Bug 70485 – [GTK] fast/events/drag-selects-image.html fails

• Component: UI Events
• Status: RESOLVED FIXED

[Philippe Normand]

Test added in http://trac.webkit.org/changeset/97878 fails consistently:

--- /home/slave/webkitgtk/gtk-linux-64-debug/build/layout-test-results/fast/events/drag-selects-image-expected.txt 
+++ /home/slave/webkitgtk/gtk-linux-64-debug/build/layout-test-results/fast/events/drag-selects-image-actual.txt 
@@ -11,6 +11,4 @@
 Dragging image in non-editable area...
 0 range(s) selected
 Dragging image in editable area...
-1 range(s) selected
-imageTwo is selected


I'll skip it for now.

[Arko Saha]

Created attachment 115550 [details]
Proposed patch

[Philippe Normand]

So the test is wrong? Can you confirm, Daniel?

[Daniel Cheng]

Do you know why your change fixes the issue for GTK? The result should be the same, as long as we're exceeding the image drag hysteresis.

[Arko Saha]

I tried to reproduce the issue in GtkLauncher. Following are the steps:
1. Drag the image with id: "imageTwo" out of the browser window, and drag it back into the GtkLauncher.
2. Drop the image into GtkLauncher.
3. The image is loaded in the browser window as main url.

Same is observed while debugging DRT.

Also in the test case if we comment out the line below :
dragNowhere(document.getElementById('imageOne')); then the test case executes properly. 

I think this is an issue with DRT or Gtk platform and not in the Webcore drag drop module.
Please let me know your thoughts on this.

[Daniel Cheng]

Hm. It's strange that commenting out the first test case helps it work. If unintended navigation is causing the GTK DRT to fail, one thing you could try is setting an ondrop handler for the page and preventing the default.

[Ryosuke Niwa]

Comment on attachment 115550 [details]
Proposed patch

View in context: https://bugs.webkit.org/attachment.cgi?id=115550&action=review

> LayoutTests/ChangeLog:8
> +        Fixed typo error.

This isn't really a typo, right? Please explain what's wrong with the existing code and how you fixed it.

[Eric Seidel (no email)]

Comment on attachment 115550 [details]
Proposed patch

Yes, this needs more explanation.  Otherwise the change looks reasonable.  Do other platforms not fail this?

[Martin Robinson]

Seems to be passing now.