Open Bug 1398766 Opened 2 years ago Updated 10 months ago

Intermittent toolkit/mozapps/extensions/test/browser/browser_gmpProvider.js | Test timed out -


(Toolkit :: Add-ons Manager, defect, P5)





(Reporter: intermittent-bug-filer, Unassigned, NeedInfo)


(Keywords: intermittent-failure, leave-open, Whiteboard: [stockwell disabled])


(1 file)

Chris, this is showing up pretty frequent on Beta57 at the moment. Any chance you could take a look or refer it to someone else?
Flags: needinfo?(cpearce)
James, are you able to take a look at this test?
Flags: needinfo?(cpearce) → needinfo?(jacheng)
Nearly, but not quite, permaorange on Linux starting with the landing of bug 1384330 on autoland. I suppose it's possible that patch is actually at fault, but particularly given Ryan's experience with 57-as-beta, I think it's vastly more likely that instead it's the shuffling of chunks due to that patch adding a test which caused it to start failing, and that it depends on either running after something, or on not having something run before it, to pass.
Sure, I will look into this.
Flags: needinfo?(jacheng)
I have no idea and no clue to find out the root cause of this issue.

I can reproduce it only if I run the "./mach mochitest toolkit/mozapps/extensions/test/browser/browser_gmpProvider.js --run-until-failure" and make the browser into background manually and repeatedly. If I run the test and don't interrupt it, it will never fail(I think "try" runs just like this).

The timeout is caused by that we didn't get the "popupshown" of the contextmenu [1], so we didn't resolve the promise.

But there is no reason that 

EventUtils.synthesizeMouse(el, 4, 4, { }, gManagerWindow);
EventUtils.synthesizeMouse(el, 4, 4, { type: "contextmenu", button: 2 }, gManagerWindow);

takes no effect.

I searched from the full log and there is no other browser tests failure(I suspected that the issue has some dependencies with other tests).

The only thing I can do is to remove the "synthesizeMouse" code since it is not stable on try.

Hi Chris,
Could I remove this two unstable synthesizing code and ignore the event callback(I don't think it is necessary)?
Or you can suggest anyone who better understand "EventUtils.synthesizeMouse" that he can tell me why it might not take effect.


Flags: needinfo?(cpearce)
I'm not sure if that's a good idea or not. spohl, or someone else in the `hg log toolkit/mozapps/extensions/test/browser/browser_gmpProvider.js` might be able to advise you on that.
Flags: needinfo?(cpearce)
Hi spohl,

According to comment 8, may I please have your advice on the question that I've ask cpearce?

If you agree with removing the unstable code, I will do it and let you review the patch.

Thank you.
Flags: needinfo?(spohl.mozilla.bugs)
Whiteboard: [stockwell needswork]
Attachment #8909648 - Flags: review?(spohl.mozilla.bugs)
I think it would be good to remain this test case but removing the unstable part to eliminate the fail rate.

Please review this patch, thanks.
Comment on attachment 8909648 [details]
Bug 1398766 - Remove the unstable mouse event synthesizing to make the test stable.

The code in question was added in bug 1056035. Did something change that made this fail? Like you said, EventUtils.synthesizeMouse shouldn't fail and may indicate an underlying problem. Removing test code simply because it doesn't always pass is a questionable approach. Can you investigate why the popupshown event doesn't fire? You may also want to check with the original author ( or the reviewer (:gfritzsche) of this code.
Flags: needinfo?(spohl.mozilla.bugs)
Attachment #8909648 - Flags: review?(spohl.mozilla.bugs)
Pushed by
Disable toolkit/mozapps/extensions/test/browser/browser_gmpProvider.js on linux debug for frequent failures. r=me, a=test-only
disabled this test to reduce the impact on the trees- please remember to enable this when testing a fix on try and landing!
Keywords: leave-open
Whiteboard: [stockwell disable-recommended] → [stockwell disabled]
Since gfritzsche is on PTO so I cannot ask for his opinions until Oct 3.

I would like to ask why "EventUtils.synthesizeMouse" is unstable now on try first.
Hi smaug,

According to the comment 6(adding a new unrelated browser test) and comment 8(make this test unstable),

I'd like to ask for your help what is the common mistake or root cause that 

makes EventUtils.synthesizeMouse intermittent failure or it successfully synthesized the mouse event but miss the popup event?

We currently disable the whole test but if we can make EventUtils.synthesizeMouse stable, we should enable it on try.

Could you please help us or forward the question to the right person since I don't know who can help on this?

Thank you.
Flags: needinfo?(bugs)
synthesizeMouse is stable.
Is it possible that the layout underneath isn't stable? Do you need to explicitly flush it?
Or something else is happening asynchronously?
Have you tried spinning event loop before or between those synthesizeMouse calls? That might reveal something useful to debug.
Flags: needinfo?(bugs)
The leave-open keyword is there and there is no activity for 6 months.
:ddurst, maybe it's time to close this bug?
Flags: needinfo?(ddurst)
Seems reasonable to me, not sure if there's a follow-up that needs to happen from #c21.
Flags: needinfo?(ddurst) → needinfo?(ayumiqmazaky)
You need to log in before you can comment on or make changes to this bug.