For the first comment, I'm seeing:


(INFO) 2016-03-10 13:51:08,689 cmd_helper._ValidateAndLogCommand:153  [host]> /b/build/slave/Android_One_Perf__2_/build/src/third_party/catapult/devil/bin/deps/linux2/x86_64/bin/adb -s AG8604420IG4GGC shell '( logcat -d -t 500 );echo %$?'
Traceback (most recent call last):
  File "/b/build/slave/Android_One_Perf__2_/build/src/third_party/android_platform/development/scripts/stack", line 258, in <module>
    sys.exit(main(sys.argv[1:]))
  File "/b/build/slave/Android_One_Perf__2_/build/src/third_party/android_platform/development/scripts/stack", line 194, in main
    constants.CheckOutputDirectory()
  File "/b/build/slave/Android_One_Perf__2_/build/src/third_party/android_platform/development/scripts/../../../../build/android/pylib/constants/__init__.py", line 264, in CheckOutputDirectory
    raise Exception('Neither CHROMIUM_OUTPUT_DIR nor CHROMIUM_OUT_DIR '
Exception: Neither CHROMIUM_OUTPUT_DIR nor CHROMIUM_OUT_DIR has been set
Traceback (most recent call last):
  File "/b/build/slave/Android_One_Perf__2_/build/src/build/android/tombstones.py", line 273, in <module>
    sys.exit(main())
  File "/b/build/slave/Android_One_Perf__2_/build/src/build/android/tombstones.py", line 255, in main
    constants.CheckOutputDirectory()
  File "/b/build/slave/Android_One_Perf__2_/build/src/build/android/pylib/constants/__init__.py", line 264, in CheckOutputDirectory
    raise Exception('Neither CHROMIUM_OUTPUT_DIR nor CHROMIUM_OUT_DIR '
Exception: Neither CHROMIUM_OUTPUT_DIR nor CHROMIUM_OUT_DIR has been set

Are these separate issues? Should this one be duped against  bug 594105  and a new one filed for the ref build?

Latest failing test is power.tough_ad_cases.reference here: 
https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%282%29/builds/2397

Disabling test.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/db5df850029cd0578e9ec391fd3de28eb5ba5c1c

commit db5df850029cd0578e9ec391fd3de28eb5ba5c1c
Author: picksi <picksi@google.com>
Date: Fri Mar 11 18:18:15 2016

Disabling power.touch_ad_cases to turn Android One green.

BUG= 593973 
CQ_EXTRA_TRYBOTS=tryserver.chromium.perf:android_s5_perf_cq;tryserver.chromium.perf:winx64_10_perf_cq;tryserver.chromium.perf:mac_retina_perf_cq;tryserver.chromium.perf:linux_perf_cq

Review URL: https://codereview.chromium.org/1783233004

Cr-Commit-Position: refs/heads/master@{#380672}

[modify] https://crrev.com/db5df850029cd0578e9ec391fd3de28eb5ba5c1c/tools/perf/benchmarks/power.py

Temporarily relabeling Performance-Sheriff-BotHealth issues to Performance-BotHealth while we wait to migrate Performance-BotHealth issues to Performance-Sheriff-BotHealth.

Android One bot is still unhappy, complaining about:

 File "/b/build/slave/Android_One_Perf__2_/build/src/build/android/pylib/constants/__init__.py", line 264, in CheckOutputDirectory
    raise Exception('Neither CHROMIUM_OUTPUT_DIR nor CHROMIUM_OUT_DIR '

Adding john as I see another CHROMIUM_OUTPUR_DIR bug assigned to him.

Here's the link to the bot:

https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%282%29?numbuilds=200

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/c91b537eb16a73f49e61b08e8d24cfbc4649ed75

commit c91b537eb16a73f49e61b08e8d24cfbc4649ed75
Author: agrieve <agrieve@chromium.org>
Date: Mon Mar 14 19:04:59 2016

test_runner.py: Auto-detect output dir when running on bots

BUG= 594105 , 593973 

Review URL: https://codereview.chromium.org/1799073004

Cr-Commit-Position: refs/heads/master@{#381032}

[modify] https://crrev.com/c91b537eb16a73f49e61b08e8d24cfbc4649ed75/build/android/pylib/constants/__init__.py

The CHROMIUM_OUTPUR_DIR bug is fixed. It was only triggered by the error handling logic though, so the underlying failure still exists. I don't see any obvious stack traces in the output though.

