Bug 101998 – IndexedDB: non-trivial test conversion batch 2

• Status: RESOLVED
• Resolution: FIXED
• Component: New Bugs

[David Grogan]

IndexedDB: non-trivial test conversion batch 2

[David Grogan]

Created attachment 173740 [details]
Patch

[David Grogan]

Created attachment 173743 [details]
Patch

[David Grogan]

Created attachment 173744 [details]
Patch

[David Grogan]

Josh, could you take a look at these 5 tests?

delete-closed-database-object doesn't work in nrwt/DRT or stand-alone content_shell, only in nrwt/content_shell.  If I let it stay loaded in stand-alone content_shell the connection eventually does get gc'd, allowing the other to proceed.  When you wrote this test, did you struggle with getting gc to collect the dead objects?  Any suggestions?

[kov's GTK+ EWS bot]

Comment on attachment 173744 [details]
Patch

Attachment 173744 [details] did not pass gtk-ews (gtk):
Output: http://queues.webkit.org/results/14803914

[Joshua Bell]

Comment on attachment 173744 [details]
Patch

lgtm...

View in context: https://bugs.webkit.org/attachment.cgi?id=173744&action=review

> LayoutTests/storage/indexeddb/resources/database-basics.js:58
>      setVersionTrans.onabort = checkMetadata;

Can you change the  setVersionTrans.abort(); call following this line to use evalAndLog() ?

> LayoutTests/storage/indexeddb/resources/database-closepending-flag.js:-74
> -        // FIXME: Should verify 'blocked' not fired. http://webkit.org/b/71130

Blocked event doesn't fire in the new code because we (incorrectly) enter a "pending" state before firing blocked when opening connections, right?

[David Grogan]

Comment on attachment 173744 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=173744&action=review

Thanks for the review.  Any thoughts on the delete-closed-database-object troubles from comment 4?

>> LayoutTests/storage/indexeddb/resources/database-basics.js:58
>>      setVersionTrans.onabort = checkMetadata;
> 
> Can you change the  setVersionTrans.abort(); call following this line to use evalAndLog() ?

Will do.

>> LayoutTests/storage/indexeddb/resources/database-closepending-flag.js:-74
>> -        // FIXME: Should verify 'blocked' not fired. http://webkit.org/b/71130
> 
> Blocked event doesn't fire in the new code because we (incorrectly) enter a "pending" state before firing blocked when opening connections, right?

I think we don't fire it because there are no other connections.  The previous connection is closed at line 46/25.

[WebKit Review Bot]

Comment on attachment 173744 [details]
Patch

Attachment 173744 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/14817540

New failing tests:
storage/indexeddb/delete-closed-database-object.html

[David Grogan]

Created attachment 174016 [details]
Patch

[David Grogan]

Tony, could you review this?

Patch 4 adds the requested evalAndLog and removes delete-closed-database-object test.

[WebKit Review Bot]

Comment on attachment 174016 [details]
Patch

Clearing flags on attachment: 174016

Committed r134533: <http://trac.webkit.org/changeset/134533>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.

[Kentaro Hara]

This patch causes a layout test failure on chromium:
http://code.google.com/p/chromium/issues/detail?id=160961

For the time being, we updated expectation results (in order to roll out WebKit DEPS). Would you fix the bug and re-update expectation results? See the above chromium issue for more details.

storage/indexeddb/database-basics.html now shows on content_browsertests. I've temporarily disabled the test in chromium to not block the wk roll even longer

FAIL Error function called unexpectedly: (AbortError) Version change transaction was aborted in upgradeneeded event handler.