An odd interaction between OS.File.writeAtomic and existing code is causing crashes in tests. I've tracked the crashes down to a conflict between unit tests that install a mock version of the @mozilla.org/plugin/host;1 service (nsIPluginHost), and core C++ code that assumes the plugin host service will always be the singleton C++ nsPluginHost. From bug 657952: According to DXR, tests that mock "@mozilla.org/plugin/host;1" are: http://dxr.mozilla.org/mozilla-central/source/toolkit/components/telemetry/tests/unit/test_TelemetryPing.js#l435 http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/test/xpcshell/test_pluginchange.js#l121 http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/test/xpcshell/test_bug455906.js#l182 http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/test/xpcshell/test_bug514327_3.js#l102 http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/test/xpcshell/test_bug449027.js#l430 http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/test/xpcshell/test_duplicateplugins.js#l96 See the comments on Bug 657952 beginning at https://bugzilla.mozilla.org/show_bug.cgi?id=657952#c5 for additional information. I need to disable the crashing tests in my patches for bug 853388; I'll refer to this bug in the xpcshell.ini

(In reply to :Irving Reid from comment #0) > An odd interaction between OS.File.writeAtomic and existing code is causing > crashes in tests. I've tracked the crashes down to a conflict between unit > tests that install a mock version of the @mozilla.org/plugin/host;1 service > (nsIPluginHost), and core C++ code that assumes the plugin host service will > always be the singleton C++ nsPluginHost. > > From bug 657952: > > According to DXR, tests that mock "@mozilla.org/plugin/host;1" are: > > http://dxr.mozilla.org/mozilla-central/source/toolkit/components/telemetry/ > tests/unit/test_TelemetryPing.js#l435 > > http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/ > test/xpcshell/test_pluginchange.js#l121 > > http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/ > test/xpcshell/test_bug455906.js#l182 > > http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/ > test/xpcshell/test_bug514327_3.js#l102 > > http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/ > test/xpcshell/test_bug449027.js#l430 > > http://dxr.mozilla.org/mozilla-central/source/toolkit/mozapps/extensions/ > test/xpcshell/test_duplicateplugins.js#l96 > > > See the comments on Bug 657952 beginning at > https://bugzilla.mozilla.org/show_bug.cgi?id=657952#c5 for additional > information. > > > I need to disable the crashing tests in my patches for bug 853388; I'll > refer to this bug in the xpcshell.ini Can we just fix these as benjamin suggested. Set a pref in those tests with a new contract id and in PluginProvider.jsm have it check the pref and use that contract instead of the regular pluginhost.

Is this something we do often? It feels rather dirty to me... Do we need the pref, or could we just have the code in question try to get "@mozilla.org/plugin/host-TEST-MOCK;1" and fall back to ".../host;1" if the mock isn't there?

Mocking in a complex interdependent system *is* dirty, let's not hide it too much. I don't think we should do automatic fallback unless there's some technical reason we have to.

The fix for bug 881578 has caused the crashes to stop for the time being, because we no longer invoke the plugin handler code for .jsm files.