Bug 54697 - REGRESSION(r78685-r78794): Mouse events no longer fired on canvas
Summary: REGRESSION(r78685-r78794): Mouse events no longer fired on canvas
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: Canvas (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P1 Blocker
Assignee: Dominic Mazzoni
URL: http://www.raysanweb.com/main.htm
Keywords:
Depends on: 50126 55406
Blocks:
  Show dependency treegraph
 
Reported: 2011-02-17 17:40 PST by sheets
Modified: 2011-02-28 13:36 PST (History)
9 users (show)

See Also:


Attachments
Fix problem and add new layout test to prevent future regression. (5.14 KB, patch)
2011-02-22 16:18 PST, Dominic Mazzoni
kbr: review-
Details | Formatted Diff | Diff
Add skips for all wk2 ports that don't implement eventSender. (7.15 KB, patch)
2011-02-22 16:55 PST, Dominic Mazzoni
no flags Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description sheets 2011-02-17 17:40:55 PST
Many mouse events (down, up, move... more?) are no longer fired on WebGL canvas contexts.

To reproduce, navigate to http://www.raysanweb.com/main.htm and click on the top-rightmost button in the menu block to turn on the WebGL menu.

Notice that mouseover highlights and mouse presses are not registered. Compare to r78685 (or Chrome 9) in which this behavior works.

In brief testing, it appears that the event handlers are never called.

I am uncertain which subsystem is responsible and have left this as a "New Bug".
Comment 1 sheets 2011-02-21 17:20:17 PST
Still in r79249
Comment 2 Antoine Mercadal 2011-02-22 02:03:20 PST
confirmed in r79278
Comment 3 Antoine Mercadal 2011-02-22 02:21:33 PST
Also, I confirm that this problem appear first in r78794 (and follows) and is also appear in 2D canvas. For example, you can try http://primalmotion.github.com/VNCCappuccino-example/ (manual here https://github.com/kanaka/noVNC).

In this example, you should be able to move the mouse.
Comment 4 Kenneth Russell 2011-02-22 11:51:49 PST
Confirmed with the latest WebKit nightly build.
Comment 5 Kenneth Russell 2011-02-22 14:04:21 PST
This affects the latest Chromium builds as well.
Comment 6 Dominic Mazzoni 2011-02-22 14:18:02 PST
Looks like my patch was responsible. I'll take a look.
Comment 7 Kenneth Russell 2011-02-22 14:36:39 PST
I've confirmed by rolling out r78789 locally that that patch introduced the regression.

If the fix will take a long period of time the appropriate thing to do would be to roll out that patch and then re-apply it once the fix is ready.
Comment 8 Dominic Mazzoni 2011-02-22 15:57:14 PST
I have a patch that fixes it, just checking layout tests now...
Comment 9 Dominic Mazzoni 2011-02-22 16:18:17 PST
Created attachment 83408 [details]
Fix problem and add new layout test to prevent future regression.
Comment 10 Kenneth Russell 2011-02-22 16:34:06 PST
Comment on attachment 83408 [details]
Fix problem and add new layout test to prevent future regression.

Looks good overall. Thanks for adding the layout test. To keep the bots green you'll need to add your new test at least to LayoutTests/platform/mac-wk2/Skipped, because it looks like WebKit2 doesn't implement EventSender yet. Might be a good idea to add it to the other wk2 ports' Skipped lists too.
Comment 11 Dominic Mazzoni 2011-02-22 16:55:57 PST
Created attachment 83416 [details]
Add skips for all wk2 ports that don't implement eventSender.
Comment 12 Kenneth Russell 2011-02-22 16:58:18 PST
Comment on attachment 83416 [details]
Add skips for all wk2 ports that don't implement eventSender.

Looks good to me.
Comment 13 WebKit Commit Bot 2011-02-23 01:31:11 PST
Comment on attachment 83416 [details]
Add skips for all wk2 ports that don't implement eventSender.

Clearing flags on attachment: 83416

Committed r79425: <http://trac.webkit.org/changeset/79425>
Comment 14 WebKit Commit Bot 2011-02-23 01:31:16 PST
All reviewed patches have been landed.  Closing bug.