Chrome apps may disappear for a while if launched right after user session starts |
|||||||||||||||||
Issue descriptionChrome Version: 72.0.3592.0 What steps will reproduce the problem? (1) Log in a user, right after session starts, hit "alt+shift+i" to open the Feedback app. Or, open the wallpaper picker (needs to pin the app icon in the shelf in order to open it quickly enough.) (2) Alternatively, use a device in demo mode, and observe the screensaver app that's auto-launched right after session starts. What is the expected result? The app windows should be shown, and the screensaver video should play normally. What happens instead? 1) The wallpaper picker window disappears for several seconds, and then appears again. Video: https://drive.google.com/file/d/1GAf5h0uk2_AuyTuDzf-BsplxHYRpGfX5/view?usp=sharing 2) The screensaver disappears, leaving an empty background, and then appear again. Video: https://drive.google.com/file/d/1-6gWVwa0eFsC_UCZDs12XhQZtFDaxMJr/view?usp=sharing 3) The Feedback app window becomes transparent, and does not recover automatically. User must click on the window, or the browser window behind in order to restore. (Note that the contents in the browser tab disappears as well. Could this be a Blink issue?) Video: https://drive.google.com/file/d/1-9CeJPnEHaYn4rInWY0M1Kdsh8uCE8Pm/view?usp=sharing Please note: 1) The issue does not reproduce on accounts. But for accounts that's reproducible, it appears 100% of the time. 2) It consistently reproduces in demo account, so it's a big polish issue for the demo experience. I didn't see anything suspicious in the lo. I also need to check if the issue exists in M71. lazyboy@, could you help triage this bug? Thanks so much!
,
Nov 2
Seeing the issue on both Samus 72.0.3598.0 and Eve 72.0.3592.0. Added the issue to go/demobugs lazyboy@, could you help triage? Thanks a lot!
,
Nov 2
For me this repros on eve with black screen and no wallpaper visible. Chrome: 72.0.3599.0 Video: https://drive.google.com/open?id=1-Parubz0aLqJIkhrupr175cvSpUmPbH-
,
Nov 2
Does this reproduce on any chromeos device?
,
Nov 3
Reassigning to hoegsberg@ since bisect shows https://chromium-review.googlesource.com/c/chromium/src/+/1045256 is the first bad commit. Could you take a look? Thanks! This is a blocker for Chrome OS demo mode. Some data points: 1) It's confusing that the issue still exists after the patch is reverted locally! There may be another bad commit? On the other hand, this commit is definitely problematic because this issue is not reproducible when the commit position is at its parent, but 100% reproducible after the commit. 2) I believe the black screen issue (#3) and the transparent window issue (#0) is the same? Sometimes I saw a combination of both. 3) Issue is reproducible at least on Eve and Samus, but is not reproducible on Kevin? (Not sure which version michaelpg@'s device is on, needs verification.)
,
Nov 3
Adding @google.com account since their @chromium.org account shows last visit > 30 days ago.
,
Nov 3
The CL linked in #5 should not affect samus at all since HW overlays are disabled on that device.
,
Nov 3
OK. I will verify it on Samus again, since I used Eve to do the bisect.
,
Nov 5
Another bisect on Samus shows https://chromium-review.googlesource.com/c/chromium/src/+/1277826 is the culprit. Reverting it on Samus fixes the issue. On Eve, both this CL and the CL in #5 need to be reverted to fix the issue. Shall I go ahead and do that?
,
Nov 5
I'd revert. Do we know if that CL ended up in 71? Assigning to the author of crrev.com/c/1277826
,
Nov 5
Neither of the two CLs are in M71. I've created a revert of crrev.com/c/1045256 (crrev.com/c/1318547 ), resolved the conflicts, and it's in CQ. For crrev.com/c/1277826, there're two many merge conflicts and it's hard to resolve without context. samans@, please help, thanks! (crrev.com/c/1318671)
,
Nov 5
The revert could be messy. I'll spend a day trying to repro and fix this issue before I attempt a revert.
,
Nov 5
Thanks! Reverting crrev.com/c/1318547 is also messy because of the CQ failure. hoegsberg@ could you kindly take a look?
,
Nov 5
Are you sure that CL:1045256 is still a problem?
,
Nov 5
Yes, 100% sure. Looks like we'll have to revert crrev.com/c/1291590 and crrev.com/c/1292822 as well. Or could you think of a better fix?
