Hi Am 12.01.2015 um 18:44 schrieb Erik Vold: > Hey Thomas, > > The cause is somehow related to how the emulator tests end up using > the sdk, so the issue may be with the test suite and not the emulator > itself. I would like to know how it is that addon-sdk code changes > are affecting the emulator. If the sdk is being used during the > emulator startup for tests then I would like to know why and how that is.
The emulator uses the same code for Gecko and Gaia as a regular hardware device. If your SDK works on real hardware, I'd expect it wo work on the emulator as well. If tests fail, it's either related to (1) broken tests, (2) broken emulator configuration (kernel, etc), (3) timeouts, due to slowness. Option (2) is unlikely. Either your SDK breaks the test infrastructure (1) or you run into something like bug 1108271 (3). For (1) you could ask on #ateam. The A team maintains the test frameworks. They might be able to help you. > > Do you know who setup these emulator test suites on treeherder? I don't know, but I'd also ask in #ateam. > > It is sad news that it will be so difficult to try changes. I'm not > sure where to make that change mentioned in bug 1108271 comment 6 either.. I think was in gaia/build/config/common-settings.json: 'screen.timeout' contains the timeout in seconds. Best regards Thomas > > Erik > > On Mon, Jan 12, 2015 at 9:24 AM, Thomas Zimmermann > <[email protected] <mailto:[email protected]>> wrote: > > Hi Erik, > > I'd say that the cause is not directly in the emulator itself. The > emulator is just another device but extremely slow, because it's > all in > software. > > The workaround from bug 1108271, comment 6, could help to figure > out if > the screen timeout causes this bug. However, I don't know how to push > Gaia and Gecko changes to a try server at the same time. > > Best regards > Thomas > > > Am 12.01.2015 um 17:43 schrieb Erik Vold: > > Hey Thomas, > > > > It looks like the tests are failing with a "timeout waiting for > port" > > message, which looks related to bug 905324 [1], but that sounds > > related to the bug that you are mentioning they way that you've > > described it. > > > > To answer your questions: > > > > 1. What is the sdk doing exactly? > > > > The sdk doesn't do anything by design unless an add-on uses it, in > > which case it provides a CommonJS loader instance to the add-on, and > > some modules which add-on devs can opt-in to use. > > > > I don't know all of the ways that other people use the loader, and I > > was suspecting the the emulator is using an sdk loader (we have > two, a > > legacy one, and a new one) that we did not expect to support. For > > example, devtools uses the sdk by creating an instance of our new > > loader (toolkit/loader) and using that to enable them to write > > CommonJS modules for their code, and so that they can use some > of the > > sdk modules that were designed for add-ons. > > > > Perhaps devtools code is running at startup? > > > > So the short answer is that the sdk is designed to do nothing except > > when an add-on needs it, but in practice it used it a lot of places > > and nobody has been able to keep track of that. > > > > > > 2. Could it affect the time that the system takes for booting? > > > > It's possible, but I don't see anything in the updated range that > > would affect startup time: > > > > > > https://github.com/mozilla/addon-sdk/compare/1aee56d5ef161f47beb5ee6197cc4598956e6010...4d7d33e3f594d10d0d13c3b97781e20316139b20#diff-92 > > > > > > 3. Can you reproduce the issue locally? > > > > Like I said before, I don't know where this emulator code is, and I > > don't know how to run the tests, so no I haven't been able to > > reproduce the issue locally. > > > > > > Erik > > > > [1] https://bugzilla.mozilla.org/show_bug.cgi?id=905324 > > > > On Mon, Jan 12, 2015 at 8:23 AM, Thomas Zimmermann > > <[email protected] <mailto:[email protected]> > <mailto:[email protected] <mailto:[email protected]>>> > wrote: > > > > Hi Erik, > > > > We have a race condition between the Gaia startup code and > the screen > > timeout. If startup takes too long, the screen will switch > off and the > > system suspends (including its networking). The host system > will then > > see a timeout at some point. See [1] for more information. Maybe > > you see > > the same issue. Could you answer some of the questions below. > > > > - What exactly is the SDK doing? > > - Could it affect the time that the system takes for booting? > > - Can you reproduce the issue locally? > > > > Best regards > > Thomas > > > > [1] https://bugzilla.mozilla.org/show_bug.cgi?id=1108271 > > > > > > Am 12.01.2015 um 16:51 schrieb Erik Vold: > > > Hey there, > > > > > > I could really use some help.. > > > > > > When I did the uplift last week, in bug 1114752, all of > the emulator > > > opt tests failed for some reason, which I need help > understanding. > > > > > > The uplift only changes code in the addon-sdk directory, I > haven't > > > changed any of the emulator code, and the sdk should have > tests for > > > all of the use cases that it is supposed to support.. > > > > > > I've made bug 1120427 for this issue, which I will need > help with, > > > since I don't know anything about the emulator project, like > > where the > > > code is, how to run tests, why and how it uses the sdk, > and I can't > > > see anything useful from the failing test logs so far.. > > > > > > https://bugzilla.mozilla.org/show_bug.cgi?id=1120427 > > > > > > These are the test results: > > > > > > > > > > https://treeherder.mozilla.org/ui/#/jobs?repo=fx-team&revision=2815bfd2ccb5 > > > > > > We would like to get the uplift (bug 1114752) done today, > so I would > > > really appreciate some help here.. we'll probably need an > additional > > > sdk test written too. > > > > > > Erik > > > > > > _______________________________________________ dev-b2g mailing list [email protected] https://lists.mozilla.org/listinfo/dev-b2g
