bug#17293: 24.4.50; C-g often fails in minibuffer under X

bug#17293: 24.4.50; C-g often fails in minibuffer under X

[raman]

I remember Stefan addressing this a while ago, but it appears to
bite me only when running under X.

Specifically:
When running under X (Ubuntu Precise) and using StumpWM  as the
window manager (neither of those specific bits might be relevant) 
often, if I have a minibuffer completion  up, hitting C-g does
not dismiss the minibuffer, I actually have to switch away from
the minibuffer and type M-x top-level to get emacs back to being normal.

bug#17293: 24.4.50; C-g often fails in minibuffer under X

[Stefan Monnier]

> often, if I have a minibuffer completion  up, hitting C-g does
> not dismiss the minibuffer, I actually have to switch away from

If you're in the minibuffer, then C-g should normally dismiss the
minibuffer, yes.  Of course, we have the problem that C-g only does that
if it runs keyboard-quit (i.e. if Emacs was idle), and not if it signals
"quit" (i.e. Emacs was busy with something).

So apparently, there's been more cases where Emacs is "busy" (probably
just waiting in a sit-for or something).  So to fix this, we need to
figure out what that C-g interrupts.


        Stefan

bug#17293: 24.4.50; C-g often fails in minibuffer under X

[T.V Raman]

Stefan, Makes sense.

I've been chasing a separate problem in emacs  at work that is
keeping emacs busy --see below.

I run emacs-jabber to connect to Google Talk. 

I have a large number of contacts there -- order of magnitude
100s  -- many of whom constantly keep changing status.

emacs-jabber uses a continues fsm ping mechanism and if I leave
emacs idle ie dont touch the keyboard, profiler shows that emacs
is permanently busy in that handler.  I've been trying various
advice tricks to keep that fsm tick from happening except  when
jabber  roster is selected, and surprize surprize that got the
c-g problem to diminish -- I could not figure out why till I saw
your note today.

Stefan Monnier writes:
 > > often, if I have a minibuffer completion  up, hitting C-g does
 > > not dismiss the minibuffer, I actually have to switch away from
 > 
 > If you're in the minibuffer, then C-g should normally dismiss the
 > minibuffer, yes.  Of course, we have the problem that C-g only does that
 > if it runs keyboard-quit (i.e. if Emacs was idle), and not if it signals
 > "quit" (i.e. Emacs was busy with something).
 > 
 > So apparently, there's been more cases where Emacs is "busy" (probably
 > just waiting in a sit-for or something).  So to fix this, we need to
 > figure out what that C-g interrupts.
 > 
 > 
 >         Stefan