Bugzilla

(In reply to Nika Layzell [:nika] (ni? for response) from comment #12)
> If we previously would've never created a new process, but are now occasionally creating a new process, I could see this potentially happening, as I believe we only load, parse, and compile a chrome:// URI once per-process (https://searchfox.org/mozilla-central/rev/2c4b830b924f42283632b70f39a60fd36433dd4d/dom/base/nsFrameMessageManager.cpp#1191-1198), so we'll probably only hit the assertion the first time the framescript is loaded in a content process, and subsequent times we'll load it from cache.

But why wouldn't the script load? It must be registered, because the callsite for `loadFrameScript` is in the same chrome package. Are we basically racing the loading of the framescript in the new process with the chrome mapping being synced into the new process from the parent, or something, as this isn't an omni.ja mapping? Is that possible? I would have thought we'd proxy getting the contents of the script to the parent, but maybe that's just me being naive...

(In reply to Nika Layzell [:nika] (ni? for response) from comment #12)
> If we previously would've never created a new process, but are now occasionally creating a new process, I could see this potentially happening, as I believe we only load, parse, and compile a chrome:// URI once per-process (https://searchfox.org/mozilla-central/rev/2c4b830b924f42283632b70f39a60fd36433dd4d/dom/base/nsFrameMessageManager.cpp#1191-1198), so we'll probably only hit the assertion the first time the framescript is loaded in a content process, and subsequent times we'll load it from cache.

But why wouldn't the script load? The chrome package for it must be registered, because the callsite for `loadFrameScript` is in the same chrome package. Are we basically racing the loading of the framescript in the new process with the chrome mapping being synced into the new process from the parent, or something, as this isn't an omni.ja mapping? Is that possible? I would have thought we'd proxy getting the contents of the script to the parent, but maybe that's just me being naive...