1176663 - [e10s] Marionette.navigate() hangs when first tab gets closed with: NS_ERROR_ILLEGAL_VALUE in nsIMessageSender.sendAsyncMessage

Status: RESOLVED FIXED

(Remote Protocol :: Marionette, defect)

Description

[Henrik Skupin [:whimboo][⌚️UTC+1]]

marionette.navigate() will hang in getting the current URL after the first tab has been closed in e10s mode. It does not happen in non-e10s mode. To reproduce this issue simply run the attached testcase with Marionette and e10s enabled.

Here the Javascript failure:

JavaScript error: chrome://marionette/content/listener.js, line 374: NS_ERROR_ILLEGAL_VALUE: Component returned failure code: 0x80070057 (NS_ERROR_ILLEGAL_VALUE) [nsIMessageSender.sendAsyncMessage]

With Marionette logging turned on you can see it fails when trying to execute "get":

*** Loading URL which will freeze the execution of the test
1434962441252	Marionette	DEBUG	conn4 -> ({"to":"0","sessionId":"{d5939d11-da6f-4471-8373-fe04c4bfcf62}","name":"getContext"})
1434962441253	Marionette	DEBUG	conn4 client <- ({"from":"0","value":"chrome","status":"success","sessionId":"{d5939d11-da6f-4471-8373-fe04c4bfcf62}"})
1434962441254	Marionette	DEBUG	conn4 -> ({"to":"0","sessionId":"{d5939d11-da6f-4471-8373-fe04c4bfcf62}","name":"setContext","parameters":{"value":"content"}})
1434962441254	Marionette	DEBUG	conn4 client <- ({"from":"0","ok":true})
1434962441255	Marionette	DEBUG	conn4 -> ({"to":"0","sessionId":"{d5939d11-da6f-4471-8373-fe04c4bfcf62}","name":"get","parameters":{"url":"http://mozqa.com"}})

After the last line Marionette hangs and does not proceed with the test anymore. A quick test has shown that this is a regression for Firefox 41. I will check for a closer range.

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+1]]

This regressed between 06-11 and 06-12. The pushlog from mozilla-central:

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=9ebd530c5843&tochange=8cf9d3e497f9

There are no e10s related changes in that inbound merge which could have caused such changed behavior. So I strongly feel that the fix from Chris on bug 1169600 regressed the behavior in Marionette itself.

Chris, can you please have a look at this? Thanks.

Comment 2

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Attachment: test_marionette.py

Comment 3

[Chris Manchester (limited bugmail, email directly)]

This is the underlying problem in bug 1169600 (which was not addressed in the original patch), I'm going to post a new fix there.

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Issue got fixed in bug 1169600.