1200334 - Opening a tab in a message handler (e10s: OFF) corrupts tab handling

Status: UNCONFIRMED

(Firefox :: Tabbed Browser, defect)

Description

[Anthony Lieuallen]

Context: https://github.com/greasemonkey/greasemonkey/issues/2234

We've got a function [1] handling a message [2] which opens a tab.  When doing so, with E10S *off*, Firefox logs to the browser console:


NS_ERROR_UNEXPECTED: Component returned failure code: 0x8000ffff (NS_ERROR_UNEXPECTED) [nsISyncMessageSender.sendSyncMessage] tab-content.js:442:0
NS_ERROR_UNEXPECTED: Component returned failure code: 0x8000ffff (NS_ERROR_UNEXPECTED) [nsISyncMessageSender.sendSyncMessage] content-sessionStore.js:216:0


And the tab state is corrupted severely enough that the tab cannot be closed.  This appears to be a race condition related to message sending/handling.  A timeout around the code that adds the tab [3] is sufficient to work around the failure, but should not be necessary.  When E10S is on, this issue is not triggered.

Confirmed in Mozilla/5.0 (X11; Linux x86_64; rv:43.0) Gecko/20100101 Firefox/43.0 .  Not sure what other reporters (in the GitHub issue) are running.


[1] https://github.com/greasemonkey/greasemonkey/blob/a3ec30452dd359d0c2e1b2246988d813bb1fc3a7/content/browser.js#L68
[2] https://github.com/greasemonkey/greasemonkey/blob/a3ec30452dd359d0c2e1b2246988d813bb1fc3a7/content/browser.js#L15
[3] https://github.com/greasemonkey/greasemonkey/pull/2262/files

Comment 1

[Loic]

Any STR with Greasemonkey?

Comment 2

[Anthony Lieuallen]

1. Install http://roger21.free.fr/test/test_gm_openintab_2.user.js
2. Open https://github.com/greasemonkey/greasemonkey/issues/2107
3. Click the bright buttons at the top

Maybe one in particular, for me all three seemed to do it.

Comment 3

[Simon Lydell]

I’m experiencing this as well, while working on the VimFx add-on. I get this error when doing almost _anything_ tab related, not only when opening tabs. Have a look at this patch to see what I mean:

https://github.com/akhodakivskiy/VimFx/commit/f856db46867305ec34fab89e1d6fc956dacb37af

Comment 4

[janekptacijarabaci]

I think it's fixed...

__Last bad: 47.0a1 (2016-02-21)__
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:47.0) Gecko/20100101 Firefox/47.0
https://hg.mozilla.org/mozilla-central/rev/af6356a3e8c56036b74ba097395356d9c6e6c5a3

__First good: 47.0a1 (2016-02-22)__
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:47.0) Gecko/20100101 Firefox/47.0
https://hg.mozilla.org/mozilla-central/rev/e1cf617a1f2813b6cd66f460313a61c223406c9b

Pushlog
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=af6356a3e8c56036b74ba097395356d9c6e6c5a3&tochange=e1cf617a1f2813b6cd66f460313a61c223406c9b

Bug 1245153 is the suspect.

Comment 5

[janekptacijarabaci]

I'm sorry...

Bug 1195295 is the suspect. The first was a typo...

Comment 6

[Simon Lydell]

I removed the workarounds in VimFx and tried it out in Firefox Developer Edition 48.0a2 (2016-05-06) with e10s disabled, and I couldn’t spot any problems. So, I think it’s fixed too. Thanks!