This will cut our Amazon costs. Developers don't care about these builds. I'm not working on this specifically but just filing it.

per mtg w/ctalbert, jgriffin and others just now, developers dont seem to focus on keeping these tests green, so fixing them only to watch them get broken seems pointless. However, lets leave these enabled for now, until we get formal approval to drop. Whether developers should be backing out / fixing bustages, or whether we all decide to drop panda-on-b2g is a bigger question. Grabbing while we socialize this topic with developers+QA.

See bug 868156 for broken builds on Gaia-Master.

per meetings w/fabrice, faramarz, dylan and james lal, they want to now investigate how much work is needed to fix and unhide these tests, and keep these tests green. Meanwhile, do not proceed further with disabling panda builds. jlal is actively investigating an initial list of panda bugs - more info soon. If jlal can fixup these tests to be green again, we hope to reenable these disabled b2g-panda-tests again.

Having a gaia tryserver available is an absolute prerequisite to unhiding any panda tests, since otherwise most developers will have zero ability to debug or reproduce any failures. The only tests that were ever running on the panda were the gaia-ui-tests, which are the same tests we currently have running on b2g desktop builds. We never got a stable set of tests on the pandas, primarily due to problems with the panda builds, so before any tests could be unhidden, some non-trivial time would have to be spent fixing the tests (assuming all the build problems are taken care of).

James: we were wondering the following: * how is your panda-greening work going? * is there a bug tracking that work, or is this it?

I just got my test vm without any panda access- My primary priority is to get something working for b2g-desktop in the next 8 weeks.

per mtg w/dylan, faramarz and ladamski, the panda builds and tests are officially no longer needed. Dylan also sent a post to dev-gaia@. We now have consensus on next steps: * RelEng to stop generating panda builds across all branches, as well as stop running any remaining b2g-on-panda tests not already disabled. * Once all b2g-panda builds+tests are stopped, these machines should be freed up for use elsewhere within RelEng infrastructure. My work is done here, so unassigning myself from this bug.

(In reply to John O'Duinn [:joduinn] from comment #7) > per mtg w/dylan, faramarz and ladamski, the panda builds and tests are > officially no longer needed. Dylan also sent a post to dev-gaia@. We now > have consensus on next steps: > > * RelEng to stop generating panda builds across all branches, as well as > stop running any remaining b2g-on-panda tests not already disabled. > * Once all b2g-panda builds+tests are stopped, these machines should be > freed up for use elsewhere within RelEng infrastructure. > > My work is done here, so unassigning myself from this bug. Lets wait until end of week (26jul) to give people time to raise objections to the dev-gaia post. If no objections by end-of-week, we'll drop the axe. As part of disabling, we should also cleanup code, maybe as part of this bug. Separately, we should also reallocate any b2g-pandas to help us better handle fennec-panda load. Filed bug#897592 to track.

I'm developping application to control hardware via Gaia/Gecko. For the development, Pandaboard is useful because it has GPIO. Now I added navigator.mozGpio and using it. So I don't want to disable it.

Hi Masashi, Thanks for letting us know; Unfortunately the FirefoxOS development teams don't have a need to support the pandaboards as they don't help us further the FirefoxOS initiative. The fact that they don't have radio makes them have no advantage over emulator tests. I'm sure there will be community members ensuring that b2g still works on the pandboards, however, we will not be spending thousands of dollars a day to create panda builds. We're currently building other platforms [1] that meet the automation testing and actual mobile devices requirements to further the FirefoxOS. I hope you understand this decision as we support the Mozilla mission. [1] https://tbpl.mozilla.org/?jobname=b2g_&showall=1

> The fact that they don't have radio Which radio do you mean ? FM radio or Wi-Fi ? Wi-Fi works on the Pandaboard. > We're currently building other platforms [1] I don't know the mean of "Unagi", "Inari", "Leo" and so on. Is there any device which has GPIO interface in the list ? If no, is there any plan to support any device which has GPIO interface ? > I'm sure there will be community members ensuring that b2g still works on the pandboards, > however, we will not be spending thousands of dollars a day to create panda builds. Will the pandaboard be erased from git tree (or config.sh) ? If no, will the pandaboard be out of test target ?