,
Nov 5
OK, I'll revert those two as well.
,
Nov 5
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/043bdd225bcd415dbf81ba078b7981b51eef1408 commit 043bdd225bcd415dbf81ba078b7981b51eef1408 Author: Wenzhao (Colin) Zang <wzang@chromium.org> Date: Mon Nov 05 23:36:45 2018 Revert "viz: add unittests for CL:1291590" This reverts commit f8dfbe74c1f5c79051461cdf448abf7f8188336f. Reason for revert: crbug.com/900373 Original change's description: > viz: add unittests for CL:1291590 > > BUG=896945 > TEST=SingleOverlayOnTopTest.AllowVideoNormalTransformWithOutputSurfaceOverlay > > Change-Id: Iaa46909657f3e11f1d982f94aaff102676a74b4a > Reviewed-on: https://chromium-review.googlesource.com/c/1292822 > Commit-Queue: Daniele Castagna <dcastagna@chromium.org> > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Cr-Commit-Position: refs/heads/master@{#602548} TBR=dongseong.hwang@intel.com,dcastagna@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: 896945, 900373 Change-Id: I57a2c9770df54447e207e7a1d1079006087829ae Reviewed-on: https://chromium-review.googlesource.com/c/1318359 Reviewed-by: Wenzhao (Colin) Zang <wzang@chromium.org> Commit-Queue: Wenzhao (Colin) Zang <wzang@chromium.org> Cr-Commit-Position: refs/heads/master@{#605517} [modify] https://crrev.com/043bdd225bcd415dbf81ba078b7981b51eef1408/components/viz/service/display/overlay_unittest.cc
,
Nov 5
I'm not following, in #13 you mentioned a CL, then you reverted different ones.
,
Nov 5
Could you read #9 again, thanks!
,
Nov 5
..and #15
,
Nov 5
To be clear, there are two culprit CLs: 1) crrev.com/c/1277826. samans@ is investigating on a fix to avoid the messy revert. 2) crrev.com/c/1318547, which goes together with crrev.com/c/1291590 and crrev.com/c/1292822. hoegsberg@ suggesting reverting them all in #16. Please let me know if you have more questions.
,
Nov 6
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/ec93e1108c6b20b6df8b54a88e48b043f00b1f59 commit ec93e1108c6b20b6df8b54a88e48b043f00b1f59 Author: Wenzhao (Colin) Zang <wzang@chromium.org> Date: Tue Nov 06 01:33:30 2018 Revert "viz: Fix wrong uv_rect calculation" This reverts commit 4bf9d01dbf34f17c899095a937dabbb884ec87fe. Reason for revert: crbug.com/900373 Original change's description: > viz: Fix wrong uv_rect calculation > > CL:1045256 calculates |uv_rect| of the primary plane in wrong way. It causes > drmModeAtomicAddProperty failure. This CL fixes it. > > BUG=896945 > > Change-Id: I6e914707cb961e8294e8ddd160b75e48ae5a2b64 > Reviewed-on: https://chromium-review.googlesource.com/c/1291590 > Commit-Queue: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Reviewed-by: Kristian H. Kristensen <hoegsberg@chromium.org> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Cr-Commit-Position: refs/heads/master@{#601392} TBR=dongseong.hwang@intel.com,dcastagna@chromium.org,hoegsberg@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: 896945, 900373 Change-Id: Ibab22a4fafb7be1ea0921f0cf774b5973d92b0b2 Reviewed-on: https://chromium-review.googlesource.com/c/1318796 Reviewed-by: Wenzhao (Colin) Zang <wzang@chromium.org> Commit-Queue: Wenzhao (Colin) Zang <wzang@chromium.org> Cr-Commit-Position: refs/heads/master@{#605559} [modify] https://crrev.com/ec93e1108c6b20b6df8b54a88e48b043f00b1f59/components/viz/service/display/overlay_processor.cc [modify] https://crrev.com/ec93e1108c6b20b6df8b54a88e48b043f00b1f59/components/viz/service/display/overlay_unittest.cc
,
Nov 6
