Remove Fission-incompatible test of UniversalXPConnect gunk
Categories
(Core :: XPConnect, task, P2)
Tracking
()
Tracking | Status | |
---|---|---|
firefox71 | --- | fixed |
People
(Reporter: kmag, Assigned: kmag)
References
(Blocks 1 open bug)
Details
Attachments
(1 file)
We don't support UniversalXPConnect outside of tests anymore, and even in tests we want to get rid of it, so let's just remove this test rather than updating it.
Assignee | ||
Comment 1•5 years ago
|
||
![]() |
||
Comment 2•5 years ago
|
||
I'm not sure I'm OK with this until we actually stop using UniversalXPConnect...
Should we just disable this test when Fission is enabled for the moment? And if we don't plan to stop using UniversalXPConnect before Fission ships, we probably do want to fix this test....
Assignee | ||
Comment 3•5 years ago
|
||
(In reply to Boris Zbarsky [:bzbarsky, bz on IRC] from comment #2)
I'm not sure I'm OK with this until we actually stop using UniversalXPConnect...
Should we just disable this test when Fission is enabled for the moment? And if we don't plan to stop using UniversalXPConnect before Fission ships, we probably do want to fix this test....
We could just change it to use a cross-origin, same-process frame, but I'd honestly rather just delete it. The only things that still rely on UniversalXPConnect are a few scattered tests, and I don't think we need to ensure it preserves its current exact semantics strictly for their sake.
That said, I also don't think this test makes any kind of sense anymore. It calls enablePrivilege
in a function scope, and assumes that the privilege only lasts to the end of that scope. But that only made sense when enablePrivilege
worked by examining the stack for signed code which had added privileges. What remains of it now just works at the compartment level. It doesn't care about stacks, and it certainly doesn't go away at the end of the scope that it was called in.
![]() |
||
Comment 4•5 years ago
|
||
The only things that still rely on UniversalXPConnect are a few scattered tests
I thought all our Talos tests had a fundamental dependency on UniversalXPConnect to work properly. In fact, that's the one real blocker for removing UniversalXPConnect altogether that I know of.
Will look into the test details tomorrow.
Assignee | ||
Comment 5•5 years ago
|
||
(In reply to Boris Zbarsky [:bzbarsky, bz on IRC] from comment #4)
The only things that still rely on UniversalXPConnect are a few scattered tests
I thought all our Talos tests had a fundamental dependency on UniversalXPConnect to work properly. In fact, that's the one real blocker for removing UniversalXPConnect altogether that I know of.
There are a few scattered parts of talos that still use it (mainly the profiler code and the a11y tests), but I wouldn't call it fundamental. In any case, none of them should be using it to reach into cross-origin documents (and from what I can tell, none of them do). If they do, they'll be broken by Fission long before they're broken by any changes to UniversalXPConnect.
Updated•5 years ago
|
Updated•5 years ago
|
Comment 7•5 years ago
|
||
bugherder |
Description
•