Closed
Bug 1194979
Opened 9 years ago
Closed 9 years ago
Use Mulet in Gaia
Categories
(Release Engineering :: General, defect)
Release Engineering
General
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: gerard-majax, Assigned: gerard-majax)
References
Details
Attachments
(1 file)
No description provided.
Assignee | ||
Updated•9 years ago
|
Summary: Update mozilla-get-url to be able to get a Mulet → Update mozilla-download to be able to get a Mulet
Assignee | ||
Comment 1•9 years ago
|
||
So mozilla-download seems to be ready. Gaia needs to make use of it.
Summary: Update mozilla-download to be able to get a Mulet → Use Mulet in Gaia
Comment 2•9 years ago
|
||
Assignee | ||
Comment 3•9 years ago
|
||
https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=2475331609d4e9555d604310fbbc7feb97f6f1a4 mozilla-runner is not finding the b2g binary. How can we teach him to use Mulet?
Flags: needinfo?(gaye)
Flags: needinfo?(aus)
Assignee | ||
Comment 4•9 years ago
|
||
After some hacks, it's progressing ... https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=16f73a6dfdeab42e70c7c53dca51089f34455e54
Assignee | ||
Comment 5•9 years ago
|
||
Right, that's better: https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=29b7063f7e3c9ab374429a38ef1320d6a3b5fe23 The JSM failures seems to come from mozilla-runner which is still trying to run b2g.
Assignee | ||
Comment 6•9 years ago
|
||
Greg, I know you already have some Gij to investigate, but you might be interested in the faiures on GU25 there: https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=29b7063f7e3c9ab374429a38ef1320d6a3b5fe23 It seems to be pretty permanent
Flags: needinfo?(gweng)
Comment 7•9 years ago
|
||
See it. But I'm dealing with Gij that I finally have some clues, so I will switch to this when it get solved or stuck again (hope not).
Assignee | ||
Comment 8•9 years ago
|
||
Mostly green, fixed JSM perma fail: https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=c1aaa0fc7cac159024aab6027a63623298fd0428
Flags: needinfo?(gaye)
Flags: needinfo?(aus)
Assignee | ||
Comment 9•9 years ago
|
||
Comment on attachment 8648369 [details] [review] [gaia] lissyx:bug1194979 > mozilla-b2g:master Is it the proper way to switch ? Specifically for the npm modules.
Attachment #8648369 -
Flags: feedback?(gaye)
Attachment #8648369 -
Flags: feedback?(aus)
Comment 10•9 years ago
|
||
Just an update: since I still have no luck at debugging intermittent Gij, I now try to solve the unit test failure first. Debugging.
Comment 11•9 years ago
|
||
Comment on attachment 8648369 [details] [review] [gaia] lissyx:bug1194979 > mozilla-b2g:master The node module related changes look good to me. The rest of the patch also seems alright although I haven't tried it myself.
Attachment #8648369 -
Flags: feedback?(aus) → feedback+
Assignee | ||
Comment 12•9 years ago
|
||
https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=8815380d4d3b718c2e52a792077214cb2e3d0857 Calendar and Keyboard unit tests failures.
Flags: needinfo?(timdream)
Flags: needinfo?(dylan)
Assignee | ||
Comment 13•9 years ago
|
||
(In reply to Alexandre LISSY :gerard-majax from comment #12) > https://treeherder.mozilla.org/#/ > jobs?repo=gaia&revision=8815380d4d3b718c2e52a792077214cb2e3d0857 > > Calendar and Keyboard unit tests failures. False alarm, a lot of retriggers later they are a bit intermittent but that's all.