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/ae83af95226ea1f66770683a36a2d27903c81d3b commit ae83af95226ea1f66770683a36a2d27903c81d3b Author: Wenzhao Zang <wzang@chromium.org> Date: Tue Nov 06 04:14:37 2018 Revert "viz: Turn off or crop primary plane when possible" This reverts commit 2bdc43adce184e5298c0b1cc6a450b69b6997da3. Reason for revert: crbug.com/900373 Original change's description: > viz: Turn off or crop primary plane when possible > > For many fullscreen cases, the rest of the renderpass quads are solid > black quads. In that case, we can turn off the primary plane, and not > having to scan out an full screen of black pixels saves a good deal of > power. In other cases, we can crop down the primary plane down to the > content rectangle (youtube channel logo, for example) and save power. > > Averaging battery draw measurement over about a minute of play back > for a fullscreen youtube video, power savings for soraka is around > 500mW (6.5W to 6.0W) and 300mW for kevin (4.4W to 4.1W), when we > turn off the primary plane. > > Change-Id: I7946d7acab334bffdf2bb49a1d9c424cf46fc610 > Reviewed-on: https://chromium-review.googlesource.com/c/1045256 > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Commit-Queue: Kristian H. Kristensen <hoegsberg@chromium.org> > Cr-Commit-Position: refs/heads/master@{#600415} TBR=dongseong.hwang@intel.com,dcastagna@chromium.org,hoegsberg@chromium.org # Not skipping CQ checks because original CL landed > 1 day ago. Bug: 900373 Change-Id: I37857bca0cda656c2e2a83dc04312b9c599976e7 Reviewed-on: https://chromium-review.googlesource.com/c/1318547 Commit-Queue: Wenzhao (Colin) Zang <wzang@chromium.org> Reviewed-by: Wenzhao (Colin) Zang <wzang@chromium.org> Cr-Commit-Position: refs/heads/master@{#605591} [modify] https://crrev.com/ae83af95226ea1f66770683a36a2d27903c81d3b/components/viz/service/display/gl_renderer_unittest.cc [modify] https://crrev.com/ae83af95226ea1f66770683a36a2d27903c81d3b/components/viz/service/display/overlay_processor.cc [modify] https://crrev.com/ae83af95226ea1f66770683a36a2d27903c81d3b/components/viz/service/display/overlay_processor.h [modify] https://crrev.com/ae83af95226ea1f66770683a36a2d27903c81d3b/components/viz/service/display/overlay_strategy_single_on_top.cc [modify] https://crrev.com/ae83af95226ea1f66770683a36a2d27903c81d3b/components/viz/service/display/overlay_strategy_underlay.cc [modify] https://crrev.com/ae83af95226ea1f66770683a36a2d27903c81d3b/components/viz/service/display/overlay_unittest.cc
,
Nov 6
Do all the bugs in #0 still happen after the revert, or only a subset of them?
,
Nov 6
I'm asking because only the third issue sounds like it could be caused by me.
,
Nov 6
2) is extremely reproducible after the CL in #9. Sorry I didn't verify 1) and 3) during the bisecting because only 2) looks like a serious UI issue, and it's difficult to verify all of them together. At this point, I think we should focus on fixing 2). If you have a Samus device, you could verify 2) yourself. I could provide the instructions to enroll the device in demo mode. If you don't have access to it, please let me know what the best next step you think should be. Thanks.
,
Nov 6
Will Eve work too? I'll see if I can find one of them in my office. Is there a way to try the demo mode on my linux workstation? I know this bug probably doesn't reproduce there but I want to try something.
,
Nov 6
Eve will work. Please make sure to include the three reverts in the build. And you should be able to verify the issue still exists. Unfortunately, we don't have a reliable way to simulate demo mode in Linux workstation. To set up demo mode on device: 1) Start with the OOBE welcome screen (if the device already has an owner, hit "ctrl+alt+shift+R" to powerwash). 2) On the welcome screen, hit "ctrl+alt+D" and follow the steps. Please make sure the device is connected to network the whole time. 3) After entering the demo session, observe the auto-launched screensaver has the issue described in #0. (During the first time, the screensaver requires downloading, so the screensaver will show up after an arbitrary amount of time and the problem may not be visible. You can manually exit the session, and the next time the screensaver will be launched right after session starts, and shows the problem.) Please let me know if you have questions. Thanks!
,
Nov 6
Thank you for the instructions. I expect to have the Samus machine tomorrow. I'll start investigating as soon as I have it.
