Closed Bug 1124330 Opened 10 years ago Closed 8 years ago

Unsafe usage of CPOW and possible race condition in browser/devtools/netmonitor tests

Categories

(DevTools :: Netmonitor, defect)

x86
Linux
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED DUPLICATE of bug 1297362

People

(Reporter: ahal, Unassigned)

References

Details

Bug 794984 got backed out for causing devtools failures in the browser/devtools/netmonitor tests. Bug 794984 is a mozprocess change that uses dedicated threads for reading lines off of stdout and stderr, so that there is no chance that the subprocess (in this case Gecko) will ever fill up it's output buffer and block when writing to stdout. In other words, it can potentially speed up the subprocess. With that change applied, netmonitor test failures are highly reproducible (though still intermittent): https://treeherder.mozilla.org/#/jobs?repo=try&revision=f3813e0279a9 I did a bit of investigation (from bug 794984 comment 32): > Here's a new push that turns on debugger logging and calls > SimpleTest.requestCompleteLog(): > https://treeherder.mozilla.org/#/jobs?repo=try&revision=078067dacd1c > > Of note are lots of "unsafe usage of CPOW" warnings and eventually this: > "JavaScript error: , line 0: Error: operation not possible on dead CPOW" > > Also weird is the fact that there are tons of TEST-PASS logs that contain > messages like: > "The tooltip status is incorrect." > > I'll file a devtools bug, maybe someone there can help us understand what is > happening. I'm not very familiar with devtools or CPOW, any help/tips for debugging this would be much appreciated!
(In reply to Andrew Halberstadt [:ahal] from comment #0) > > Of note are lots of "unsafe usage of CPOW" warnings and eventually this: > > "JavaScript error: , line 0: Error: operation not possible on dead CPOW" I haven't looked into it, but this may well be pointing to the actual problem. > > Also weird is the fact that there are tons of TEST-PASS logs that contain > > messages like: > > "The tooltip status is incorrect." I think this is just Victor's preferred style for messages, presumably based on the assumption that people would only look at these messages when they fail, in which case the text would be accurate.
related to bug 1119884?
If that message is happening in other bc chunks, then it is a possible red herring. Bug 794984 specifically breaks the devtools netmonitor tests, all other browser chrome tests pass as normal.
Blocks: 794984
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → DUPLICATE
Product: Firefox → DevTools
You need to log in before you can comment on or make changes to this bug.