(In reply to Masashi Honma from comment #11) > > The fact that they don't have radio > > Which radio do you mean ? FM radio or Wi-Fi ? > Wi-Fi works on the Pandaboard. > Radio to make calls and send text messages. Cellphones do; panda boards not. > > We're currently building other platforms [1] > > I don't know the mean of "Unagi", "Inari", "Leo" and so on. Those are codenames for mobile devices that are similar or the same as what our partners have released or will be releasing. > Is there any device which has GPIO interface in the list ? > If no, is there any plan to support any device which has GPIO interface ? > I have no idea. Maybe someone from dev.b2g would know. > > I'm sure there will be community members ensuring that b2g still works on the pandboards, > > however, we will not be spending thousands of dollars a day to create panda builds. > > Will the pandaboard be erased from git tree (or config.sh) ? > If no, will the pandaboard be out of test target ? I don't think that anyone will take the trouble to go around and strip the panda code from our codebase. This bug is not about that. This bug is about stop creating the builds on tbpl on our Amazon EC2 instances.

> I don't think that anyone will take the trouble to go around and strip the panda code > from our codebase. This bug is not about that. This bug is about stop creating the builds > on tbpl on our Amazon EC2 instances. What is tbpl ? I am not familiar with Mozilla internal system because I am in outside of Mozilla. I am only a user of Firefox OS on Pandaboard. I want to know what will happen after you stop to support of pandaboard. 1) "./config.sh pandaboard" works ? 2) will Source codes and binaries for pandaboard be removed from the git tree ?

(In reply to Masashi Honma from comment #13) > > I don't think that anyone will take the trouble to go around and strip the panda code > > from our codebase. This bug is not about that. This bug is about stop creating the builds > > on tbpl on our Amazon EC2 instances. > > What is tbpl ? I am not familiar with Mozilla internal system because I am > in outside of Mozilla. I am only a user of Firefox OS on Pandaboard. > I want to know what will happen after you stop to support of pandaboard. > tbpl refers to https://tbpl.mozilla.org It allows developers to determine that their changes are passing or failing. > 1) "./config.sh pandaboard" works ? > 2) will Source codes and binaries for pandaboard be removed from the git > tree ? I assume so (since we're not changing any of that in here) but I would ask in the dev.b2g mailing list to be sure. In this bug we're only modifying what runs on tbpl.

Masashi-san: I do understand that the panda boards might be useful to you. Note that this bug isn't about forbidding or making it impossible for you to keep doing so. This is just about whether we are going to have panda boards in our testing infrastructure. We don't have enough engineering resources to keep running tests for all types of hardware that anyone wants to run Firefox OS on. So we have to pick the platforms that are most helpful to the development that we are doing. In general, our testing for B2G is in an *incredibly* sad state. So I think any place that we can simplify without sacrificing too much functionality we have to take that opportunity.

I understand Pandaboard will be removed from automated test environment. Can I continue to use "./config.sh pandaboard" after that ?

(In reply to Masashi Honma from comment #16) > I understand Pandaboard will be removed from automated test environment. Thanks for your understanding. We'll now proceed with this. > Can I continue to use "./config.sh pandaboard" after that ? Yes. This bug is only about stopping our automated builds (currently being generated) and tests (already stopped). We are not removing the "config.sh pandaboard" build option.

The tests are already disabled. Tested on my dev-master.

Comment on attachment 786365 [details] [diff] [review] patch to disable b2g builds for pandas Review of attachment 786365 [details] [diff] [review]: ----------------------------------------------------------------- Just the nits. LGTM. ::: mozilla/BuildSlaves.py.template @@ +10,5 @@ > 'android': 'pass', > 'android-armv6': 'pass', > 'android-noion': 'pass', > 'android-x86': 'pass', > + 'mock': 'pass', Nit: there's one extra white space in that line. ::: mozilla/b2g_config.py @@ +686,5 @@ > '--summary', > '--gecko-languages-file', 'build/b2g/locales/all-locales', > ], > 'gecko_languages_file': 'build/b2g/locales/all-locales', > + }, Nit: 3 extra white spaces. @@ +929,5 @@ > 'win32_gecko': {}, > 'linux32_gecko_localizer': {}, > 'linux64_gecko_localizer': {}, > 'macosx64_gecko_localizer': {}, > + 'win32_gecko_localizer': {}, Nit: white spaces.

In production

verified in tbpl