Re: Pretest next week

[Adrian Robert <adrian.b.robert@gmail.com>]

On Jan 22, 2009, at 3:49 PM, Chong Yidong wrote:

> YAMAMOTO Mitsuharu <mituharu@math.s.chiba-u.ac.jp> writes:
>
>> Still the Cocoa/GNUstep port doesn't handle C-g properly.  I think
>> this issue should be addressed before the pretest because it is
>> expected to require nontrivial changes in its event handling part.
>
> Adrian, could you comment?  Thanks.

Sure.  I don't think "nontrivial changes" would be required.  As I  
outlined earlier  (http://thread.gmane.org/gmane.emacs.devel/105688/ 
focus=105694):


1) Complete removal of all Feval() calls in bad places as discussed  
earlier on this list.  A quick check reveals four places, relating  
to: emacs termination, deadkey input handling, toolbar toggle, and  
preferences-help.

- Mostly straightforward (but taking some time) by using custom  
'nonascii-keystroke' events, except for the case of input-method  
handling when in isearch mode, where passing such an event breaks out  
of the isearch entry; I could use some help on this.


2) Go through the code comparing nsterm and macterm input handling to  
make sure all ctrl-g related processing is similar

- I did this at one point, but it should be re-checked if problems  
remain after (1) and (3).


3) Make sure some kind of interrupt mechanism is in place to pick up  
ctrl-g events when emacs core is busy processing and does not itself  
make passes through the NS code event loop (colored spinning disk  
shown in gui).

- I did some experimentation for (3) using a signal handler a while  
ago but something didn't seem to work, and some investigation is  
needed.  I don't think major changes to the port event-handling would  
be needed  since the issue is interrupts during emacs core  
processing, not inside the Cocoa processing.  It should just come  
down to getting whichever interrupt mechanism worked in the Carbon  
port on OS X during emacs-core processing to be active in the Cocoa  
port.  I would welcome help on this.


Recently I've been spending the time I've had for the port on fixing  
the bugs reported by users (http://emacsbugs.donarmstrong.com/cgi-bin/ 
pkgreport.cgi?package=ns), so I haven't made any progress on these yet.