[BlackBerry] crash in CookieDatabaseBackingStore.

• Component: WebKit BlackBerry
• Status: RESOLVED FIXED

[Jason Liu]

playing around with the 'setting' on HTML browser
swipe down the 'setting' up and down more than 5 times and going through
different categories. eventually, browser chrome is turn into grey color.

[Jason Liu]

Created attachment 150135 [details]
Patch

[George Staikos]

If the db is null, won't it cause massive behavioural failure?

[Jason Liu]

(In reply to comment #2)
> If the db is null, won't it cause massive behavioural failure?

There are not cookies from database. So users may input something again if this happens.

[Rob Buis]

Comment on attachment 150135 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=150135&action=review

> Source/WebCore/ChangeLog:8
> +        It is hard to reproduce it now. And there is not a bad data base to analyse.

Why is it hard to reproduce? Is there an internal PR?

> Source/WebCore/ChangeLog:9
> +        So add a guard for cookies pointer.

Add an empty line here.

> Source/WebCore/ChangeLog:10
> +        No new tests, behavior has not changed.

Not strictly true, behavior has changed for when cookies is null.

[Jason Liu]

(In reply to comment #4)
> (From update of attachment 150135 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=150135&action=review
> 
> > Source/WebCore/ChangeLog:8
> > +        It is hard to reproduce it now. And there is not a bad data base to analyse.
> 
> Why is it hard to reproduce? Is there an internal PR?
> 
PR 158159, I didn't reproduce it. And Wei said it is hard, too.
> > Source/WebCore/ChangeLog:9
> > +        So add a guard for cookies pointer.
> 
> Add an empty line here.
> 
> > Source/WebCore/ChangeLog:10
> > +        No new tests, behavior has not changed.
> 
> Not strictly true, behavior has changed for when cookies is null.
Because it is hard to reproduce, so I am not sure a test case is needed.

[Rob Buis]

Comment on attachment 150135 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=150135&action=review

R- for now, please mention internal PR and reword the ChangeLog a bit.

>> Source/WebCore/ChangeLog:10
>> +        No new tests, behavior has not changed.
> 
> Not strictly true, behavior has changed for when cookies is null.

I still think you need to explain better, maybe have somebody in the office help you with the ChangeLog?

[Lyon Chen]

I think following message is the real issue, somehow the database file is locked and we are unable to create a table. We need to fix this.

Jul 04 19:25:28    3 10023     1 Thread 8 (warning!): ERROR: Could not create the table to store the cookies into. No cookie will be stored!
/home/lyon/work/rebase/webkit/Source/WebCore/platform/blackberry/CookieDatabaseBackingStore/CookieDatabaseBackingStore.cpp(223) : void WebCore::CookieDatabaseBackingStore::invokeOpen(const WTF::String&)

Jul 04 19:25:28    3 10023     1 Thread 8 (warning!): ERROR: SQLite Error Message: database is locked

[Jason Liu]

Created attachment 151622 [details]
Patch

[Jason Liu]

Created attachment 151642 [details]
Patch

[George Staikos]

Comment on attachment 151642 [details]
Patch

Seems ok.  Technically you could have put the delete inside the if (cookies) too but it doesn't make any practical difference.

[WebKit Review Bot]

Comment on attachment 151642 [details]
Patch

Clearing flags on attachment: 151642

Committed r122328: <http://trac.webkit.org/changeset/122328>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.