is in dev mode and you can access the VT, `chrome --version` would give you the Chromium version number.

Added UI>Shell>StartScreen mostly as FYI.
Possible duplicate of  bug 736807  ?
(Oh, maybe not.  When I saw 736807 on x86 platforms, it didn't even boot to the login screen.)
Hi Michael,

I actually did try logging in -- I clicked on what I assumed to be the user pod and typed in my password -- nothing perceivable happened.

I retrieved logs via SSH (my chromebook's ethernet IP address stays consistent for several days). I can also access the VT2 terminal just fine (it correctly renders).
Could be related to https://chromium-review.googlesource.com/c/538277/. Mali driver should have been updated to support this properly but maybe minnie didn't get this change for some reason. 
Hi David,

I believe you're correct. I built and deployed chrome at your change, https://chromium.googlesource.com/chromium/src/+/4e06ceca760fefd5a551237cb57d6aeb8f48c03a, and observed this issue. I then built and deployed chrome at the change right before, https://chromium.googlesource.com/chromium/src/+/7e18c32950f83dc1d94420eee6299869ad4c44b3, and did not observe this issue.

Please let me know if you need more information / logs / etc.

Comment 10 by r...@chromium.org, Jun 27 2017

Cc: wzang@chromium.org jdufault@chromium.org
What version of ChromeOS are you using on this device?
I'm seeing this issue on canary 9690.0.0.
That's surprising as 9690.0.0 should include https://chrome-internal-review.googlesource.com/c/369508/
Project Member

Comment 15 by bugdroid1@chromium.org, Jun 28 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/942b6421b57e47ab17c087f51436cbb3e4355f60

commit 942b6421b57e47ab17c087f51436cbb3e4355f60
Author: Daniele Castagna <dcastagna@chromium.org>
Date: Wed Jun 28 05:55:18 2017

ozone: s/DLOG/LOG on GLImage init failure.

Always log the EGL error string when eglCreateImageKHR fails.

Bug:  737235 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Change-Id: I418bc56e352b622209eb90223f3494488e677bf1
Reviewed-on: https://chromium-review.googlesource.com/551315
Reviewed-by: David Reveman <reveman@chromium.org>
Reviewed-by: Robert Kroeger <rjkroege@chromium.org>
Commit-Queue: Daniele Castagna <dcastagna@chromium.org>
Cr-Commit-Position: refs/heads/master@{#482898}
[modify] https://crrev.com/942b6421b57e47ab17c087f51436cbb3e4355f60/ui/gl/gl_image_egl.cc

Project Member

Comment 16 by bugdroid1@chromium.org, Jun 28 2017

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/24d239973fefa7bfd729aac06898e88ce92c90bb

commit 24d239973fefa7bfd729aac06898e88ce92c90bb
Author: David Reveman <reveman@chromium.org>
Date: Wed Jun 28 16:36:28 2017

Revert "ui: Use EGL_EXT_image_flush_external correctly when creating EGLImage."

This reverts commit 4e06ceca760fefd5a551237cb57d6aeb8f48c03a.

Reason for revert: Causing GLImage create failure on Minnie.

BUG= 737235 

Original change's description:
> ui: Use EGL_EXT_image_flush_external correctly when creating EGLImage.
> 
> EGL_IMAGE_EXTERNAL_FLUSH_EXT attribute should be set to EGL_TRUE
> when creating an image that support eglImageFlushExternalEXT.
> 
> BUG= 720045 
> TEST=gl_unittests
> 
> Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
> Change-Id: I7e01f5cdb964da5c1599a0fe62efb8440b765852
> Reviewed-on: https://chromium-review.googlesource.com/538277
> Reviewed-by: Kenneth Russell <kbr@chromium.org>
> Reviewed-by: Robert Kroeger <rjkroege@chromium.org>
> Reviewed-by: Daniele Castagna <dcastagna@chromium.org>
> Commit-Queue: David Reveman <reveman@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#482134}

TBR=rjkroege@chromium.org,reveman@chromium.org,kbr@chromium.org,dcastagna@chromium.org

# Not skipping CQ checks because original CL landed > 1 day ago.

Bug:  720045 
Change-Id: I9ffde8b9f48de0d7fc2d6288aad8bc823705120a
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel;master.tryserver.chromium.linux:linux_optional_gpu_tests_rel;master.tryserver.chromium.mac:mac_optional_gpu_tests_rel;master.tryserver.chromium.win:win_optional_gpu_tests_rel
Reviewed-on: https://chromium-review.googlesource.com/551181
Reviewed-by: David Reveman <reveman@chromium.org>
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Reviewed-by: Robert Kroeger <rjkroege@chromium.org>
Commit-Queue: David Reveman <reveman@chromium.org>
Cr-Commit-Position: refs/heads/master@{#483014}
[modify] https://crrev.com/24d239973fefa7bfd729aac06898e88ce92c90bb/third_party/khronos/EGL/eglext.h
[modify] https://crrev.com/24d239973fefa7bfd729aac06898e88ce92c90bb/ui/gl/gl_image_native_pixmap.cc
[modify] https://crrev.com/24d239973fefa7bfd729aac06898e88ce92c90bb/ui/gl/test/gl_image_test_template.h
[modify] https://crrev.com/24d239973fefa7bfd729aac06898e88ce92c90bb/ui/ozone/gl/gl_image_ozone_native_pixmap_unittest.cc

I just built and deployed chrome on my veyron-minnie and I'm no longer seeing this issue. 

Thank you for your help David! Would you like me to close this bug or do you have other CLs related to this bug planned?
Status: Fixed (was: Untriaged)
Labels: ReleaseBlock-Dev
Issue is seen on 61.0.3142.0/9695.0.0 dev-channel Candy,Daisy and Minnie devices
Cc: songsuk@chromium.org pucchakayala@chromium.org
Labels: M-61

Comment 21 by ajha@chromium.org, Jun 29 2017

Cc: josa...@chromium.org gkihumba@chromium.org
Issue is working fine for  61.0.3144.0/9693.1.0 dev-channel Daisy,Kip and Minnie
But for Candy Device Issue is seen and filed a new Issue 737945 
Status: Verified (was: Fixed)

Sign in to add a comment