Open Bug 1191134 Opened 4 years ago Updated 3 years ago

Spurious load event from test harness in e10s

Categories

(Testing :: Mochitest, defect)

34 Branch
defect
Not set

Tracking

(e10s+)

Tracking Status
e10s + ---

People

(Reporter: billm, Unassigned)

Details

Attachments

(1 file)

Attached patch patchSplinter Review
I noticed this problem while debugging some tests locally. For some reason it doesn't show up on tinderbox. Unfortunately, it causes some orange on tinderbox that I can't reproduce locally. I'm posting it here so that I don't lose the patch, but I'm not sure when I'll be able to get back to it.

The basic problem is this code from the test harness:
    // Replace the last tab with a fresh one
    if (window.gBrowser) {
      gBrowser.addTab("about:blank", { skipAnimation: true });
      gBrowser.removeCurrentTab();
      gBrowser.stop();
    }

The gBrowser.stop() is supposed to prevent the about:blank load from triggering a load event. In non-e10s it does. But in e10s, the stop() function just sends a message to the content process. This message is processed in a separate turn of the event loop from the message to create a new tab. During that time, a load event could have fired and been propagated to the parent. If that happens, then the next test that runs will be given a load event for about:blank that it doesn't need or want.
> Unfortunately, it causes some orange on tinderbox

Should have said "Unfortunately, my fix for the problem causes some orange on tinderbox".
Bill, is this still current?
Component: General → Mochitest
Flags: needinfo?(wmccloskey)
Product: Firefox → Testing
Yes, I think it's still a bug. I haven't reproduced it recently.
Flags: needinfo?(wmccloskey)
Assignee: wmccloskey → nobody
You need to log in before you can comment on or make changes to this bug.