Flags: needinfo?(timdream)
Flags: needinfo?(dylan)
Assignee | ||
Comment 14•9 years ago
|
||
So I am reproducing your GU25 100% locally: - get a mulet runtime, mine is built from source in ../gecko/obj-mulet/ - run: RUNTIME=../gecko/obj-mulet/dist/firefox/firefox ./bin/gaia-test - open ./apps/system/test/unit/lockscreen/lockscreen_notifications_test.js and save it to trigger tests That is 100% bingo repro of: 1) [system-test/unit/lockscreen/lockscreen_notifications_test.js] system/LockScreenNotifications Remove top mask when actionable notification is at the visual top ofthe container viewport top-actionable class is correctly added in onNotificationHighlighted test on 0-th notification: Error: with notification index 0 container should not have the top-actionable class: expected true to equal false at chaiAssert (app://system.gaiamobile.org/common/test/helper.js:32:1) at equal (app://system.gaiamobile.org/common/vendor/chai/chai.js:2438:1) at testTopActionableClass (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:543:1) at (anonymous) (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:560:1) at wrapper (app://system.gaiamobile.org/common/test/mocha_generators.js:62:19) at run (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3709:7) at runTest (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4081:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4127:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4007:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4016:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3964:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3984:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4932:28) 2) [system-test/unit/lockscreen/lockscreen_notifications_test.js] system/LockScreenNotifications Remove top mask when actionable notification is at the visual top ofthe container viewport top-actionable class is correctly added in onNotificationHighlighted test on 1-th notification: Error: with notification index 1 container should have the top-actionable class: expected false to equal true at chaiAssert (app://system.gaiamobile.org/common/test/helper.js:32:1) at equal (app://system.gaiamobile.org/common/vendor/chai/chai.js:2438:1) at testTopActionableClass (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:543:1) at (anonymous) (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:560:1) at wrapper (app://system.gaiamobile.org/common/test/mocha_generators.js:62:19) at run (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3709:7) at runTest (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4081:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4127:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4007:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4016:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3964:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3984:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4932:28) 3) [system-test/unit/lockscreen/lockscreen_notifications_test.js] system/LockScreenNotifications Remove top mask when actionable notification is at the visual top ofthe container viewport top-actionable class is correctly removed with tryAddTopMaskByNotification test on 0-th notification: Error: with notification index 0 container should have the top-actionable class: expected false to equal true at chaiAssert (app://system.gaiamobile.org/common/test/helper.js:32:1) at equal (app://system.gaiamobile.org/common/vendor/chai/chai.js:2438:1) at testTopActionableClass (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:664:1) at (anonymous) (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:681:1) at wrapper (app://system.gaiamobile.org/common/test/mocha_generators.js:62:19) at run (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3709:7) at runTest (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4081:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4127:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4007:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4016:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3964:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3979:7) at done (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3700:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3712:9) at (anonymous) (app://system.gaiamobile.org/common/test/mocha_generators.js:46:20) at wrapper (app://system.gaiamobile.org/common/test/mocha_generators.js:73:15) at run (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3709:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3973:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3984:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4932:28) 4) [system-test/unit/lockscreen/lockscreen_notifications_test.js] system/LockScreenNotifications Remove top mask when actionable notification is at the visual top ofthe container viewport top-actionable class is correctly removed with tryAddTopMaskByNotification test on 1-th notification: Error: with notification index 1 container should not have the top-actionable class: expected true to equal false at chaiAssert (app://system.gaiamobile.org/common/test/helper.js:32:1) at equal (app://system.gaiamobile.org/common/vendor/chai/chai.js:2438:1) at testTopActionableClass (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:664:1) at (anonymous) (app://system.gaiamobile.org/test/unit/lockscreen/lockscreen_notifications_test.js:681:1) at wrapper (app://system.gaiamobile.org/common/test/mocha_generators.js:62:19) at run (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3709:7) at runTest (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4081:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4127:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4007:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4016:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3964:1) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3979:7) at done (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3700:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3712:9) at (anonymous) (app://system.gaiamobile.org/common/test/mocha_generators.js:46:20) at wrapper (app://system.gaiamobile.org/common/test/mocha_generators.js:73:15) at run (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3709:7) at next (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3973:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:3984:5) at (anonymous) (app://system.gaiamobile.org/common/vendor/mocha/mocha.js:4932:28)
Flags: needinfo?(gweng)
Assignee | ||
Comment 15•9 years ago
|
||
gasp, sorry for this the env variable should be FIREFOX= and not RUNTIME= ...
Assignee | ||
Comment 16•9 years ago
|
||
And local repro is stil here if I use a mulet downloaded from TaskCluster ...
