Closed Bug 904742 Opened 11 years ago Closed 10 years ago

Intermittent "System JS : ERROR chrome://satchel/content/formSubmitListener.js:15 ReferenceError: XPCOMUtils is not defined" when running B2G mochitest-3

Categories

(Toolkit :: Form Manager, defect)

ARM
Gonk (Firefox OS)
defect
Not set
normal

Tracking

()

RESOLVED WORKSFORME
Tracking Status
firefox24 --- unaffected
firefox25 --- unaffected
firefox26 --- affected

People

(Reporter: RyanVM, Unassigned)

References

Details

(Keywords: intermittent-failure)

Due to recent improvements to the TBPL log parser, we are catching JS exceptions that previously went unnoticed. I recently noticed that we hit the following exception intermittently on B2G mochitest-3 runs. https://tbpl.mozilla.org/php/getParsedLog.php?id=26496785&tree=Mozilla-Inbound 10:23:10 INFO - System JS : ERROR chrome://satchel/content/formSubmitListener.js:15 10:23:10 INFO - ReferenceError: XPCOMUtils is not defined The other exceptions shown in that log are already filed and can be disregarded here. Dave, can you suggest someone who might look into this? We are hoping to make these exceptions fatal at some point, at which point this will become an intermittent orange.
Flags: needinfo?(dtownsend+bugmail)
Dolske and I watch this component and know this code.
Flags: needinfo?(dtownsend+bugmail)
Do I need to appease you with offerings of meece to get you to take it?
I can't take this until we're done fighting perf issues with Australis.
OK, fine. Meece *and* maple syrup.
I can't find the Like button for comment 4.
I suspect there's something B2G specific here, or at least something terrifyingly wrong if Cu.import() isn't working. Not sure who would be good to look into this, let's start with Blake since he probably knows a bit of this and has touched B2G. FWIW, I think I saw a similar error when working on bug 901409, if I dropped an actually-still-need enablePrivledge() from a mochitest that that was using XPCOMUtils in exactly this way (for generateQI). But the code here should always be running with chrome privs, and the Ci.import() shouldn't ever fail. So... Bizarre B2G bug? JS bug?
Flags: needinfo?(mrbkap)
This is almost certainly a bug with reuseglobal. I bet we have tests that use Components.utils.import, whose functions get collected meaning that the script gets reused for a different function altogether and we end up importing onto the wrong global object or something (I haven't investigated this too closely). It should be easy to hack something up that get called back when JS scripts get destroyed and removes the mapping in the components loaders mThisObject map.
Flags: needinfo?(mrbkap)
No longer blocks: 892958
Blocks: log-SnR
Blocks: 920191
No longer blocks: log-SnR
This should be fixed by the patch for bug 952183.
Depends on: 952183
Intermittent failure not seen for >3 months; filter on mass-intermittent-wfm-20140812.
Status: NEW → RESOLVED
Closed: 10 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.