I see somewhat similar behavior as well.  I thought it might be related to
autosave but have no verification yet.  I did confirm the behavior using
emacs -q and will gladly go after the problem with gdb but I would very
much appreciate some suggestions.  I rebuilt twice in the past week using
the latest bazaar changes and saw the behavior in both builds.  My previous
build is about 3 months old and does not show the bad behavior, fwiw.

-pmr


On Mon, Jan 7, 2013 at 2:01 PM, Ivan Andrus <darthandrus@gmail.com> wrote:

> On Jan 7, 2013, at 6:36 PM, Harald Hanche-Olsen <hanche@math.ntnu.no>
> wrote:
>
> > Once in a while, some frame stops receiving keyboard events. It keeps
> > receiving mouse events, so I can move the cursor around and so forth,
> > but typing into the frame has no effect.
> >
> > Apparently, if I just keep typing random junk into the frame, the
> > problem will resolve itself, with most of the input lost. (This is a
> > new observation, not yet thoroughly confirmed. Previously, I would
> > just kill the frame and open a new one.) In any case, the problem
> > only affects one frame, and other frames work as usual.
> >
> > This is happening on OS X, --with-ns, recent builds from trunk. I
> > think the problem has been around for quite a while (i.e., weeks), but
> > I just haven't gotten around to reporting it until now.
> >
> > I'd file a bug report, but the problem is that I have no idea how to
> > reproduce the problem. It just happens at random times.
> >
> > Do other OS X users see this? Can you suggest something I could try to
> > get more information next time it happens? I gather that the event
> > loop is notoriously hard to debug, however.
>
> I have noticed this too.  For me the problem "goes away" when I press a
> letter.  In other words only control or meta (or hyper probably) keys don't
> work. When I press a regular letter then it inputs that character and I am
> able to type control characters as normal.
>
> Sadly, I can't offer any advice on how to debug it, but thought this extra
> data point might be useful.  I haven't noticed any clues as to what might
> cause it, and it's rather rare.
>
> -Ivan
>