Comment 17•9 years ago
|
||
Update and bad news (to me): I installed another fresh Ubuntu 14.04 VM, built Gaia and ran the test exactly followed what you described, but it just worked well. No error occurred. The steps: 1. Install Ubuntu 14.04 VM on my Mac in Parallels Desktop (don't tell me this software makes the difference...) 2. Install git, python-pip from apt-get 3. Install nvm[1] for the case I may need to switch node versions since Gaia always prints the log "please 0.10 or you may encounter problems" during building and running tests, but I have an opposite experience: 0.10 never worked on my Ubuntu 14.04, and only 0.12 can work well 4. Run 'make' once to make sure everything is fine - Then I run the test- 5. Download Mulet from the updated link on MDN[2]: I fetched the one just ran about 7 hours ago, so it's relatively "fresh" one, not like the 3 months old one I fetched once 6. Unpack Mulet and put it under /home/ubuntu/mulet 7. In Gaia directory, execute the command: FIREFOX=/home/ubuntu/mulet/firefox ./bin/gaia-test 8. After test-agent shows up, open the ./apps/system/test/unit/lockscreen/lockscreen_notifications_test.js and trigger the test 9. It shows 29 passed...even for those failed tests as your log shows. The next step? I submitted a loan request to test it on the remote server, which now has been granted. So I will connect to that and repeat the steps to see if I can find any different results. But before that I will to check code again to see if it could pass all incorrectly (false negative). If so I will try to fix the code and push to Mulet via Gecko try configs. And I also suspect if nvm/node/npm caused the issue. But I have no other options since 0.10 just refused to work on my machine, which showed the dependencies were broken while I try to make it, which will invoke `npm install`. However, the most weird thing is if I install them manually one-by-one, they're all good to install. But since I need to install at least 20 packages manually...it's just too painful and sometimes the dependencies of dependencies still failed, so I gave up to think it's "good", and switched to 0.12 solved all the problems. --- [1]: https://github.com/creationix/nvm [2]: https://tools.taskcluster.net/index/artifacts/#gecko.v1.mozilla-central.latest.linux.mulet/gecko.v1.mozilla-central.latest.linux.mulet.opt
Comment 18•9 years ago
|
||
The loan instance now unfortunately has some connection issue. So while releng people try to solve that I'm not digging into the code, and will pick some suspicious part to do some modifications.
Comment 19•9 years ago
|
||
Okay I did some change in the test file and pushed to the try. Wait the result while I'm trying it on loan instance again.
Comment 20•9 years ago
|
||
Update: Now I finally can reproduce it on the remote instance. I'm debugging it.
Comment 21•9 years ago
|
||
And it looks like: 1. With MacOSX + node 0.12 and latest Mulet[1] it won't reproduce 2. With Ubuntu 15.04(latest), 14.04 (LTS) + node 0.12 and latest Mulet it won't reproduce 3. With Ubuntu 12.04(which TaskCluster uses) + node 0.12 and latest Mulet it CAN reproduce I don't know if TaskCluster instance make other differences. If not, the only difference is the version of Ubuntu, which means Mulet is possibly relying on some system features which may affect the test results. However, since I'm not expert of TaskCluster and Mulet, I can only raise what I doubted here, and fix the test first. [1]: https://tools.taskcluster.net/task-inspector/#If5ja8xATC2-XaB_-2GTYw/
Comment 22•9 years ago
|
||
Update: on my local console, with Ubuntu 12.04 + node 0.12 and the latest Mulet, it won't reproduce... It looks like a TaskCluster only bug. Anyway, I will still try solve it.
Comment 23•9 years ago
|
||
Update: I've found the main failure point is on B2G desktop, in LockScreenNotifications#onNotificationHighlighted The `_getNumNotificationInContainerViewport()` returns '3', but on Mulet it returns 4. This is a direct failure cause, so I will try to dig in that.
Comment 24•9 years ago
|
||
I think I can debug it now and I need to send a patch. So I now fire another bug to track and patch it.
Assignee | ||
Comment 25•9 years ago
|
||
(In reply to Greg Weng [:snowmantw][:gweng][:λ] from comment #21) > And it looks like: > > 1. With MacOSX + node 0.12 and latest Mulet[1] it won't reproduce > 2. With Ubuntu 15.04(latest), 14.04 (LTS) + node 0.12 and latest Mulet it > won't reproduce > 3. With Ubuntu 12.04(which TaskCluster uses) + node 0.12 and latest Mulet it > CAN reproduce > > I don't know if TaskCluster instance make other differences. If not, the > only difference is the version of Ubuntu, which means Mulet is possibly > relying on some system features which may affect the test results. However, > since I'm not expert of TaskCluster and Mulet, I can only raise what I > doubted here, and fix the test first. Or it's because of some race and some environment exposes it more easily ... > > [1]: https://tools.taskcluster.net/task-inspector/#If5ja8xATC2-XaB_-2GTYw/
Comment 26•9 years ago
|
||
Still don't know the cause of different behavior, but I've fixed the root cause. Maybe Alexandre can verify that? I've done it manually on my loaner machine, but more eyes on that makes sure the result is really verified.