,
Nov 7
FYI this revert also fixed another issue I was chasing: https://b.corp.google.com/issues/118630028
,
Nov 8
,
Nov 8
,
Nov 8
,
Nov 8
,
Nov 8
This revert also fixed two additional issues I was chasing: https://b.corp.google.com/issues/118503559 https://b.corp.google.com/issues/118502516 so three in total.
,
Nov 9
,
Nov 12
This seems to be an existing bug but my CL has changed the way the bug manifests itself. The screensaver is told to hide while it's still being shown (RenderWidgetHostViewAura::WasHidden is called). Before my CL, this used to cause a few seconds of freezing; now we actually hide the screensaver.
,
Nov 12
samans@, thanks! We indeed have a video freezing issue (crbug.com/884285). Any chance we can fix the two issues together? (If not, I think it's still worthwhile fixing the disappearance issue first, because it's much worse than freezing)
,
Nov 12
I don't think I can fix both problems (the root cause is probably somewhere in the Chrome OS code that I'm not familiar with), but I'll do my best to see what can be done about the disappearance issue.
,
Nov 13
OK, I think I know why this is happening. My CL is not doing anything wrong; it's just revealing some nastiness in Chrome OS. When I log in (at least in the demo mode), I see 30+ renderer processes created that die after about 10 seconds and don't draw anything in the meantime (I've attached the list of URLs corresponding to these renderers; they're all extensions). The large number of renderers guarantees that any extra renderer process (the screensaver in this case) that is created during this time will have its content evicted as soon as it's marked hidden. Here are some possible fixes: (1) Don't mark the screensaver hidden while it's still being shown. (2) Don't create so many renderer processes that don't actually do anything. (3) If we have to create the renderer processes for whatever reason, at least mark them hidden. I personally don't believe a revert is appropriate, because it will break other stuff that I've been working on and does not really fix the problem at its core (we will still have freezes, we still create unnecessary renderer processes, etc). I don't think I have the knowledge to fix these issues, so passing back to wzang@.
,
Nov 13
I think these are background pages that are spun up when extensions are started, then shut down after idling. lazyboy: Can we avoid creating these renderers in the first place at session start? If not, can they be marked as hidden, to improve the reliability of apps that need to appear visibly?
,
Nov 13
Please note that the real bug is that we tell the screensaver it's hidden while it's still visible. The renderer processes spawned at login time just make it worse by causing disappearance instead of a freeze. A hack should be possible but please be aware of the real issue and work on it.
,
Nov 14
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/27ac690d2b92b3fe1635036c1035db5c37d5abe9 commit 27ac690d2b92b3fe1635036c1035db5c37d5abe9 Author: Saman Sami <samans@chromium.org> Date: Wed Nov 14 18:14:14 2018 Hack for demo screensaver disappearing momentarily after login There is an existing bug in Chrome OS where the screensaver is told to hide even though it's still visible. Before crrev.com/c/1277826 this used to only cause a freeze, but since then the screensaver disappears altogether. The reason is that during the first 10-20 seconds after logging into Chrome OS, there are 30+ renderers spawned in the background that don't actually draw anything. These renderers overwhelm FrameEvictionManager such that any renderer that goes hidden (e.g. the screensaver) will have its surface immediately evicted. This used to work fine before because notifying FrameEvictionManager used to happen in OnFirstSurfaceActivation as opposed to EmbedSurface, and therefore these useless renderers would not interfere with frame eviction. The Chrome OS team is worried they won't have a proper fix in time, so for the time being implement a behaviour similar to what we had before: only report DelegatedFrameHosts to FrameEvictionManager that have seen their first surface activation. Bug: 900373 Change-Id: Ic98929f8e1c834de2af22aa299805395be6a9072 Reviewed-on: https://chromium-review.googlesource.com/c/1334590 Commit-Queue: Saman Sami <samans@chromium.org> Reviewed-by: Fady Samuel <fsamuel@chromium.org> Cr-Commit-Position: refs/heads/master@{#608041} [modify] https://crrev.com/27ac690d2b92b3fe1635036c1035db5c37d5abe9/content/browser/renderer_host/delegated_frame_host.cc [modify] https://crrev.com/27ac690d2b92b3fe1635036c1035db5c37d5abe9/content/browser/renderer_host/delegated_frame_host.h [modify] https://crrev.com/27ac690d2b92b3fe1635036c1035db5c37d5abe9/content/browser/renderer_host/render_widget_host_view_aura_unittest.cc
,
Nov 14
samans@, thanks for the fix.
