View Full Version : Major usability problem: insensitive screen edges
denisw
February 2nd, 2008, 11:01 AM
Hi,
Although I like Compiz very much, one thing about it is very annoying: when using it, a one-pixel border around the screen is insensitive to clicks. This is a big violation of Fitt's Law: for instance, you cannot make a quick move downwards to hit a taskbar item because you likely hit the insensitive bottom screen edge. This problem is exclusive to Compiz; with other window managers, clicking on items at the screen edge is just fine.
I opened up a bug about this on freedesktop bugzilla:
https://bugs.freedesktop.org/show_bug.cgi?id=14178
For reference, here is the Ubuntu Launchpad bug about the same problem:
https://bugs.launchpad.net/compiz/+bug/103306
adamk
February 2nd, 2008, 12:43 PM
I'm not sure I understand. I just threw my mouse to the bottom and to select a window in my xfce4 taskbar, and didn't have any problems. I was at the very bottom edge.
simpson-fan
February 2nd, 2008, 12:47 PM
I can confirm this. I'm using Ubuntu Hardy testing.
It's a very nervous bug.
Bye, simpson-fan
edit: look here: https://bugs.launchpad.net/ubuntu/+source/compiz/+bug/103306
adamk
February 2nd, 2008, 01:08 PM
So, from what I see, this problem happens with the right border for me, but not the top or bottom borders.
adam
NoSkill
February 2nd, 2008, 02:17 PM
So, from what I see, this problem happens with the right border for me, but not the top or bottom borders.
adam
Awwh! Now why'd ya have to go mention this you guys, I'm seeing it now too! :(
Using Fedora7, so it's not a distro problem.
From what I'm seeing, you move a window to the left or right edges so that one of the window title icons (close, restore, minimize, sticky, or set above) is on the left or right edges. Then if you move your mouse to the extreme edge, you'll find you won't be able to select the icon until you move in by one pixel. Is that what you guys are seeing too?
Oasisgames
February 2nd, 2008, 02:48 PM
The problem is with Rotate Cube and Wall. Wall has the "problem" in all edges, but Rotate only has it on the left and right. It's caused by the single pixel border these plugins need to detect a window being dragged over to the next viewport.
I really don't know why these are otherwise insensitive, but I can talk with the higher-ups about it.
simpson-fan
February 2nd, 2008, 03:34 PM
The problem is with Rotate Cube and Wall. Wall has the "problem" in all edges, but Rotate only has it on the left and right. It's caused by the single pixel border these plugins need to detect a window being dragged over to the next viewport.
I really don't know why these are otherwise insensitive, but I can talk with the higher-ups about it.
I can confirm this, it's true!
NoSkill
February 6th, 2008, 02:34 AM
The problem is with Rotate Cube and Wall. Wall has the "problem" in all edges, but Rotate only has it on the left and right. It's caused by the single pixel border these plugins need to detect a window being dragged over to the next viewport.
I really don't know why these are otherwise insensitive, but I can talk with the higher-ups about it.
Did you have a talk with the big-wigs? Any reasons given why these plugins are causing the issue?
some-guy
February 6th, 2008, 02:59 AM
It's the same here :(
I wish I had a custom user title, but if given the opportunity to have one, I have no idea what I'd set it to. :|just make ~300-400 posts :p
maniac
February 6th, 2008, 02:25 PM
Did you have a talk with the big-wigs? Any reasons given why these plugins are causing the issue?
Because rotate and wall use those windows to initiate edge actions (Edge flip Pointer/Move/DnD). If you don't need this kind of functionality, disable the corresponding edges in ccsm (Flip left/right/up/down).
SmSpillaz
February 6th, 2008, 10:27 PM
Is there a particular reason as to why clicks cannot be captured in these areas? The only reason I can think of is that these areas have been set up is to trigger an event in the case that the cursor moves over it to prevent mouse polling.
In which case, the problem should be resolvable if/when input redirection makes it out. Or am I mistaken?
Amaranth
February 8th, 2008, 05:02 AM
The problem is you'd have to have these windows capture all events and then for the ones they don't care about figure out what window is below them and emulate an event to that window. It's very ugly, breaks easily, and will not be accepted in compiz.
SmSpillaz
February 8th, 2008, 06:23 AM
That is probably a temporary solution which does sound rather ugly code wise (considering that some windows are created with toolkits that don't really conform to the X standards).
Like I said, am I correct in saying that these edges wait for events? If so, I'd imagine a proper solution would be to wait for input redirection, right?
maniac
February 8th, 2008, 01:49 PM
That is probably a temporary solution which does sound rather ugly code wise (considering that some windows are created with toolkits that don't really conform to the X standards).
Like I said, am I correct in saying that these edges wait for events? If so, I'd imagine a proper solution would be to wait for input redirection, right?
The best solution is waiting for software cursor (which means we have to track the cursor position at all times anyway) and drop the invisible windows completely ;)
SmSpillaz
February 8th, 2008, 02:52 PM
The best solution is waiting for software cursor (which means we have to track the cursor position at all times anyway) and drop the invisible windows completely ;)
Sort-of off-topic, but would this make mouse-polling obsolete?
maniac
February 8th, 2008, 05:35 PM
Sort-of off-topic, but would this make mouse-polling obsolete?
Yes, it would.
b0le
February 8th, 2008, 10:48 PM
Would this be the point of the keyboard, pointer and input objects in the OF branch (they are not actually used for anything yet, at least as far as I can tell from git-grep)
gehrehmee
February 12th, 2008, 10:43 PM
This also hits auto-hide panels on the edges of the screen, or other windows that appear only when mouse is on the edge.
Curiously, these programs don't seem to be affected by brightside, which is able to map events to edge motions, but still allows auto-hide panels to be opened. Maybe it's worth looking at how brightside detects pointers-on-edges?
GGLucas
March 13th, 2008, 06:49 PM
I noticed this while going from ubuntu 7.10 to arch linux, and ubuntu (at least gutsy), doesn't seem to have this problem with the rotate cube or desktop wall plugins enabled, shouldn't that mean that there's a way to make it work properly ? Not being able to click the edges really bothers me, I like desktop wall but I won't enable it while this bug still exists :(
vBulletin® v3.7.3, Copyright ©2000-2008, Jelsoft Enterprises Ltd.