 Issue 595416  has been merged into this issue.

blink_perf.canvas is still failing on Android One: https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%282%29
I'm trying a patch which should disable it only on Android One: https://codereview.chromium.org/1806843004
The benchmark has been failing since it was enabled on Android: crrev.com/379686

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/69b38e03d4714e45603211a9b8b198d39b6e9546

commit 69b38e03d4714e45603211a9b8b198d39b6e9546
Author: petrcermak <petrcermak@chromium.org>
Date: Thu Mar 17 17:53:01 2016

Disable blink_perf.canvas on Android One

Rationale: The benchmark is consistently failing on the "Android One
Perf (2)" bot (see the associated bug).

BUG= 593973 
TBR=skyostil
CQ_EXTRA_TRYBOTS=tryserver.chromium.perf:android_s5_perf_cq;tryserver.chromium.perf:winx64_10_perf_cq;tryserver.chromium.perf:mac_retina_perf_cq;tryserver.chromium.perf:linux_perf_cq

Review URL: https://codereview.chromium.org/1804413004

Cr-Commit-Position: refs/heads/master@{#381741}

[modify] https://crrev.com/69b38e03d4714e45603211a9b8b198d39b6e9546/tools/perf/benchmarks/blink_perf.py

Now that the benchmark has been disabled on Android One, I'm reducing this bug's priority and reassigning it to the author of the CL that enabled it (http://crrev.com/379686).

petrcermak, could you copy&paste error log before #13. It's because external contributor don't have the permission.
This build is good candidate; https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%282%29/builds/2464

The error log (especially the logcat) is quite long. Are you sure you can't access it through https://build.chromium.org/p/chromium.perf/builders/Android%20One%20Perf%20%282%29/builds/2464/steps/blink_perf.canvas/logs/stdio? I have no trouble accessing it even in incognito mode.

#16 - nope, only googler can access it :'(

Here it is (attached).

Deleted: blink_perf_canvas_failure_log 58.0 KB

blink_perf_canvas_failure_log 58.0 KB View Download

The log says out of memory crashes chromium. It's strange. canver_perf doesn't require exceptional memory space...

I/ActivityManager(  555): Process android.process.media (pid 3079) has died
E/lowmemorykiller(  133): Error opening /proc/2039/oom_score_adj; errno=2
I/ActivityManager(  555): Process android.process.acore (pid 2039) has died
.......
W/ActivityManager(  555): Scheduling restart of crashed service com.google.android.youtube/com.google.android.apps.youtube.app.offline.transfer.OfflineTransferService in 1000ms
I/WindowState(  555): WIN DEATH: Window{1779727e u0 com.android.launcher/com.android.launcher2.Launcher}
I/ActivityManager(  555): Process com.android.launcher (pid 1212) has died
I/ActivityManager(  555): Low on memory:
.......
I/ActivityManager(  555):   Free RAM: 94416 kB
I/ActivityManager(  555):   Used RAM: 357403 kB
I/ActivityManager(  555):   Lost RAM: -27251 kB

gentle ping.. 
could someone please check and update the issue.

According to #19, does this mean that there is memory regression on Android?

Ping dongseong.hwang@. What to do about the memory regression?

nednguyen@, sorry for delaying replying. I cannot debug it more because I have only Nexus 5, on which I can not reproduce it.

Sami mentioned it's memory issue. It might be because old Nexus one doesn't have enough memory.
https://codereview.chromium.org/1804413004#msg20

I'm not sure if it's regression. Recently  issue 496707  enabled canvas perf test on Android. Before the issue, Android never have been covered properly.

IMO, this issue is fixed, and Nexus one is just the device which cannot support the perf test.

excluded nexus one which don't have memory enough to support canvas perf test.

Issue 650070 has been merged into this issue.