,
Nov 15
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/3c6137e7e6ae4bc7e83699491b666b38c0d86d78 commit 3c6137e7e6ae4bc7e83699491b666b38c0d86d78 Author: Saman Sami <samans@chromium.org> Date: Thu Nov 15 00:45:54 2018 Fix wrong OS guard in DelegatedFrameHost Bug: 900373 Change-Id: Ibaaeda2fcff69b5bf45eb7de767731bf53177f56 Reviewed-on: https://chromium-review.googlesource.com/c/1336510 Reviewed-by: Fady Samuel <fsamuel@chromium.org> Commit-Queue: Saman Sami <samans@chromium.org> Cr-Commit-Position: refs/heads/master@{#608198} [modify] https://crrev.com/3c6137e7e6ae4bc7e83699491b666b38c0d86d78/content/browser/renderer_host/delegated_frame_host.cc
,
Nov 15
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/3c6137e7e6ae4bc7e83699491b666b38c0d86d78 commit 3c6137e7e6ae4bc7e83699491b666b38c0d86d78 Author: Saman Sami <samans@chromium.org> Date: Thu Nov 15 00:45:54 2018 Fix wrong OS guard in DelegatedFrameHost Bug: 900373 Change-Id: Ibaaeda2fcff69b5bf45eb7de767731bf53177f56 Reviewed-on: https://chromium-review.googlesource.com/c/1336510 Reviewed-by: Fady Samuel <fsamuel@chromium.org> Commit-Queue: Saman Sami <samans@chromium.org> Cr-Commit-Position: refs/heads/master@{#608198} [modify] https://crrev.com/3c6137e7e6ae4bc7e83699491b666b38c0d86d78/content/browser/renderer_host/delegated_frame_host.cc
,
Nov 27
I submitted the reland CL with the new change fixing this issue. https://chromium-review.googlesource.com/c/chromium/src/+/1346997
,
Nov 28
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/1ceb07933f73defaa73c30153fcbe812ad1965e0 commit 1ceb07933f73defaa73c30153fcbe812ad1965e0 Author: Dongseong Hwang <dongseong.hwang@intel.com> Date: Wed Nov 28 01:15:08 2018 Reland "viz: Turn off or crop primary plane when possible" This is a reland of I7946d7acab334bffdf2bb49a1d9c424cf46fc610 and the following bug fixes as follows; "viz: Fix wrong uv_rect calculation" 4bf9d01dbf34f17c899095a937dabbb884ec87fe "viz: add unittests for CL:1291590" f8dfbe74c1f5c79051461cdf448abf7f8188336f Additional fix: - Fix the flickering issue because of primary |uv_rect| cropping even though an overlay attemption fails mainly due to limited pixel rate (e.g. 4k video). - Fix wrong |uv_rect| calculation which causes atomic page flip failures with "No space left on device (28)" error. Original change's description: > viz: Turn off or crop primary plane when possible > > For many fullscreen cases, the rest of the renderpass quads are solid > black quads. In that case, we can turn off the primary plane, and not > having to scan out an full screen of black pixels saves a good deal of > power. In other cases, we can crop down the primary plane down to the > content rectangle (youtube channel logo, for example) and save power. > > Averaging battery draw measurement over about a minute of play back > for a fullscreen youtube video, power savings for soraka is around > 500mW (6.5W to 6.0W) and 300mW for kevin (4.4W to 4.1W), when we > turn off the primary plane. > > Change-Id: I7946d7acab334bffdf2bb49a1d9c424cf46fc610 > Reviewed-on: https://chromium-review.googlesource.com/c/1045256 > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Commit-Queue: Kristian H. Kristensen <hoegsberg@chromium.org> > Cr-Commit-Position: refs/heads/master@{#600415} TEST=SingleOverlayOnTopTest.AllowVideoNormalTransformWithOutputSurfaceOverlay Bug: 896945, 900373 Change-Id: I88db118dbc3ca2fed8c8eaf1f9a4337cdae12990 Reviewed-on: https://chromium-review.googlesource.com/c/1346997 Commit-Queue: Dongseong Hwang <dongseong.hwang@intel.com> Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> Reviewed-by: Kristian H. Kristensen <hoegsberg@chromium.org> Reviewed-by: Daniele Castagna <dcastagna@chromium.org> Cr-Commit-Position: refs/heads/master@{#611484} [modify] https://crrev.com/1ceb07933f73defaa73c30153fcbe812ad1965e0/components/viz/service/display/gl_renderer_unittest.cc [modify] https://crrev.com/1ceb07933f73defaa73c30153fcbe812ad1965e0/components/viz/service/display/overlay_processor.cc [modify] https://crrev.com/1ceb07933f73defaa73c30153fcbe812ad1965e0/components/viz/service/display/overlay_processor.h [modify] https://crrev.com/1ceb07933f73defaa73c30153fcbe812ad1965e0/components/viz/service/display/overlay_strategy_single_on_top.cc [modify] https://crrev.com/1ceb07933f73defaa73c30153fcbe812ad1965e0/components/viz/service/display/overlay_strategy_underlay.cc [modify] https://crrev.com/1ceb07933f73defaa73c30153fcbe812ad1965e0/components/viz/service/display/overlay_unittest.cc
