201342 - Drag initiation should use system drag thresholds where available

Status: RESOLVED FIXED

(Core :: DOM: Events, defect)

Description

[saari (gone)]

On Win32 we should use the drag threshold rect that the system is using.
Available via getsystemmetrics SM_CXDRAG and SM_CYDRAG

Do Mac and Linux have similar settings we can query?

Comment 1

[Michael Buckland]

Discussed in edt.  Plussing.

Comment 2

[Mike Pinkerton (not reading bugmail)]

mac has no similar setting, iirc.

Comment 3

[Brian Ryner (not reading)]

I can't find an API for this on Mac either.  By inspection, as you're dragging
the mouse (on a Finder icon), the item deselects after you move 4 pixels; the
item starts to move on the 5th pixel.

On linux, the threshold is hardcoded in gtk+ (1.x, anyway) to start the drag if
you've moved more than 3 pixels in any direction (gtk+ 1.2.10, gtkdnd.c, line
2433).  blizzard, is there something better we can do for gtk2?

Comment 4

[Brian Ryner (not reading)]

Ah, indeed there is.  We can do something like:

  gint drag_threshold;
  g_object_get (gtk_widget_get_settings (widget),
                "gtk-dnd-drag-threshold", &drag_threshold,
                NULL);    

on gtk2.

Comment 5

[Brian Ryner (not reading)]

Attachment: patch

Use system setting for drag threshold on Win32 and gtk2; hardcode observed
threshold for Mac and gtk1.

Comment 6

[Christopher Blizzard (:blizzard)]

Comment on attachment 122233 [details] [diff] [review]
patch


>+            gint threshold;
>+            g_object_get(gtk_widget_get_settings(box),
>+                         "gtk-dnd-drag-threshold", &threshold,
>+                         NULL);

Initialize threshold to zero in case that object_get() fails?  Other than that,
r=blizzard.

Comment 7

[Brian Ryner (not reading)]

Comment on attachment 122233 [details] [diff] [review]
patch

Can we get this in for 1.4b? It's a pretty low-risk fix that makes our behavior
more platform-native.

Comment 8

[(not reading, please use [e-mail] instead)]

Comment on attachment 122233 [details] [diff] [review]
patch

a=sspitzer

I know that nsEventStateManager.cpp just loves to regress, but if bryner thinks
it is low risk, that's enough for me.

Comment 9

[Brian Ryner (not reading)]

checked in.