725544 - (android_4.0_testing) [tracking bug] Android 4.0 testing

Status: RESOLVED FIXED

(Infrastructure & Operations Graveyard :: CIDuty, task, P4)

Description

[Armen [:armenzg]]

We're looking on having panda boards to be doing Android 4.0 testing.

We will be having requirements gathered in this wiki page (wip):
https://wiki.mozilla.org/ReleaseEngineering/Projects/PandaBoards

Comment 1

[Armen [:armenzg]]

We're not ready to be getting rid of the foppies setup but to be able to scale we will be switching to do them with Linux VMs - bug 725534.
We might have to live with both mini and Linux foopies for now.

Comment 2

[Armen [:armenzg]]

From https://wiki.mozilla.org/Mobile/Testing/02_15_12#Panda_Boards

ctalbert/jmaher/wlach - have had minimal luck getting pandaboards to boot and run tests (mochitest and reftest samples so far). ctalbert and jmaher can't even get an image to boot.
* require proper power supply
* requiring building from tip of ICS
* require panda video driver
* possibly require specific type(s) of sdcard 

concerns:
* the boards sometimes become unstable (much more than the tegras)
* armenzg: do we have an alternative to the panda boards?
** blassey: no fallback plan

Comment 3

[Armen [:armenzg]]

https://wiki.mozilla.org/Auto-tools/Projects/Pandaboard_Setup

Comment 4

[Armen [:armenzg]]

From: https://wiki.mozilla.org/Mobile/Testing/02_29_12#Panda_Boards
* [P1] [NEW]: bug 731669 (ctalbert) we can't reboot a panda board without using a PDU to reboot it
** hopefully the ES boards make things better 
** jmaher does not hit this issue
* [P1] [NEW]: bug 731670 (ctalbert) MAC address gets re-created on every reboot
*high enough screen resolution
** are there more details about this? 
* [P1] [DONE] bug 729323 - board racked on haxxor for releng to mess with
** armenzg to verify that I can reach the board 
* [P1] [DONE] bug 725846 - 8 boards have been received
* [P2] [DONE]: bug 725859 - document process for re-imaging and remote re-imaging

concerns:
* IT is concerned about not having remote re-imaging
* the boards sometimes become unstable (much more than the tegras)
** each a-team member has a different experience on how to make things work
** hopefully the ES boards make things better 
* armenzg: do we have an alternative to the panda boards?
** blassey: no fallback plan

Comment 5

[Armen [:armenzg]]

changing priority just to make "my bugs" assigned easier to look at.

Comment 6

[William Lachance (:wlach)]

Found a problem with installing stuff on Android ICS which will affect the pandaboards. Already fixed, but we'll need to be sure to use a new sutagent.

Comment 7

[Armen [:armenzg]]

(In reply to William Lachance (:wlach) from comment #6)
> Found a problem with installing stuff on Android ICS which will affect the
> pandaboards. Already fixed, but we'll need to be sure to use a new sutagent.

Is that change already on SUT agent 1.09?

Comment 8

[Joel Maher ( :jmaher ) (UTC -8)]

yes, this change will be in 1.09

Comment 9

[William Lachance (:wlach)]

We will need to fix bug 763497 (or come up with a workaround), otherwise the logcat code won't work with the Panda. My inclination is just to fix.

Comment 10

[William Lachance (:wlach)]

Watcher needs a new permission to work properly with ICS.

Comment 11

[Ed Morley [:emorley]]

I've just been reviewing what's hidden on TBPL for Android on Panda:

* mozilla-central:
mochitest-1
mochitest-2
plain-reftest-1
plain-reftest-2
plain-reftest-3
plain-reftest-4
robocop
trobocheck
trobocheck-2

* mozilla-inbound:
(same as mozill-central)

* try:
mochitest-1
mochitest-2
plain-reftest-1
plain-reftest-2
plain-reftest-3
plain-reftest-4
robocop

Look at the results for some of these hidden runs...

* mochitest-2 looks green to me:
https://tbpl.mozilla.org/?tree=Mozilla-Inbound&noignore=1&jobname=Android%204.0%20Panda%20mozilla-inbound%20opt%20test%20mochitest-2

* trobocheck$ (not trobocheck2) also looking green:
https://tbpl.mozilla.org/?tree=Mozilla-Inbound&noignore=1&jobname=Android%204.0%20Panda%20mozilla-inbound%20talos%20remote-trobocheck

* mochitest-1 is indeed orange (presume bug 781792):
https://tbpl.mozilla.org/?tree=Mozilla-Inbound&noignore=1&jobname=Android%204.0%20Panda%20mozilla-inbound%20opt%20test%20mochitest-1

* robocop also orange (can't find a bug on this):
https://tbpl.mozilla.org/?tree=Mozilla-Inbound&noignore=1&jobname=Android%204.0%20Panda%20mozilla-inbound%20opt%20test%20robocop

* reftest1-4:
No longer being run.

As such, I have:
* Unhidden mochtitest-2 + trobocheck[^2] on m-c, inbound and try
* Hidden trobocheck2 on Try, to bring it in sync with the other trees

Comment 12

[Ed Morley [:emorley]]

s/Look/Looking/

Comment 13

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

bug#860028 is expected to reduce intermittent errors.

Comment 14

[Justin Wood (:Callek)]

Panda's have been running Android 4.0 tests in production for months now, despite the few dependancies I think this has outlived its usefulness.