772288 - SpecialPowers.wrap(Components).utils.import(url, window) is broken

Status: RESOLVED FIXED

(Core :: XPConnect, defect)

Description

[Matthew N. [:MattN]]

Attachment: Test case

See the attached mochitest-plain test.

Workaround is to do:

const Cu = SpecialPowers.wrap(Components).utils;
const Services = Cu.import("resource://gre/modules/Services.jsm").Services;

Comment 1

[Bobby Holley (:bholley)]

Alternatively, you can do Cu.import(url, window), which imports it into |window|'s scope. See the docs here: https://developer.mozilla.org/en/Components.utils.import

There's not a lot we can do here to fix the bug as filed. Cu.import(url) imports the JSM into the caller's scope. SpecialPowers.wrap explicitly proxies operations into a special chrome scope so that they run with privileges. Matt, do you have any ideas on how this could be made nicer? Or should we WONTFIX?

Comment 2

[Matthew N. [:MattN]]

Attachment: Test case v2 [added Cu.import(url, window) test]

(In reply to Bobby Holley (:bholley) from comment #1)
> Alternatively, you can do Cu.import(url, window), which imports it into
> |window|'s scope. See the docs here:
> https://developer.mozilla.org/en/Components.utils.import

Yep, that didn't work either but I forgot to include it in the testcase.  I guess I should have mentioned that.  I tried:
Cu.import(url, window)
and 
Cu.import(url, this) (which would have been the window) as well.

I've now updated the test case. Is Cu.import(url, window) supposed to work there?

> There's not a lot we can do here to fix the bug as filed. Cu.import(url)
> imports the JSM into the caller's scope. SpecialPowers.wrap explicitly
> proxies operations into a special chrome scope so that they run with
> privileges. Matt, do you have any ideas on how this could be made nicer? Or
> should we WONTFIX?

If Cu.import(url, window) would work then it would be nicer.

Comment 3

[Bobby Holley (:bholley)]

(In reply to Matthew N. [:MattN] from comment #2)

> If Cu.import(url, window) would work then it would be nicer.

Agreed! Let's morph this bug into that issue.

Comment 4

[Bobby Holley (:bholley)]

Ah, so the properties are ending up on the Xray holder for the chrome view on the content window.

I think we should just unwrap here and enter the compartment. Patching coming right up.

Comment 5

[Bobby Holley (:bholley)]

Attachment: Tests. v2 r=bholley

Updated tests. r=bholley

Comment 6

[Bobby Holley (:bholley)]

Attachment: Unwrap target objects for Cu.import. v1

Comment 7

[Blake Kaplan (:mrbkap) (inactive)]

Comment on attachment 640568 [details] [diff] [review]
Unwrap target objects for Cu.import. v1

Review of attachment 640568 [details] [diff] [review]:
-----------------------------------------------------------------

Doesn't this risk introducing bug 653926 here? If I'm reading the component loader properly, we'll end up with scripts with script->principal being system, but the compartment being non-system. I think we need to avoid that.

Comment 8

[Bobby Holley (:bholley)]

Attachment: Waive Xray for target objects in Cu.import. v1

I'm not sure I grok all of the implications of the old approach, but I totally agree
that cavalier unwrapping should be avoided. Let's waive Xray instead.

Comment 9

[Bobby Holley (:bholley)]

(fixed the 'dump callers' typo locally)

Comment 10

[Blake Kaplan (:mrbkap) (inactive)]

Comment on attachment 644344 [details] [diff] [review]
Waive Xray for target objects in Cu.import. v1

Thanks.

Comment 11

[Bobby Holley (:bholley)]

Twiddled with null-ness and repushed to try:
https://tbpl.mozilla.org/?tree=Try&rev=230fd7f7903c

Comment 12

[Bobby Holley (:bholley)]

And that one was busted because I asserted AccessCheck::callerIsChrome(), which fails for tests that use UniversalXPConnect to call Cu.import. Switched to nsContentUtils::CallerHasUniversalXPConnect() and repushed:


https://tbpl.mozilla.org/?tree=Try&rev=cd60bc464319

Comment 13

[Bobby Holley (:bholley)]

Looks green. Pushed to m-i:

http://hg.mozilla.org/integration/mozilla-inbound/rev/8485eb46a4d6
http://hg.mozilla.org/integration/mozilla-inbound/rev/cd5e3b073284

Comment 14

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/cd5e3b073284
https://hg.mozilla.org/mozilla-central/rev/8485eb46a4d6