Jesse Bunch Reporter
	Description • 11 years ago

(note: the attached test case demonstrates this)

1. Create div#1 with contenteditable=true and tabindex=0
2. Create div#2 with some text
3. Focus the div#1 by calling .focus()
4. Set contenteditable=false on div#1
5. Focus div#2 by calling .focus()

Whammy.

What happens:
div#2 is focused and all of it's text is selected.

What should happen: 
div#2 should gain focus and it's text should not be focused.

Workaround:
Call .blur() on div#1 before setting contenteditable=false

Other Notes:
Reproduces on both OSX (v27) and Windows (v27)

	Mats Palmgren (inactive) Assignee
	Comment 1 • 11 years ago

The editor has this focus/blur listener that manipulates the Selection:
http://hg.mozilla.org/mozilla-central/annotate/7133bb431eba/editor/libeditor/base/nsEditorEventListener.cpp#l955
Among other things, it's installs an ancestor limiter on focus and
removes it on blur.  In the testcase at hand we don't blur though,
we simply disconnect this listener leaving the ancestor limiter in place.
So when the next node is focused we fail to Collapse the selection
because the new node is not a descendent of the limiter and is rejected.
This happens here:
http://hg.mozilla.org/mozilla-central/annotate/7133bb431eba/dom/base/nsFocusManager.cpp#l2124
The AddRange succeeds and selects the contents of the focused node.
The CollapseToStart fails, and leaves the selected state as is.
http://hg.mozilla.org/mozilla-central/annotate/7133bb431eba/dom/base/nsFocusManager.cpp#l2125

https://tbpl.mozilla.org/?tree=Try&rev=5f1951267a2c

	Mats Palmgren (inactive) Assignee
	Comment 2 • 11 years ago

This seems like a rather major bug and I'm surprised we haven't caught it until now.
Is it a regression perhaps?

	Jesse Bunch Reporter
	Comment 3 • 11 years ago

By regressing with previous release builds, it looks like this first happens with Firefox 11.

	(no longer active)
	Comment 4 • 11 years ago

Comment on attachment 8373060 [details] [diff] [review]
fix+reftest

Review of attachment 8373060 [details] [diff] [review]:
-----------------------------------------------------------------

I'm pretty sure that this has been broken since forever. :(  Thanks for the investigation and the patch!

	Mats Palmgren (inactive) Assignee
	Comment 5 • 11 years ago

https://hg.mozilla.org/integration/mozilla-inbound/rev/9009d48cb9c7

	Carsten Book [:Tomcat]
	Comment 6 • 11 years ago

https://hg.mozilla.org/mozilla-central/rev/9009d48cb9c7

	(no longer active)
	Comment 7 • 11 years ago

Mats, how do you feel about uplifting this to Aurora?  (See bug 965690 comment 6)

	Mats Palmgren (inactive) Assignee
	Comment 8 • 11 years ago

This patch is low risk and likely applies as is, so it should be fine to uplift.