Assignee | ||
Comment 27•9 years ago
|
||
(In reply to Greg Weng [:snowmantw][:gweng][:λ] from comment #26) > Still don't know the cause of different behavior, but I've fixed the root > cause. Maybe Alexandre can verify that? I've done it manually on my loaner > machine, but more eyes on that makes sure the result is really verified. Thanks! I have rebased my PR on top of master: https://treeherder.mozilla.org/#/jobs?repo=gaia&revision=94a3e639c7139ef0f3336cbeceaa5410deac291c
Assignee | ||
Comment 28•9 years ago
|
||
(In reply to Alexandre LISSY :gerard-majax from comment #27) > (In reply to Greg Weng [:snowmantw][:gweng][:λ] from comment #26) > > Still don't know the cause of different behavior, but I've fixed the root > > cause. Maybe Alexandre can verify that? I've done it manually on my loaner > > machine, but more eyes on that makes sure the result is really verified. > > Thanks! I have rebased my PR on top of master: > https://treeherder.mozilla.org/#/ > jobs?repo=gaia&revision=94a3e639c7139ef0f3336cbeceaa5410deac291c And we have GU25 green \o/
Assignee | ||
Comment 29•9 years ago
|
||
(In reply to Alexandre LISSY :gerard-majax from comment #27) > (In reply to Greg Weng [:snowmantw][:gweng][:λ] from comment #26) > > Still don't know the cause of different behavior, but I've fixed the root > > cause. Maybe Alexandre can verify that? I've done it manually on my loaner > > machine, but more eyes on that makes sure the result is really verified. > > Thanks! I have rebased my PR on top of master: > https://treeherder.mozilla.org/#/ > jobs?repo=gaia&revision=94a3e639c7139ef0f3336cbeceaa5410deac291c Looks like we have two new failures on GU9 in two FTU unit tests :(
Flags: needinfo?(sfoster)
Comment 30•9 years ago
|
||
I cant reproduce these test failures - in nightly or using mulet. The FTU tutorial appears to play normally on device with the latest gaia+gecko. I do however see a new error in the console when running the FTU on nightly and mulet: HTTP "Content-Type" of "video/mp4" is not supported. Load of media resource app://ftu.gaiamobile.org/style/images/tutorial/VerticalScroll.mp4 failed. index.html Failed to load tutorial media: /style/images/tutorial/VerticalScroll.mp4 tutorial.js:32:11 We try not to block the tutorial - or the unit tests - on the successful loading of the videos to avoid effectively bricking the phone with a bad path or asset. However, that is concerning, and might be implicated in the test failures.
Flags: needinfo?(sfoster)
Comment 31•9 years ago
|
||
If I use mulet to browse to the assets in gaia/apps/ftu/style/images/tutorial/ I get the issue - the mp4s are downloaded rather than played.
Comment 32•9 years ago
|
||
I filed bug 1199348 for the above.
Assignee | ||
Comment 33•9 years ago
|
||
(In reply to Sam Foster [:sfoster] from comment #31) > If I use mulet to browse to the assets in > gaia/apps/ftu/style/images/tutorial/ I get the issue - the mp4s are > downloaded rather than played. I'm not sure this is relevant: I do have this behavior when I try this locally too. But a |make APP=ftu test-agent-test| does pass with Mulet on my system. Do we know if video playback is working well on the gaia-try infra ?
Comment 35•9 years ago
|
||
Fixed by bug 1199429.
Comment 36•9 years ago
|
||
I pushed a follow-up to add the firefox directory to the gitignore: https://github.com/mozilla-b2g/gaia/commit/ef5a291e44fea82f0e3793ac4f17eaa2188af7e0
Assignee | ||
Updated•9 years ago
|
Assignee: nobody → lissyx+mozillians
Assignee | ||
Updated•8 years ago
|
Attachment #8648369 -
Flags: feedback?(gaye)
Updated•7 years ago
|
Component: Tools → General
You need to log in
before you can comment on or make changes to this bug.
Description
•