Reproducible on latest Nightly (BuildID: 20140817030204) Not reproducible on latest Aurora (BuildID: 20140817004001) Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:34.0) Gecko/20100101 Firefox/34.0 Steps to reproduce: 1. Launch Firefox and wait for OpenH264 plugin to install 2. Go to http://mozilla.github.io/webrtc-landing/pc_test.html 3. Start test with "require h264 video" checkbox marked 4. Crash the plugin using 'media.gmp.plugin.crash';'true' Expected results: "Submit a crash report" button is available in the OpenH264 crash reporting notification. Actual results: "Submit a crash report" button is not available in the OpenH264 crash reporting notification. Notes: 1. On Windows 7 64bit and Ubuntu 14.04 32bit, the button is shown. 2. Logs: https://pastebin.mozilla.org/6036491 3. Initial crash in about:crashes 48ECDFBB-334C-43E3-939C-74E878DDC076, after I click the link it submits https://crash-stats.mozilla.com/report/index/20965cf7-d7ee-4be5-95ef-b14f62140818

Can we get this in the next iteration?

It appears that we are successfully crashing and even have a crash report, but it's not associated with the GMP crash properly. It's a bit strange that there is almost no metadata with the report.

Added to the 34.3 spreadsheet.

Sounds like I should tag this for uplift to Fx 33 when it's fixed?

http://hg.mozilla.org/mozilla-central/annotate/4d94eeca89f3/content/media/gmp/GMPChild.cpp#l133 is early-returning before we hook up the crash reporter.

Comment on attachment 8475372 [details] [diff] [review] bug1054965-mac-gmpcrash This looks fine to me.

Verified as fixed on Mac OS X 10.9.4 using latest Nightly (buildID: 20140821030201) from 2014-08-21.

Comment on attachment 8475372 [details] [diff] [review] bug1054965-mac-gmpcrash Approval Request Comment [Feature/regressing bug #]: bug 1012949 [User impact if declined]: no crash submissions for mac OpenH264 [Describe test coverage new/current, TBPL]: manual QA passed from nightly landing [Risks and why]: Very low-risk: reordered a method so that mac matches existing window/linux codepath. [String/UUID change made/needed]: None gfritzsche/felipe, do either of you know why this wasn't caught by browser_globalplugin_crashinfobar.js ? Do we need to add a check in that test?

We don't have GMP-specific coverage yet as the fake GMP plugin was not available soon enough. I filed bug 1046052 on that coverage.

Bug 1012949 hasn't landed on Aurora yet, so there's nothing to do here.

I added a whiteboard note to that bug, since it's a likely uplift candidate.

Already verified on 34 based on comment 9.

I see that bug 1046052 has been created for a mochitest. I don't see a need for us to create a Mozmill test. Liz, please always check back with the dev about an in-tree test first before requesting a test in an external test framework. Thanks.

I've requested upload to aurora for bug 1012949 and a bunch of other patches listed in bug 1012949 comment #110 (including this one). I'll wait to uplift this one until permission is granted to uplift the other patches.

Comment on attachment 8475372 [details] [diff] [review] bug1054965-mac-gmpcrash Landed on mozilla-aurora: https://hg.mozilla.org/releases/mozilla-aurora/rev/ea8195633864

Also verified as fixed on latest Aurora 33.0a2, the crash submission button appears in the notification.