Bug 94119 – VoidCallback should not be a special snowflake

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

[Csaba Osztrogonác]

https://trac.webkit.org/changeset/125592 made storage/websql/database-lock-after-reload.html
flakey on Qt. I got it with manual bisecting. Sometimes it is flakey (TEXT PASS) and sometimes
it fails twice with the following diff:

--- /ramdisk/qt-linux-64-release/build/layout-test-results/storage/websql/database-lock-after-reload-expected.txt 
+++ /ramdisk/qt-linux-64-release/build/layout-test-results/storage/websql/database-lock-after-reload-actual.txt 
@@ -1,3 +1,1 @@
-Inserting some data
-Test part 2 Complete


I saw the following flakiness on the GTK bot:

--- /home/slave/webkitgtk/gtk-linux-64-release/build/layout-test-results/storage/websql/database-lock-after-reload-expected.txt 
+++ /home/slave/webkitgtk/gtk-linux-64-release/build/layout-test-results/storage/websql/database-lock-after-reload-actual.txt 
@@ -1,3 +1,4 @@
-Inserting some data
-Test part 2 Complete
+CONSOLE MESSAGE: line 52: TypeError: 'undefined' is not an object (evaluating 'database.transaction')
+Error - could not open database
+Test part 1 Complete
 

Could you check it, please?

[Csaba Osztrogonác]

You can easily reproduce it on Qt:
Tools/Scripts/old-run-webkit-tests storage/websql/database-lock-after-reload.html --iterations 100 --singly

32 test cases (32%) succeeded
68 test cases (68%) had incorrect layout

[Adam Barth]

dhcp-172-31-55-91:abarth-webkit abarth$ Tools/Scripts/run-webkit-tests --debug storage/websql/database-lock-after-reload.html --iterations 1000
Using port 'mac-future'
Test configuration: <future, x86_64, debug>
Placing test results in None
Baseline search path: mac -> generic
Using Debug build
Pixel tests disabled
Regular timeout: 35000, slow test timeout: 175000
Command line: /Users/abarth/git/abarth-webkit/WebKitBuild/Debug/DumpRenderTree -

Found 1 test; running 1000 (1000 times each: --repeat-each=1 --iterations=1000), skipping -999.
Running 1 DumpRenderTree over 1 shard.

[233/1000] storage/websql/database-lock-after-reload.html failed unexpectedly (text diff)
[518/1000] storage/websql/database-lock-after-reload.html failed unexpectedly (text diff)
[551/1000] storage/websql/database-lock-after-reload.html failed unexpectedly (text diff)
[684/1000] storage/websql/database-lock-after-reload.html failed unexpectedly (text diff)
                                                                
Retrying 1 unexpected failure(s) ...

Running 1 DumpRenderTree over 1 shard.

                                                    
996 tests ran as expected, 4 didn't:


Unexpected flakiness: text failures (1)
  storage/websql/database-lock-after-reload.html = TEXT PASS

[Adam Barth]

--- /Users/abarth/git/abarth-webkit/WebKitBuild/Debug/layout-test-results/storage/websql/database-lock-after-reload-expected.txt
+++ /Users/abarth/git/abarth-webkit/WebKitBuild/Debug/layout-test-results/storage/websql/database-lock-after-reload-actual.txt
@@ -1,3 +1 @@
-Inserting some data
-Test part 2 Complete

[Adam Barth]

Looks like it's flaky on apple-mac as well, but much less so.

[Zan Dobersek]

http://test-results.appspot.com/dashboards/flakiness_dashboard.html#group=%40ToT%20-%20webkit.org&tests=storage%2Fwebsql%2Fdatabase-lock

Stating the obvious, at least for the Qt and GTK ports, it's only flaky in the release builds.

[Adam Barth]

> Stating the obvious, at least for the Qt and GTK ports, it's only flaky in the release builds.

Oh!  Thanks for stating that.  It's wasn't obvious to me!  :)

[Adam Barth]

Ok.  It's much flakier release, even on apple-mac

88 tests ran as expected, 12 didn't:

[Adam Barth]

I understand the issue.  Looking for the right fix.

[Adam Barth]

Created attachment 158654 [details]
Patch

[Eric Seidel (no email)]

Comment on attachment 158654 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=158654&action=review

LGTM.  Except the XCode oddity of course.

> Source/WebCore/WebCore.xcodeproj/project.pbxproj:1963
> +		7C522D4C15B478B2009B7C95 /* (null) in Headers */ = {isa = PBXBuildFile; settings = {ATTRIBUTES = (Private, ); }; };

This looks wrong.

> Source/WebCore/WebCore.xcodeproj/project.pbxproj:23634
> +				7C522D4C15B478B2009B7C95 /* (null) in Headers */,

This also looks wrong.

[Eric Seidel (no email)]

Comment on attachment 158654 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=158654&action=review

> Source/WebCore/Modules/filesystem/SyncCallbackHelper.h:100
> +        virtual bool handleEvent()

OVERRIDE?

[Adam Barth]

(In reply to comment #11)
> (From update of attachment 158654 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=158654&action=review
> 
> > Source/WebCore/Modules/filesystem/SyncCallbackHelper.h:100
> > +        virtual bool handleEvent()
> 
> OVERRIDE?

I actually had this locally but removed it when uploading.  :)

[Adam Barth]

Created attachment 158656 [details]
Patch for landing

[Adam Barth]

Let's give the EWS a look before landing.

[WebKit Review Bot]

Comment on attachment 158656 [details]
Patch for landing

Attachment 158656 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/13503799

[Build Bot]

Comment on attachment 158656 [details]
Patch for landing

Attachment 158656 [details] did not pass win-ews (win):
Output: http://queues.webkit.org/results/13517147

[Peter Beverloo (cr-android ews)]

Comment on attachment 158656 [details]
Patch for landing

Attachment 158656 [details] did not pass cr-android-ews (chromium-android):
Output: http://queues.webkit.org/results/13514184

[Adam Barth]

Created attachment 158666 [details]
Patch

[WebKit Review Bot]

Comment on attachment 158666 [details]
Patch

Attachment 158666 [details] did not pass chromium-ews (chromium-xvfb):
Output: http://queues.webkit.org/results/13508408

[Adam Barth]

Created attachment 158673 [details]
Patch

[WebKit Review Bot]

Comment on attachment 158673 [details]
Patch

Clearing flags on attachment: 158673

Committed r125745: <http://trac.webkit.org/changeset/125745>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.