Open Bug 805694 Opened 12 years ago Updated 2 years ago

Attempts to use the parent side of an IPC channel to obtain an app ID, private browsing status, will fail after OnStopRequest

Categories

(Core :: Networking: HTTP, defect, P3)

x86_64
Linux
defect

Tracking

()

People

(Reporter: jdm, Unassigned)

Details

(Whiteboard: [necko-backlog])

Attachments

(1 file)

The actual channel calls ReleaseListeners in OnStopRequest, so there's no remaining link between it and the serialized load context in the parent actor.
This was worse than I expected. I feel dirty.
Attachment #675395 - Flags: review?(jduell.mcbugs)
Assignee: nobody → josh
Blocks: 722850
Comment on attachment 675395 [details] [diff] [review] Retain parent HTTP channel callbacks as long as the child might initiate operations that require them. Review of attachment 675395 [details] [diff] [review]: ----------------------------------------------------------------- ::: netwerk/protocol/http/HttpChannelParent.cpp @@ +387,5 @@ > +HttpChannelParent::RecvNotifyStartRequestComplete() > +{ > + if (mChannel) { > + bool pending; > + mChannel->IsPending(&pending); I am a bit worried about using IsPending. I'd rather have two state flags in HttpChannelParent telling you that you have received NotifyStartRequestComplete and OnStoprequest. When both are set -> do the cleanup. Nit: the cleanup code (3 lines) deserves its own method.
Comment on attachment 675395 [details] [diff] [review] Retain parent HTTP channel callbacks as long as the child might initiate operations that require them. I'm pretty sure this issue no longer matters once bug 805616 is fixed.
Attachment #675395 - Flags: review?(jduell.mcbugs)
No longer blocks: 722850
Assignee: josh → nobody
Whiteboard: [necko-backlog]
Priority: -- → P1
Priority: P1 → P3
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: