982600 - ASAN log spew: "AddressSanitizer: SEGV /builds/slave/m-in-l64-asan-0000000000000000/build/js/src/ctypes/typedefs.h:78 js::ctypes::ConvertToJS(JSContext*, JS::Handle<JSObject*>, JS::Handle<JSObject*>, void*, bool, bool, JS::Value*)"

Status: RESOLVED FIXED

(Core :: js-ctypes, defect)

Description

[Carsten Book [:Tomcat]]

Ubuntu ASAN VM 12.04 x64 mozilla-inbound opt test mochitest-browser-chrome on 2014-03-12 01:36:22 PDT for push 01ca94b4729f

slave: tst-linux64-spot-641

https://tbpl.mozilla.org/php/getParsedLog.php?id=35989404&tree=Mozilla-Inbound#error0

01:48:27     INFO -  SUMMARY: AddressSanitizer: SEGV /builds/slave/m-in-l64-asan-0000000000000000/build/js/src/ctypes/typedefs.h:78 js::ctypes::ConvertToJS(JSContext*, JS::Handle<JSObject*>, JS::Handle<JSObject*>, void*, bool, bool, JS::Value*)

Comment 1

[Ryan VanderMeulen [:RyanVM]]

While it's good to have gotten this on file, note that this is *NOT* currently an intermittent failure. The actual failure in the linked log was the timeout in browser_968447_bookmarks_toolbar_items_in_panel.js. The ASAN warning is just a bunch of noise that we're seeing now due to the recent TBPL log parser changes.

I'm taking intermittent-failure off this bug because I don't want it being a source of mis-stars for other failures.

Comment 2

[Andrew McCreight [:mccr8]]

If this is a bother, you could alter the TBPL regexp so it doesn't match "SUMMARY: AddressSanitizer: SEGV".

Comment 3

[Ryan VanderMeulen [:RyanVM]]

Is there a reason these shouldn't be fatal errors?

Comment 4

[Andrew McCreight [:mccr8]]

It looks like this test is intentionally crashing a child process:
  http://mxr.mozilla.org/mozilla-central/source/browser/base/content/test/social/browser_social_workercrash.js
So this behavior is probably expected.

Comment 5

[Andrew McCreight [:mccr8]]

In general, yes, we do care about a SEGV, and it could end up causing a real crash, but in general those are going to show up in non-ASAN builds, so it isn't as big of a deal to miss them.  And presumably they'd show up as regular crashes that the test harness can deal with if they happen in the parent process.

Comment 6

[Ed Morley [:emorley]]

https://tbpl.mozilla.org/php/getParsedLog.php?id=45758731&tree=B2g-Inbound#error2

Comment 7

[Ed Morley [:emorley]]

I might add this to the list that TBPL blacklists, given me forgetting this and getting confused again today, plus things like bug 985845 being misfiled as a result.

Comment 8

[Nicholas Nethercote [inactive]]

FWIW, I got confused by this today and filed bug 1057907, and we even briefly closed the tree as a result.

Comment 9

[Ryan VanderMeulen [:RyanVM]]

(In reply to Ed Morley [:edmorley] from comment #7)
> I might add this to the list that TBPL blacklists, given me forgetting this
> and getting confused again today, plus things like bug 985845 being misfiled
> as a result.

I think we need to proceed with this approach, since I don't think we're going to fix this on the harness side. It's an ongoing source of confusion for people.

Comment 10

[Andrew McCreight [:mccr8]]

Decoder, do you know if there's some way to annotate an intentional crash in ASan?

Comment 11

[Christian Holler (:decoder)]

(In reply to Andrew McCreight [:mccr8] from comment #10)
> Decoder, do you know if there's some way to annotate an intentional crash in
> ASan?

There is no way to do this. In general, crashing tests should just be disabled under ASan (there is no crash reporter anyway to do something useful with the crash)

Comment 12

[Andrew McCreight [:mccr8]]

Ah, that's a good point.

It looks like these tests are crashing if you want to disable them for ASan, Ryan:
  toolkit/components/thumbnails/test/browser_thumbnails_bg_crash_during_capture.js
  toolkit/components/thumbnails/test/browser_thumbnails_bg_crash_while_idle.js
  browser/base/content/test/social/browser_social_workercrash.js
  dom/ipc/tests/test_process_error.xul

Comment 13

[Christian Holler (:decoder)]

Pushed patch to try that disables all of these tests if no crashreporter is present:

https://tbpl.mozilla.org/?tree=Try&rev=7533382916e8

Please report back if this solves your problems.

Comment 14

[Ryan VanderMeulen [:RyanVM]]

(In reply to Christian Holler (:decoder) from comment #13)
> Pushed patch to try that disables all of these tests if no crashreporter is
> present:
> 
> https://tbpl.mozilla.org/?tree=Try&rev=7533382916e8
> 
> Please report back if this solves your problems.

No logspam in the run :) f+

Comment 15

[Christian Holler (:decoder)]

Attachment: disable-asan-tests.patch

Comment 16

[Christian Holler (:decoder)]

Requested review from both test owners based on f+ in comment 14. This patch disables the mentioned tests if no crashreporter is present (which is the case for ASan builds).

Comment 17

[Mark Hammond [:markh] [:mhammond]]

Comment on attachment 8498052 [details] [diff] [review]
disable-asan-tests.patch

Review of attachment 8498052 [details] [diff] [review]:
-----------------------------------------------------------------

Note :adw is on PTO, but I think this looks fine with the requested changes.

::: browser/base/content/test/social/browser.ini
@@ +54,5 @@
>  [browser_social_status.js]
>  skip-if = e10s # Bug 915547 (social providers don't install)
>  [browser_social_window.js]
>  [browser_social_workercrash.js]
> +run-if = crashreporter

please make this 'skip-if = !crashreporter' - see https://bugzilla.mozilla.org/show_bug.cgi?id=1074507#c2

::: dom/ipc/tests/chrome.ini
@@ +3,5 @@
>    process_error.xul
>    process_error_contentscript.js
>  
>  [test_process_error.xul]
> +skip-if = buildapp == "mulet" || (crashreporter == false)

just '|| !crashreporter' here (and the other spots) too please.

Comment 18

[Christian Holler (:decoder)]

Landed with the requested changes, thanks Mark!

https://hg.mozilla.org/integration/mozilla-inbound/rev/bef7426ecdb2

Comment 19

[Ryan VanderMeulen [:RyanVM]]

Thanks, Christian!

https://hg.mozilla.org/releases/mozilla-aurora/rev/baaf699255e1
https://hg.mozilla.org/releases/mozilla-beta/rev/449e9b0c1780

Comment 20

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/bef7426ecdb2