Chrome OS UI fails to reach login screen on all arm platforms (daisy, peach, nyan, oak, veyron) since R51-8041.0.0 |
||||||
Issue descriptionChrome Version: 51.0.2672.0 Chrome OS Version: >= R51-8041.0.0 Chrome OS Platform: any arm (daisy, peach, nyan, oak) Steps To Reproduce: (1) boot (2) (3) Expected Result: boot to Login screen Actual Result: chrome splash shows, then chrome UI shows breifly, followed by a black screen that blinks occasionally How frequently does this problem reproduce? (Always, sometimes, hard to reproduce?) Always. What is the impact to the user, and is there a workaround? If so, what is it? These versions are DOA. Please provide any additional information below. Attach a screen shot or log if possible. This bug is branched from from issue 593459 (see comment c23), because that is an auto-filed bug and way to noisy. That comment is repeated here: ----- The Chrome OS UI fails boot to login screen on all arm devices since R51-8041.0.0 It looks like this is caused by a recent change to the chromeos-chrome package on 8041 is broken. I took 8041.0.0, built my own 51.0.2672.0 chrome using the simple chrome workflow, and deployed it to 8041.0.0 and the UI works. I'm not sure what the difference is between building chrome from ebuild inside the cros chroot and using chrome-sdk. However, there were some recent changes to the chromeos-chrome ebuild: https://chromium-review.googlesource.com/#/c/330006/ For some more analysis see: http://crosbug.com/p/51023. ----- One new fact: I can boot to login screen if I disable the Chrome sandbox by adding "--no-sandbox" to /etc/chrome_dev.conf.
,
Mar 10 2016
I tested https://chromium-review.googlesource.com/#/c/330006/ for daisy. https://uberchromegw.corp.google.com/i/chromiumos.tryserver/builders/release/builds/2584 Inside the artifacts, I checked the chrome binary from stripped-packages.tar and The flags I added in the CL was used in this trybot and it passes.
,
Mar 10 2016
,
Mar 10 2016
Its reproduced on Veyron-minnie too.
,
Mar 11 2016
I tried building a peach_pit image locally, and the image worked fine. Anyone succeeded to reproduce this with locally built binary?
,
Mar 11 2016
1. Issue is seen in M51: 51.0.2672.0/8049.0.0- Dev- Channel Peppy, Daisy, Blaze, Mighty as well 2.Issue is not seen in M51: 51.0.2672.0/8049.0.0- Dev- Channel Gnawty, Quawks, Paine, Parrot as able to sign in to user and work on all the apps.
,
Mar 11 2016
,
Mar 11 2016
I have a CL https://chromium-review.googlesource.com/#/c/332221/ that should fix this issue.
,
Mar 12 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/overlays/chromiumos-overlay/+/c856736b4db1a40c760a8979b3f827da2bf08eea commit c856736b4db1a40c760a8979b3f827da2bf08eea Author: Yunlian Jiang <yunlian@google.com> Date: Fri Mar 11 00:18:54 2016 chromeos-chrome: do not split hot/cold text sections for arm. This CL disables split hot/cold text section for arm. This is because the splitting causes the boot time error for chrome on arm boards. BUG= chromium:593719 TEST=build chrome locally without these flags, the login screen appears. Change-Id: I2ff59d0acf150f696db021f566ab3afefd315d02 Reviewed-on: https://chromium-review.googlesource.com/332221 Commit-Ready: Yunlian Jiang <yunlian@chromium.org> Tested-by: Yunlian Jiang <yunlian@chromium.org> Reviewed-by: Mike Frysinger <vapier@chromium.org> [modify] https://crrev.com/c856736b4db1a40c760a8979b3f827da2bf08eea/chromeos-base/chromeos-chrome/chromeos-chrome-9999.ebuild
,
Mar 12 2016
Should we manually uprev chromeos-chrome? The fix was merged in 8053 but I tried 8054 image and it still failed with the same issue. I checked the chrome binary in 8054 and it was compiled with the flag which should be removed in the fix. Maybe the builder is still pulling the broken chromeos-chrome prebuilt?
,
Mar 13 2016
I think so, this would make the canary bot green. Can anyone do that? Thanks.
,
Mar 13 2016
PFQ is more or less clean. Only alex failing. Hopefully it will pass next time. Talked to vapier about this. He said manually uprevving the ebuild will not work because the PFQ is the only one generating the binpkg. So, to fix this we need to modify the CURRENT ebuild. Need to create a CL with the change to the current ebuild and chump it. I will wait a little longer for the PFQ to go green and, if not, do the other chnage.
,
Mar 13 2016
PFQ is not going green. So I will work on putting the fix on the current ebuild...
,
Mar 13 2016
The following revision refers to this bug: https://chromium.googlesource.com/chromiumos/overlays/chromiumos-overlay/+/be131e7f5322de6f69db88bf17f0c0894af93be5 commit be131e7f5322de6f69db88bf17f0c0894af93be5 Author: Luis Lozano <llozano@chromium.org> Date: Sun Mar 13 17:46:42 2016 Force applying of fix for chromium:593719 This puts the fix for this bug in the current ebuild. The fix is in https://chromium-review.googlesource.com/332221. We need this fix urgently and we cannot wait for this change to be uprevved by the PFQ. So, I am applying directly to the current ebuild. BUG= chromium:593719 TEST=tested previously. Just started an emerge by hand to make sure I did not break anything. Change-Id: I3f23621daa7a6e0f95127771ef498a615a847015 Reviewed-on: https://chromium-review.googlesource.com/332296 Reviewed-by: Luis Lozano <llozano@chromium.org> Commit-Queue: Luis Lozano <llozano@chromium.org> Tested-by: Luis Lozano <llozano@chromium.org> [modify] https://crrev.com/be131e7f5322de6f69db88bf17f0c0894af93be5/chromeos-base/chromeos-chrome/chromeos-chrome-51.0.2672.0_rc-r1.ebuild
,
Mar 14 2016
CL:332221 first appears in 8061.0.0. I manually verified that 8061.0.0 boots to UI on oak. Marking this one Verified.
,
Mar 14 2016
thanks for the verification. waiting for the rest of the canary builders...
,
Mar 14 2016
I have verified some of the canary builders are also fixed. (some others have not finished) |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by jorgelo@chromium.org
, Mar 10 2016