908213 - "System JS : ERROR chrome://mochitests/content/chrome/content/base/test/chrome/cpows_parent.xul:25 TypeError: data is undefined" in all Linux/Windows mochitest-other runs

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect)

Description

[Ryan VanderMeulen [:RyanVM]]

Due to recent improvements to the TBPL log parser, we are catching more exceptions in the test logs than we did previously. While they are currently non-fatal, we intend to change that eventually. CCing the various people that appear to have been involved with working on the CPOW tests.

Affects both opt and debug builds. Interestingly, this doesn't appear to hit OSX.

https://tbpl.mozilla.org/php/getParsedLog.php?id=26867837&tree=Mozilla-Central

WINNT 6.2 mozilla-central opt test mochitest-other on 2013-08-22 05:23:08 PDT for push 89294cd501d9
slave: t-w864-ix-079

05:26:33     INFO -  891 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_cpows.xul | proxy operation threw exception
05:26:33     INFO -  NOTE: child process received `Goodbye', closing down
05:26:33     INFO -  [Child 3580] WARNING: pipe error: 109: file e:/builds/moz2_slave/m-cen-w32-00000000000000000000/build/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 312
05:26:33     INFO -  892 INFO TEST-END | chrome://mochitests/content/chrome/content/base/test/chrome/test_cpows.xul | finished in 236ms
05:26:33     INFO -  893 INFO TEST-START | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html
05:26:33     INFO -  894 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | appStatus should be present in nsIPrincipal, if not the rest of this test will fail
05:26:33     INFO -  added observers
05:26:33     INFO -  ############################### browserElementPanning.js loaded
05:26:33     INFO -  ######################## BrowserElementChildPreload.js loaded
05:26:33     INFO -  895 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | same_origin : style should be allowed
05:26:33     INFO -  896 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | cross_origin : style should be allowed
05:26:33     INFO -  897 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | same_origin : script should be allowed
05:26:33     INFO -  898 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | cross_origin : script should be allowed
05:26:33     INFO -  899 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | same_origin : img should be allowed
05:26:33     INFO -  900 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | cross_origin : img should be allowed
05:26:33     INFO -  System JS : ERROR chrome://mochitests/content/chrome/content/base/test/chrome/cpows_parent.xul:25
05:26:33     INFO -                       TypeError: data is undefined
05:26:33     INFO -  System JS : ERROR chrome://mochitests/content/chrome/content/base/test/chrome/cpows_parent.xul:25
05:26:33     INFO -                       TypeError: data is undefined
05:26:33     INFO -  JavaScript error: chrome://browser/content/tabbrowser.xml, line 2303: aTab is null
05:26:33     INFO -  901 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_csp_bug768029.html | iframe principal's app status doesn't match the expected app status.


https://tbpl.mozilla.org/php/getParsedLog.php?id=26868729&tree=Mozilla-Central

Ubuntu VM 12.04 mozilla-central opt test mochitest-other on 2013-08-22 05:38:20 PDT for push 89294cd501d9
slave: tst-linux32-ec2-398

05:41:14     INFO -  998 INFO TEST-START | chrome://mochitests/content/chrome/content/base/test/chrome/test_cpows.xul
05:41:14     INFO -  Xlib:  extension "RANDR" missing on display ":0".
05:41:14     INFO -  creating 1!
05:41:14     INFO -  [TabChild] SHOW (w,h)= (10, 10)
05:41:15     INFO -  loading about:blank, 1
05:41:15     INFO -  loaded child cpow test
05:41:15     INFO -  beginning cpow sync test
05:41:15     INFO -  999 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_cpows.xul | correct json
05:41:15     INFO -  System JS : ERROR chrome://mochitests/content/chrome/content/base/test/chrome/cpows_parent.xul:25
05:41:15     INFO -                       TypeError: data is undefined
05:41:15     INFO -  beginning cpow async test
05:41:15     INFO -  1000 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_cpows.xul | correct json
05:41:15     INFO -  System JS : ERROR chrome://mochitests/content/chrome/content/base/test/chrome/cpows_parent.xul:25
05:41:15     INFO -                       TypeError: data is undefined
05:41:15     INFO -  loaded child cpow test
05:41:15     INFO -  beginning cpow sync test
05:41:15     INFO -  1001 INFO TEST-PASS | chrome://mochitests/content/chrome/content/base/test/chrome/test_cpows.xul | correct json

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Yep, started on evilpie's push.
https://tbpl.mozilla.org/?tree=Mozilla-Inbound&rev=bf471c1e49ba

Comment 2

[Tom S [:evilpie]]

Okay, I think I know what is going on, will fix this ASAP.

Comment 3

[Tom S [:evilpie]]

Attachment: enable browser.tabs.remote

Comment 4

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

I don't think this is a good idea. Setting browser.tabs.remote has all sorts of consequences throughout the browser. I think it would be better to have a separate pref to force CPOWs to be enabled.

Comment 5

[Tom S [:evilpie]]

Well in bug 902013 I added a pref to "force-disable" CPOWs, but I guess we could use the reverse as well. Actually what we might be able to do is create the browser element on the fly and just set browser.tabs.remote during its creating, that should have minimal effects on anything else.

Comment 6

[Tom S [:evilpie]]

Attachment: better fix

- Changed dom.ipc.cpows.force-disabled to dom.ipc.cpows.force-enabled and use it to enable CPOWs
- Added onerror handler so that an error actually makes the test fail
- Changed the last "done" message from the child to be synchronous to prevent (real) timing issues.

Comment 7

[David Anderson [:dvander] - inactive, e-mail if emergency]

Comment on attachment 794176 [details] [diff] [review]
better fix

Review of attachment 794176 [details] [diff] [review]:
-----------------------------------------------------------------

Nice, thanks for fixing these tests up.

Comment 8

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/89f609571f44