Re: [RFC]: Window focus changes

[Benhur Stein]

On 2003-10-31 03:28:40 -0200 Adam Fedor <fedor@doc.com> wrote:

> Perhaps you can try with --GNU-Debug=NSKeyEvent and we could try to
> interpret what is going on with the keysyms.
Found it!

I have the following defaults for the control keys:

benhur@piazito gui $ defaults read NSGlobalDomain|grep Key
NSGlobalDomain GSFirstControlKey 'Control_L'
NSGlobalDomain GSSecondAlternateKey 'Meta_L'
NSGlobalDomain GSSecondCommandKey 'No Symbol'
NSGlobalDomain GSSecondControlKey 'Control_R'
NSGlobalDomain GSFirstAlternateKey 'Alt_R'
NSGlobalDomain GSFirstCommandKey 'Alt_L'

I don't know why, but Meta_L and Alt_L have the same keyCode (64) 
making the second alternate key
the same as the first command key.

When a new window is created with Command-N (Alt_L), the new window 
receives a KeyMapNotify event,
where it initializes _command_pressed and _alt_pressed, because it 
if's are independent.
When the key is released, only _command_pressed is reset, because in 
process_key_event the
if's are kept together by else's. This leaves _alt_pressed set and 
all the strange keys I was seeing.

I didn't change the code, but I think that a test should be done 
when the defaults are read to see
if two commands are mapped to the same keyCode, and unify the 
treatment, to consider the
keys independently or with a precedence in all cases.

The attached patch contains a small correction on a off-by-one test 
in a vector.
(all this in XGServerEvent.m)

> I can't replicate this, but perhaps I don't understand exactly what 
> you
> are doing.
Stupid me, I had only compiled back-xlib, and I was testing back-art!
Now I can't replicate it neither. Sorry.

> When the window initially gets the click, it calls
> -activateIgnoringOtherApps:, which orders all the windows front and
> makes the key window key again (i.e. the previous key window). Then 
> the
> window that got the click makes itself key.  I'm not sure how to 
> solve
> that one.
Why not just calling activate.. after setting the window as Key?
The attached patch does that.
The activate.. method is put inside the if (!closed). In which 
situation would a closed
window receive a mouseDown? And should it activate the app then?

Benhur

<back.diff><gui.diff>

back.diff
Description: Binary data

gui.diff
Description: Binary data