424684 - GTK fires spurious drag exit event after a drop

Status: RESOLVED FIXED

(Core :: Widget: Gtk, defect)

Description

[[e-mail]]

nsEventStateManager fires a drag exit event after a drop. However the GTK code fires its own drag exit event. This confuses code that expects matching drag events. For instance, attachment 310239 [details] [diff] [review] in bug 420341 does not work on GTK builds. Note that it is possible to work around the problem (attachment 311270 [details] [diff] [review] does just that) but I don't think our events should differ between platforms.

Comment 1

[[e-mail]]

Attachment: Proposed patch

Comment 2

[Serge Gautherie (:sgautherie)]

'in‑testsuite=?':
I think checking what the events are and their order would be wanted, if possible.
For example, attachment 310239 [details] [diff] [review] in bug 420341 depends on getting only one DragExit event, and that it happens after the Drop event.

Comment 3

[Damon Sicore (:damons)]

Not blocking as it's not a regression.  Also, won't take the patch unless there are sufficient tests here per comment #2.

Comment 4

[Andrew Schultz]

It seems that mochitest's synthesizeMouse is not sufficient to fool drag 'n drop.  I can trigger ondraggesture, but beyond that drag/drop code (conceivably in widget/) couldn't care less about the synthesized events (since it's working off the real cursor).

Comment 5

[[e-mail]]

(In reply to comment #2)
>For example, attachment 310239 [details] [diff] [review] in bug 420341 depends on getting only one
>DragExit event, and that it happens after the Drop event.
If we check in that attachment with this patch, that could then become a litmus test in SeaMonkey for this bug; would that suffice?

Comment 6

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 311282 [details] [diff] [review]
Proposed patch

a=beltzner

Comment 7

[Mike Beltzner [:beltzner, not reading bugmail]]

Sure, describing a Litmus case is fine ... just make sure to do it :)

Comment 8

[[e-mail]]

Fix checked in.

Comment 9

[Andrew Schultz]

I added this as litmus testcase 5274