Thank you for listing those reports. We don't have those exact devices with us but tried reproing on following close devices. But, no luck.
- Huawei Media Pad / 702L / 6.0.0
- HTC One M8 / 6.0.0
Most of the reports are on JB devices though.

kbr@/ kainino@ - Can you please take a look at this one?
The BOVE three 'In-app feedback reports' have logs attached.

We've received two additional reports saying bookmark or history page is not displayed and the screenshot in the feedback report shows as black page. I'm wondering if these two reports are related.  
https://listnr.corp.google.com/report/78403684854
https://listnr.corp.google.com/report/80153467828

It does look like a duplicate of  crbug.com/767913 . hongchichang@? What are the other affected devices ? kainino@ please can you confirm whether this issue is a duplicate of  crbug.com/767913 ?

I'm "Not a product user" on those listnr links so I have no way to do anything with this bug yet. I've requested access.

If anyone else might be able to reproduce, please note that it would appear on M62 but not M63; the fix appears on M63 at 63.0.3239.17 and on M64 at 64.0.3245.0.

kainino@, you should be able to access those listnr now 

Tried on Huawei Honor 5X (KIM-22L) / 6.0.1 and few JB devices but regrettably not reproducible. Thanks.

+a couple of Chrome Android graphics folks as FYI

https://listnr.corp.google.com/report/79384869565 <- this one looks to be corrupted gpu raster tile, adreno 200 gpu

The other two in-app ones show pretty much all black, and they both have Vivante GC4000 gpu. gpu raster is already is already disabled there, so presumably this is something else

we have a huawei honor 2 in the cabinet with Vivante GC4000, the hardware navigation buttons don't work, so it's borderline unusable :|

the honor 2 is still running 4.0.4, which chrome no longer supports, and there is no update available

guess we'll have to just go with educated guessing. anything changed in 62?

boliu: thanks for looking into this. Any chance you could take this issue? I'm still WFH today after being sick and don't have access to test devices.

> https://listnr.corp.google.com/report/79384869565 <- this one looks to be corrupted gpu raster tile, adreno 200 gpu

This one is almost certainly  issue 767913 .

If the affected devices shouldn't be using the GPU at all, then I would have no idea where to look. Maybe software compositor?

But if we can't find a device to repro on, this may be impossible to pin down.

I had ordered a device with Vivante GC4000, but that won't arrive until next week. I've also reached out to Huawei TAM if any are available here.

> If the affected devices shouldn't be using the GPU at all

There is no software composite fallback on android, gpu composite is always used. It's only raster that's software for vivante, according to this anyway: https://cs.chromium.org/chromium/src/gpu/config/software_rendering_list.json?rcl=13e9b84c8187eb11f42c04969a9a57757fd720f3&l=951

I can take it, but I guess no one can actually do anything until we get one of these devices and hopefully it's an easy repro..

boliu@ do you have any update here?

No. Did we get the device yet?

Device is delayed.  We're asking Huawei for help too.  See internal issue b/68826975.

Christine any update here? Please update this bug.

It's taking longer than expected to obtain a device so likely won't be able to fix this in time for M63.

Moving to M64 for now.

Got device. Has Hisilicon Technologies Immersion 16 GPU. (hisilicon licensed vivante GC4000 according to wikipedia)

But rendering is totally broken all the way back to m60, so I'm not sure if there is anything new in m62. There's loads of these logs, so I guess that's the place to start.
[ERROR:gl_fence_egl.cc(34)] : Failed to get EGLSync attribute. error code:12290

Does chrome still support fallback to gles fences I wonder..

disable-gpu-shader-disk-cache fixed this (noticed things are broken only on second launch)

EGL fence errors went away with it as well it appears.

oh, chrome://gpu for that the device

Deleted: gpu.html 94.2 KB

gpu.html 94.2 KB View Download

Thanks Bo for tracking that down. Can you add a targeted driver bug workaround for the device which turns on either disable_program_cache or disable_program_disk_cache, whichever will work around the problem?

yep will do

Hmm...

--disable-gpu-shader-disk-cache flag works
disable_program_cache workaround works
disable_program_disk_cache workaround doesn't work

Looking at an the code, I don't have an explanation why that is. (I don't think I made any mistakes, but that's always possibility..)

Let's just use the disable_program_cache workaround, unless you can debug what's going on with disable_program_disk_cache. I actually think that only disabling the shader disk cache is not correct. The bug's probably in the device's implementation of the GL_OES_get_program_binary extension, so we shouldn't use it for either an in-memory or on-disk cache.

Figured it out. disable_program_disk_cache disables writing anything new to the disk cache, but it doesn't prevent using anything from the cache if it's already there. The flag disables both.

And I think my local installs of dev builds aren't invalidating the disk cache. I don't know in production, if a new chrome version should invalidate invalidate the cache; if yes, then I think disable_program_disk_cache should work as well.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/b2893a602c8e043dcdc7e6cfb8b4cbd2d1878a4c

commit b2893a602c8e043dcdc7e6cfb8b4cbd2d1878a4c
Author: Bo Liu <boliu@chromium.org>
Date: Sat Dec 09 01:04:40 2017

gpu: Disable program cache for vivante GPUs

They have been found to cause massive corruptions.

Bug:  780498 
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: Iabf2a5500753ddf161084dcc1c1e55b44b157612
Reviewed-on: https://chromium-review.googlesource.com/817946
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Bo <boliu@chromium.org>
Cr-Commit-Position: refs/heads/master@{#522945}
[modify] https://crrev.com/b2893a602c8e043dcdc7e6cfb8b4cbd2d1878a4c/gpu/config/gpu_driver_bug_list.json

Re #31 - the disk cache is invalidated on any version change, so either work around should work in production. But your approach is safest, so sounds good.

Requesting merging CL in #32. It's disabling a GPU feature, which is expanding an existing GPU workaround used on other devices already, so shouldn't be an issue.

This bug requires manual review: M64 has already been promoted to the beta branch, so this requires manual review
Please contact the milestone owner if you have questions.
Owners: cmasso@(Android), cmasso@(iOS), kbleicher@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Please verify that this issue is fixed in M65 prior to merging it.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/cee6a2e537b1ce5009b5161736d5c85fb22f0dd2

commit cee6a2e537b1ce5009b5161736d5c85fb22f0dd2
Author: Bo Liu <boliu@chromium.org>
Date: Tue Jan 02 15:25:29 2018

[Merge M64] gpu: Disable program cache for vivante GPUs

They have been found to cause massive corruptions.

Bug:  780498 
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: Iabf2a5500753ddf161084dcc1c1e55b44b157612
Reviewed-on: https://chromium-review.googlesource.com/817946
Reviewed-by: Kenneth Russell <kbr@chromium.org>
Commit-Queue: Bo <boliu@chromium.org>
Cr-Commit-Position: refs/heads/master@{#522945}(cherry picked from commit b2893a602c8e043dcdc7e6cfb8b4cbd2d1878a4c)

TBR=kbr@chromium.org

Change-Id: Iabf2a5500753ddf161084dcc1c1e55b44b157612
Reviewed-on: https://chromium-review.googlesource.com/846265
Reviewed-by: Bo <boliu@chromium.org>
Cr-Commit-Position: refs/branch-heads/3282@{#379}
Cr-Branched-From: 5fdc0fab22ce7efd32532ee989b223fa12f8171e-refs/heads/master@{#520840}
[modify] https://crrev.com/cee6a2e537b1ce5009b5161736d5c85fb22f0dd2/gpu/config/gpu_driver_bug_list.json