Bug 35624 - Closing a database while a transaction is in progress can cause a crash
Summary: Closing a database while a transaction is in progress can cause a crash
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: New Bugs (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P2 Normal
Assignee: Dumitru Daniliuc
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-03-02 18:06 PST by Dumitru Daniliuc
Modified: 2010-03-03 17:02 PST (History)
1 user (show)

See Also:

Attachments
patch (1.35 KB, patch)
2010-03-02 18:40 PST, Dumitru Daniliuc 		dglazkov: review+
dumi: commit-queue-
Details | Formatted Diff | Diff
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Dumitru Daniliuc 2010-03-02 18:06:50 PST 
Before running any transaction step on the DB thread or the main thread, we check if the database is closed by calling the SQLTransaction::checkAndHandleClosedDatabase() method. Parts of that method though are intended to run only on the DB thread, and could cause a crash if run on the main thread. In particular, calling SQLTransactionCoordinator::releaseLock() on the main thread after SQLTransactionCoordinator::shutdown() was called on the DB thread triggers an assertion failure.

Solution: check what thread we're on inside SQLTransaction::checkAndHandleClosedDatabase() and only run the appropriate code for that thread.
Comment 1 Dumitru Daniliuc 2010-03-02 18:40:48 PST 
Created attachment 49874 [details]
patch
Comment 2 Michael Nordman 2010-03-02 18:52:13 PST 
yup... lgtm
Comment 3 Dimitri Glazkov (Google) 2010-03-03 13:11:18 PST 
Comment on attachment 49874 [details]
patch

Per our IM discussion, please describe why it's impossible to write a deterministic regression test for it and mention the flakiness in the layout test it fixes.
Comment 4 Dumitru Daniliuc 2010-03-03 17:02:29 PST 
Landed as r55487.