524548 - Safebrowsing should shut down sooner if possible

Status: RESOLVED INVALID

(Toolkit :: Safe Browsing, defect)

Description

[Ben Turner (not reading bugmail, use the needinfo flag!)]

Attachment: Shutdown stacks

On my windows VM sometimes shutdown takes forever. When I break I see that the main thread is waiting to join the safebrowsing thread, but it's still processing lots of updates and using sqlite extensively. I don't know anything about the architecture here but would it be possible to interrupt this processing and run the updates later?

I'll attach the stacks from both threads.

Comment 1

[David Dahl :ddahl]

I think

Comment 2

[David Dahl :ddahl]

whoops, I think if we switch the mozStorage to executeAsync, we can cancel the current query? is that right sdwilsh?

Comment 3

[David Dahl :ddahl]

I'm wondering if you would create an observer that detects XPCOM shutdown each time you run the db update. If you detect XPCOM shutdown, you call reset() on the statement. Would that rollback the transaction, which would make url-classifier start over on the next startup?

Comment 4

[David Dahl :ddahl]

Looks like there is a "cancelUpdate" call in the dbService:

http://mxr.mozilla.org/mozilla-central/source/toolkit/components/url-classifier/src/nsUrlClassifierDBService.cpp#3289

Comment 5

[Shawn Wilsher :sdwilsh]

You could do that with executeAsync, but in this case it doesn't matter.  You'd just want to stop processing any updates, and do it at next startup.

Comment 6

[David Dahl :ddahl]

Looks like there is an observer registered for xpcom-shutdown-threads

http://mxr.mozilla.org/mozilla-central/source/toolkit/components/url-classifier/src/nsUrlClassifierDBService.cpp#3962

Would we need to add another observer for 'xpcom-shutdown' explicitly?

Comment 7

[Justin Dolske [:Dolske]]

The old sqlite DB is gone, I do hope the new one shuts down faster. :)