,
Nov 29
,
Dec 7
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/e1809a76301288b5b74afc6a785cc2c15da0f938 commit e1809a76301288b5b74afc6a785cc2c15da0f938 Author: Kazuhiro Inaba <kinaba@google.com> Date: Fri Dec 07 17:21:16 2018 Revert "Reland "viz: Turn off or crop primary plane when possible"" This reverts commit 1ceb07933f73defaa73c30153fcbe812ad1965e0. Reason for revert: ARC and Crostini on ARM broken (crbug.com/896945#c17) Original change's description: > Reland "viz: Turn off or crop primary plane when possible" > > This is a reland of I7946d7acab334bffdf2bb49a1d9c424cf46fc610 and > the following bug fixes as follows; > "viz: Fix wrong uv_rect calculation" 4bf9d01dbf34f17c899095a937dabbb884ec87fe > "viz: add unittests for CL:1291590" f8dfbe74c1f5c79051461cdf448abf7f8188336f > > Additional fix: > - Fix the flickering issue because of primary |uv_rect| cropping even though > an overlay attemption fails mainly due to limited pixel rate (e.g. 4k video). > - Fix wrong |uv_rect| calculation which causes atomic page flip failures with > "No space left on device (28)" error. > > Original change's description: > > viz: Turn off or crop primary plane when possible > > > > For many fullscreen cases, the rest of the renderpass quads are solid > > black quads. In that case, we can turn off the primary plane, and not > > having to scan out an full screen of black pixels saves a good deal of > > power. In other cases, we can crop down the primary plane down to the > > content rectangle (youtube channel logo, for example) and save power. > > > > Averaging battery draw measurement over about a minute of play back > > for a fullscreen youtube video, power savings for soraka is around > > 500mW (6.5W to 6.0W) and 300mW for kevin (4.4W to 4.1W), when we > > turn off the primary plane. > > > > Change-Id: I7946d7acab334bffdf2bb49a1d9c424cf46fc610 > > Reviewed-on: https://chromium-review.googlesource.com/c/1045256 > > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > > Commit-Queue: Kristian H. Kristensen <hoegsberg@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#600415} > > TEST=SingleOverlayOnTopTest.AllowVideoNormalTransformWithOutputSurfaceOverlay > > Bug: 896945, 900373 > Change-Id: I88db118dbc3ca2fed8c8eaf1f9a4337cdae12990 > Reviewed-on: https://chromium-review.googlesource.com/c/1346997 > Commit-Queue: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Kristian H. Kristensen <hoegsberg@chromium.org> > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Cr-Commit-Position: refs/heads/master@{#611484} TBR=dongseong.hwang@intel.com,dcastagna@chromium.org,hoegsberg@chromium.org Bug: 896945, 900373 Change-Id: Iec6d2812d94c286c495d90d8952b65ecffa4c010 Reviewed-on: https://chromium-review.googlesource.com/c/1363734 Reviewed-by: Kazuhiro Inaba <kinaba@chromium.org> Reviewed-by: Daniele Castagna <dcastagna@chromium.org> Commit-Queue: Daniele Castagna <dcastagna@chromium.org> Cr-Commit-Position: refs/heads/master@{#614732} [modify] https://crrev.com/e1809a76301288b5b74afc6a785cc2c15da0f938/components/viz/service/display/gl_renderer_unittest.cc [modify] https://crrev.com/e1809a76301288b5b74afc6a785cc2c15da0f938/components/viz/service/display/overlay_processor.cc [modify] https://crrev.com/e1809a76301288b5b74afc6a785cc2c15da0f938/components/viz/service/display/overlay_processor.h [modify] https://crrev.com/e1809a76301288b5b74afc6a785cc2c15da0f938/components/viz/service/display/overlay_strategy_single_on_top.cc [modify] https://crrev.com/e1809a76301288b5b74afc6a785cc2c15da0f938/components/viz/service/display/overlay_strategy_underlay.cc [modify] https://crrev.com/e1809a76301288b5b74afc6a785cc2c15da0f938/components/viz/service/display/overlay_unittest.cc
