943018 - Firefox OS will black out and become nonresponsive if Marionnette python script runs while the device is still booting up

Status: RESOLVED FIXED

(Firefox OS Graveyard :: General, defect)

Description

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

This is one possible issue that stephend and I thought we saw in regards to the nonresponsive buri devices.

The resulting symptom was that the device looked like it was turned off; trying to hit the power button on/off doesn't seem to wake the device.  You can still adb shell to the device.


I need to come up with better STRs than removing the sleep setting in the main script for hamachi scripts before the python script runs.

Also need to logcat this information and give clearer info.

I'm logging this bug so that I don't forget to do this.

Comment 1

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Work around is to wait for the device to fully boot up before running the python script.

Comment 2

[Dave Hunt [:davehunt] [he/him] ⌚BST]

Depending on the Python script, one of the first things that happens is stopping/starting b2g. If it's UI tests the profile and persistent storage is also removed. It would be good to know at what point in the Python script we reach. As you suggest, the logcat would be useful, as would the eventual Python traceback.

Comment 3

[Stephen Donner [:stephend] Not actively reading bugmail]

I/we believe we have a potential fix -- at least, we seem to have (with :fabrice's help) fixed one issue/cause of this issue:

1) In our "Erase Cache, Complete Gecko and Gaia Flash" script in Jenkins, we were hitting a race condition where B2G would start up when it shouldn't have (early), while reading in /system/b2g/webapps/webapps.json, combine that with the one in gaia/profile/webapps/webapps.json, which we push to the device (custom apps for testing Marketplace dev, UI test, etc.)

2) Fabrice added a simple:

"adb stop b2g" call in the script just before we push those, seen here in its final form, with the fix:

# XXX - Fabrice's _awesome fix to stop b2g  -- to prevent a race condition -- below_
adb shell stop b2g

echo -e "\n\033[1;30mInstalling new gaia webapps\033[0m"
adb shell mkdir -p /system/b2g/defaults/pref
adb push gaia/profile/webapps /system/b2g/webapps
adb push user.js /system/b2g/defaults/pref
adb push gaia/profile/settings.json /system/b2g/defaults

3) Without that fix, we were getting lucky, essentially -- B2G would start, and push/read in the right files it needs, before it gets to the point at which it checks that /system/b2g/webapps and its content was correct.

4) We've done a day full of runs with the above fix, and I'm optimistic (but trying not to jinx it!) that it /might/ fix what we saw, here!

Comment 4

[Stephen Donner [:stephend] Not actively reading bugmail]

...and Dave's done some further bulletproofing (last 2-3 weeks or so), so I'm resolving this as fixed.