:lightsofapollo had previously mentioned that the new taskcluster/etc. should let us run the GELAM tests pretty easily. Given the recent platform breakages experienced and the fact that the GELAM tests already work, run relatively quickly, and detected these, it would probably be good to get them running for gecko builds. When the front-end integration tests start happening/get fancified again (bug 975588) a lot of this coverage will become redundant. However, there might be an argument for keeping these more thorough tests around since they also will perform a bit like a stress test for aspects of TCPSocket, etc. (For example, we discovered/see the TCPSocket thread-race bug 1059081. Unfortunately that same bug means we'd need to hide the tests until that bug is fixed.) :lightsofapollo, can you let us know the steps we'd need to do to get this happening or point us at a wiki page that documents the procedure? Thanks!

[Tracking Requested - why for this release]:

Marking this priority as it would help us catch platform regressions -- need some help from the automation team to help out with the configuration.

Will fill in some more details... the TC only bits are documented http://docs.taskcluster.net/ but how to integrate this into the usual sheriff workflow is something we are still in the process of figuring out. :asuth do you care if the tests run and are hidden or do you want full privileges? In addition to https://wiki.mozilla.org/Sheriffing/Job_Visibility_Policy we have some extra requirements so far for non build bot jobs (they should be listed above too): - must be classifiable (new word for staring) - this means using the TBPL format for output and using buildbot style steps https://github.com/mozilla-b2g/gaia/blob/master/build/docker/gaia-taskenv/bin/buildbot_step (this is a workaround for treeherder log parsing) - RyanVM will review symbols

(In reply to James Lal [:lightsofapollo] from comment #3) > :asuth do you care if the tests run and are hidden or do you want full > privileges? In addition to > https://wiki.mozilla.org/Sheriffing/Job_Visibility_Policy we have some extra > requirements so far for non build bot jobs (they should be listed above too): Baby steps. I think it would make sense to start out: - hidden - maybe only keyed off of b2g-desktop nightly builds - ideally something that could be manually requested on try builds (that might be too hard?) so IndexedDB and MozTCPSocket changes can get in on this earlier. I would monitor during this time to be a bit more proactive about jumping on platform regressions. I think we would eventually want to be unhidden and running on all builds when: - we're not seeing intermittent failures - we're absolutely sure that the way we're getting our b2g-desktop instance is actually downloading the right thing - we're seeing good bang/buck. Specifically, our tests run fast enough and either provide useful coverage the integration tests don't provide, or they are or seem more trustworthy than our integration tests, allowing us to get faster backouts. (It wouldn't surprise me if people ignore Gij oranges on the try server.) In the event the bang/buck isn't there, this might be something we end up just running for us on a nightly basis and per-GELAM-commit basis. It sounds like the immediate next steps for us are to address the logging issues and to follow the TC instructions and cargo cult an existing docker setup, ideally one that knows how to get b2g-desktop already. I definitely won't be able to get to this before next week.

Turning this into meta bug

Okay- GELAM is now configured to use TC https://treeherder.allizom.org/ui/#/jobs?repo=try-taskcluster ... The timeline this month is very tight with bitbar (our device lab) coming online and setting up try + gecko + taskcluster so its unlikely we will pick up any of these bugs this month myself. The hope here is after we are done the above asks should be fairly easy to do in a self-service manner.

Firefox OS is not being worked on