,
Dec 10
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7 commit a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7 Author: Kazuhiro Inaba <kinaba@google.com> Date: Mon Dec 10 12:14:50 2018 Revert "Reland "viz: Turn off or crop primary plane when possible"" This reverts commit 1ceb07933f73defaa73c30153fcbe812ad1965e0. Reason for revert: ARC and Crostini on ARM broken (crbug.com/896945#c17) Original change's description: > Reland "viz: Turn off or crop primary plane when possible" > > This is a reland of I7946d7acab334bffdf2bb49a1d9c424cf46fc610 and > the following bug fixes as follows; > "viz: Fix wrong uv_rect calculation" 4bf9d01dbf34f17c899095a937dabbb884ec87fe > "viz: add unittests for CL:1291590" f8dfbe74c1f5c79051461cdf448abf7f8188336f > > Additional fix: > - Fix the flickering issue because of primary |uv_rect| cropping even though > an overlay attemption fails mainly due to limited pixel rate (e.g. 4k video). > - Fix wrong |uv_rect| calculation which causes atomic page flip failures with > "No space left on device (28)" error. > > Original change's description: > > viz: Turn off or crop primary plane when possible > > > > For many fullscreen cases, the rest of the renderpass quads are solid > > black quads. In that case, we can turn off the primary plane, and not > > having to scan out an full screen of black pixels saves a good deal of > > power. In other cases, we can crop down the primary plane down to the > > content rectangle (youtube channel logo, for example) and save power. > > > > Averaging battery draw measurement over about a minute of play back > > for a fullscreen youtube video, power savings for soraka is around > > 500mW (6.5W to 6.0W) and 300mW for kevin (4.4W to 4.1W), when we > > turn off the primary plane. > > > > Change-Id: I7946d7acab334bffdf2bb49a1d9c424cf46fc610 > > Reviewed-on: https://chromium-review.googlesource.com/c/1045256 > > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > > Commit-Queue: Kristian H. Kristensen <hoegsberg@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#600415} > > TEST=SingleOverlayOnTopTest.AllowVideoNormalTransformWithOutputSurfaceOverlay > > Bug: 896945, 900373 > Change-Id: I88db118dbc3ca2fed8c8eaf1f9a4337cdae12990 > Reviewed-on: https://chromium-review.googlesource.com/c/1346997 > Commit-Queue: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Kristian H. Kristensen <hoegsberg@chromium.org> > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Cr-Commit-Position: refs/heads/master@{#611484} TBR=dongseong.hwang@intel.com,dcastagna@chromium.org,hoegsberg@chromium.org Bug: 896945, 900373 Change-Id: Iec6d2812d94c286c495d90d8952b65ecffa4c010 Reviewed-on: https://chromium-review.googlesource.com/c/1363734 Reviewed-by: Kazuhiro Inaba <kinaba@chromium.org> Reviewed-by: Daniele Castagna <dcastagna@chromium.org> Commit-Queue: Daniele Castagna <dcastagna@chromium.org> Cr-Original-Commit-Position: refs/heads/master@{#614732}(cherry picked from commit e1809a76301288b5b74afc6a785cc2c15da0f938) Reviewed-on: https://chromium-review.googlesource.com/c/1369760 Cr-Commit-Position: refs/branch-heads/3626@{#193} Cr-Branched-From: d897fb137fbaaa9355c0c93124cc048824eb1e65-refs/heads/master@{#612437} [modify] https://crrev.com/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7/components/viz/service/display/gl_renderer_unittest.cc [modify] https://crrev.com/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7/components/viz/service/display/overlay_processor.cc [modify] https://crrev.com/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7/components/viz/service/display/overlay_processor.h [modify] https://crrev.com/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7/components/viz/service/display/overlay_strategy_single_on_top.cc [modify] https://crrev.com/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7/components/viz/service/display/overlay_strategy_underlay.cc [modify] https://crrev.com/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7/components/viz/service/display/overlay_unittest.cc
