we should only have one set of configs/factories for our mobile device testing. The n810s are currently running on the older config because they don't have the needed tools to work with the new factories. The n810s and builders+pollers are already in the configuration for the new mobile master code, so this is an image + imaging issue. They are going to need(in addition to current setup): -mercurial -cvs -dropbear instead of openssh -nginx built from source (no packages) -startup script that does all device specific setup -rf room not Mozilla-Build They should use staging-mobile-master:9020 for the default master This bug is to track the work of putting them into production. Bug 546702 is to track the actual image modifications

Looks like you should own this, John.

not working on this for a bit

Morphing.

working on this.

in addition to the packages on the n900 refimage doc, we need to build python as it does not seem to have an installable version in the repository anymore. I am going to use python 2.5.4 to keep parity with the n900

i have the base image done. I am going to work on getting our n900 start up scripts to work on the n810 making the changes needed. Because we don't have to worry about a two part imaging process, I am going to bake the scripts into the image through an HG checkout on the refimage and symlinks. My work is noted at https://wiki.mozilla.org/ReferencePlatforms/Test/Maemo

It looks like this image is working so far but I still need to configure nginx. I am going to rework the startup scripts to make it more similar to the n900. I plan to do the following in the startup procedure: 1. set up device to disable as much power saving as possible 2. fork a process that will check for uptime limit being exceeded that will run the entire time the device is up 3. fork a process that will ping the ?-mobile-master to make sure connections stay alive 4. check being able to ping ?-mobile-master and block buildbot until this happens with a sensible limit I am also going to have a script that will reboot the device intelligently like on the n900 (reboot-user).

there are a couple issues with this patch that I need to resolve before I ask for review. 1. uptime-check.py wasn't being launched. I don't know why that is but I am looking into it. 2. i need to test the uptime offset file in the wild. I did test this logic locally and it worked, but funny things happen on device 3. ensure that date/time are set properly on boot -- likely going to copy the ntp logic from the older n810 startup script

I have compiled ntpdate for the n810s. I will be testing this tomorrow morning. Because this requires network to be up before it will work, I am going to make the buildbot start script block on getting a network connection before running the ntpdate program

this patch: -adds new script to create an image on the flashing machine -modifies imaging behaviour to better deal with card mounting/unmounting -heartbeat type checks done on slaves -- no longer dependent on master<->slave connection

Comment on attachment 445812 [details] [diff] [review] new n810 service and flashing tools + echo "WARNING: DO NOT SELECT A NON-SD CARD DRIVE LETTER" + find /dev -maxdepth 1 -name sda -o -name sd? -print Hm, what if we have the backup drive connected? You prefer just hardcoding sda? for i in "$@" ; do - BATCH='yes' ./moz-image.sh $ROOTFSDIR /dev/sd${i} maemo-n810-ref & + if [[ "$i" == "a" || "$i" == "A" ]] ; then + echo Ignoring /dev/sda + fi + BATCH='yes' ./moz-image.sh $ROOTFSDIR /dev/sd${i} maemo-n810-XX & You're not actually ignoring it here. You might want a continue or put the BATCH=... in an else. +NTPCOMMAND = '/usr/local/bin/ntpdate' /me doesn't mention ntpd -q =D Really, the not ignoring sda is the only thing I found here. Otherwise, *stamp*.

This patch improves the output from the N810 imaging scripts I still need to write the moz-verify.sh script but don't want to block these changes landing

Comment on attachment 446057 [details] [diff] [review] more improvements SPIN! RPLUS

even more improvements

Comment on attachment 446100 [details] [diff] [review] more *STAMP* with the verbal s,SDDEV,CARD, fix

Anything left to do here?

they need to be connected to pmm02/smm02

I am going to get some patches together to get the new n810 image into production on mobile-rw

This patch enables unit test and talos for n810 using the new image and on the new mobile master configs. I have not removed the old configs in this patch and will submit a patch with the contents of: hg rm mobile mobile-staging hg mv mobile-rw mobile While none of the tests are going green, they are going orange in the same way that production is.

http://staging-mobile-master.build.mozilla.org:8020/builders/n900-gtk%20mozilla-1.9.2%20unit%20reftest/builds/0 sample n810 run using a 1.9.2 build on the new master.

Comment on attachment 457611 [details] [diff] [review] buildbot-configs v1 I'd prefer that the staging tarballs be uncommented by default, and pointing to http://10.250.48.136/maemo/*.tar.bz2 (maemo5 for n900s).

Comment on attachment 457611 [details] [diff] [review] buildbot-configs v1 http://hg.mozilla.org/build/buildbot-configs/rev/2573958ba012 landed. there are no slaves in the pool at this point

I have done the switch over. I have switched/moved as many devices as i could [1]. I have disabled polling on the old configs and will reimage remaining devices to point to the new configs. Devices that were moved that I wasn't able to reboot will reboot after their 10 hours is up on their own. [1] maemo-n810-%02d % [9, 18, 21, 24, 26, 30, 31, 35, 37, 38, 52, 55, 56, 58, 59, 61, 70, 78]