1217047 - make nsComponentManagerImpl::IsContractIDRegistered lie less

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Nathan Froyd [:froydnj]]

Yury and Michael were running into some issues with the component manager.  The following sequence of steps:

* RegisterFactory $CID, $CONTRACT_ID
* UnregisterFactory $CID
* IsContractIDRegistered $CONTRACT_ID

can return true, which seems nonsensical.  It makes a certain amount of sense, since UnregisterFactory doesn't know what contract ID corresponds to $CID.  But we can do better.  I know I've been bitten by this before and wasted a good amount of time figuring out what was going on.

I think we can do better, and a try push with an experimental patch seems to not break anything horribly.  I'll toss that up in a second.

Comment 1

[Nathan Froyd [:froydnj]]

Attachment: try harder in IsContractIDRegistered to return a reasonable answer

We can have a valid nsFactoryEntry in the following cases:

* We registered a known module;
* We registered a factory (mModule will be null in this case);
* We created a service via the factory.

So one of these members must be valid if we still have a entry in the
contract ID table that actually does interesting things.

Comment 2

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 8676950 [details] [diff] [review]
try harder in IsContractIDRegistered to return a reasonable answer

Worth checking whether this fixes bug 1193155.

Comment 3

[Yury Delendik (:yury)]

Comment on attachment 8676950 [details] [diff] [review]
try harder in IsContractIDRegistered to return a reasonable answer

Review of attachment 8676950 [details] [diff] [review]:
-----------------------------------------------------------------

Fixed bug 1193155 for me. (As well as bug 1193088, but it hard to tell if this or bug 1179262 fixed it)

Comment 4

[Benjamin Smedberg]

Comment on attachment 8676950 [details] [diff] [review]
try harder in IsContractIDRegistered to return a reasonable answer

We should remove this API, but as long as we've got it it might as well do something useful.

Comment 5

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/d34f0bd09295

Comment 6

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/d34f0bd09295

Comment 7

[Nathan Froyd [:froydnj]]

Comment on attachment 8676950 [details] [diff] [review]
try harder in IsContractIDRegistered to return a reasonable answer

Approval Request Comment
[Feature/regressing bug #]: bug 1179262
[User impact if declined]: can't toggle PDF.js on and off without restart (see bug 1193155)
[Describe test coverage new/current, TreeHerder]: We have a few tests in-tree that exercise code around here, though not the specific additions that were made.
[Risks and why]: Low risk.
[String/UUID change made/needed]: None.

Comment 8

[Liz Henry (:lizzard) (privacy/fingerprinting team)]

Comment on attachment 8676950 [details] [diff] [review]
try harder in IsContractIDRegistered to return a reasonable answer

Restarting is annoying for users; let's uplift this small fix to aurora and beta.

Comment 9

[Liz Henry (:lizzard) (privacy/fingerprinting team)]

Marking as affected for 42+ since this seems to have been a problem since at least 40.

Comment 10

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-aurora/rev/090108f1c594

Comment 11

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/releases/mozilla-beta/rev/c66289e84c50

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-b2g44_v2_5/rev/090108f1c594