,
Dec 19
Here's a summary of the rules that were executed: - OnlyMergeApprovedChange: Rule Failed -- Revision a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7 was merged to refs/branch-heads/3626 branch with no merge approval from a TPM! Please explain why this change was merged to the branch! - AcknowledgeMerge: Notification Required --
,
Dec 19
The following revision refers to this bug: https://chromium.googlesource.com/chromium/src.git/+/a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7 Commit: a0ffbe6455fc7eb3cfa1a80d2429ffc064b8ece7 Author: kinaba@google.com Commiter: kinaba@chromium.org Date: 2018-12-10 12:14:50 +0000 UTC Revert "Reland "viz: Turn off or crop primary plane when possible"" This reverts commit 1ceb07933f73defaa73c30153fcbe812ad1965e0. Reason for revert: ARC and Crostini on ARM broken (crbug.com/896945#c17) Original change's description: > Reland "viz: Turn off or crop primary plane when possible" > > This is a reland of I7946d7acab334bffdf2bb49a1d9c424cf46fc610 and > the following bug fixes as follows; > "viz: Fix wrong uv_rect calculation" 4bf9d01dbf34f17c899095a937dabbb884ec87fe > "viz: add unittests for CL:1291590" f8dfbe74c1f5c79051461cdf448abf7f8188336f > > Additional fix: > - Fix the flickering issue because of primary |uv_rect| cropping even though > an overlay attemption fails mainly due to limited pixel rate (e.g. 4k video). > - Fix wrong |uv_rect| calculation which causes atomic page flip failures with > "No space left on device (28)" error. > > Original change's description: > > viz: Turn off or crop primary plane when possible > > > > For many fullscreen cases, the rest of the renderpass quads are solid > > black quads. In that case, we can turn off the primary plane, and not > > having to scan out an full screen of black pixels saves a good deal of > > power. In other cases, we can crop down the primary plane down to the > > content rectangle (youtube channel logo, for example) and save power. > > > > Averaging battery draw measurement over about a minute of play back > > for a fullscreen youtube video, power savings for soraka is around > > 500mW (6.5W to 6.0W) and 300mW for kevin (4.4W to 4.1W), when we > > turn off the primary plane. > > > > Change-Id: I7946d7acab334bffdf2bb49a1d9c424cf46fc610 > > Reviewed-on: https://chromium-review.googlesource.com/c/1045256 > > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > > Commit-Queue: Kristian H. Kristensen <hoegsberg@chromium.org> > > Cr-Commit-Position: refs/heads/master@{#600415} > > TEST=SingleOverlayOnTopTest.AllowVideoNormalTransformWithOutputSurfaceOverlay > > Bug: 896945, 900373 > Change-Id: I88db118dbc3ca2fed8c8eaf1f9a4337cdae12990 > Reviewed-on: https://chromium-review.googlesource.com/c/1346997 > Commit-Queue: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Dongseong Hwang <dongseong.hwang@intel.com> > Reviewed-by: Kristian H. Kristensen <hoegsberg@chromium.org> > Reviewed-by: Daniele Castagna <dcastagna@chromium.org> > Cr-Commit-Position: refs/heads/master@{#611484} TBR=dongseong.hwang@intel.com,dcastagna@chromium.org,hoegsberg@chromium.org Bug: 896945, 900373 Change-Id: Iec6d2812d94c286c495d90d8952b65ecffa4c010 Reviewed-on: https://chromium-review.googlesource.com/c/1363734 Reviewed-by: Kazuhiro Inaba <kinaba@chromium.org> Reviewed-by: Daniele Castagna <dcastagna@chromium.org> Commit-Queue: Daniele Castagna <dcastagna@chromium.org> Cr-Original-Commit-Position: refs/heads/master@{#614732}(cherry picked from commit e1809a76301288b5b74afc6a785cc2c15da0f938) Reviewed-on: https://chromium-review.googlesource.com/c/1369760 Cr-Commit-Position: refs/branch-heads/3626@{#193} Cr-Branched-From: d897fb137fbaaa9355c0c93124cc048824eb1e65-refs/heads/master@{#612437} |
|||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||
Comment 1 by rbyers@chromium.org
, Nov 1