Follow the steps from bug 1146955,

1. Go to https://mozilla.github.io/webrtc-landing/pc_test.html
2. Check Require H.264 and then press the Start button
3. Confirm the popups to allow device usage
4. Intentionally kill the media plugin process

Expected: A notification bar should appear

Actual: the plugin area has crashed and stops drawing but no notification appears.

The test browser_globalplugin_crashinfobar.js tests that the event from the plugin opens the notification bar. This test still works so the frontend notification seems to still work. The PluginCrashed event never occurs.

This bug makes it difficult to remove the framescript toolkit/content/process-content.js as I cannot properly test my changes.

The severity field is not set for this bug.
:jimm, could you have a look please?

For more information, please visit auto_nag documentation.

Tips for bisecting:

• openh264 needs to be downloaded for us to standup the GMP -- you can force this by going to about:addons (plugins tab) -> clicking on gear -> check for updates.
• The GMP can be crashed by killing the plugin-container process or by setting the media.gmp.plugin.crash pref to true.

This looks like it's been busted for awhile. My current regression range is 2018-08-15 to 2019-08-15. Will look at getting a specific regression and seeing what can be done.

Regression window

2020-08-24T14:58:46.712000: INFO : Narrowed nightly regression window from [2018-11-07, 2018-11-09] (2 days) to [2018-11-08, 2018-11-09] (1 days) (~0 steps left)

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=5836a60614764631436bf5030c5baa34c676c7a2&tochange=d3d642b624886729636c3690a806f38b4d737731

Looks like the message is still shown for crashes where Widevine is being hosted in the GMP, but not for OpenH264.

It looks like when an OpenH264 plugin crashes we don't have any helpers so the crash doesn't get reported[0]. However, even when this did work I'm having trouble figuring out where a helper was registered. It looks like the WebRTC encoder has never passed a crash helper when getting a GMP[1], and I'm currently trying to figure out how else the WebRTC case could have been handling crashes.

[0] https://searchfox.org/mozilla-central/rev/19c23d725f27d0989e4a60f36d64004cebb39736/dom/media/gmp/GMPService.cpp#158-162
[1] https://searchfox.org/mozilla-central/rev/19c23d725f27d0989e4a60f36d64004cebb39736/media/webrtc/signaling/src/media-conduit/WebrtcGmpVideoCodec.cpp#200

Passing NI.