Re: doubling of input events on focus (X, 21.1 and later)

Re: doubling of input events on focus (X, 21.1 and later)

[Richard Stallman]

    upgraded from 20.7, about once every 5-10 times that I bring the emacs
    window to the forefront (focus the window, make it active), the next
    keystroke I enter will be doubled.

Nobody else has reported this.  Has anyone else observed this problem?

After this happens next time, please type C-h l and see if the
character appears twice there.  That will at least begin 
to diagnose where the problem happens.

Re: doubling of input events on focus (X, 21.1 and later)

[Ami Fischman]

Richard Stallman <rms@gnu.org> writes:

[...]

> After this happens next time, please type C-h l and see if the
> character appears twice there.  That will at least begin 
> to diagnose where the problem happens.

I've done this before, and the lossage buffer indeed shows the doubled
character (as if I'd typed it in twice).  I used to think that maybe this
was jittery fingers or a defective keyboard, but this symptom ONLY happens
w/ emacs, and ONLY when switching focus to it from another X app.  Also, it
happens sometimes while I'm typing common words that I never mis-spell
(such as "mail" or certain email addresses), so I'm almost 100% confident
the problem is in software somewhere...

FWIW, I'm using XFree86-4.2.0 on a modified RH7.2 PC.

I wrote the above, then saved the draft, hoping that the behaviour will
occur again soon.  I did some typing, then switched to Gnus' *Group* buffer,
and typed "4 g" (which gets new news of level 4 and lower).  Switched away
from emacs (waiting for gnus to do it's thing, since it involves several
remote nntp servers).  When I switched back, I hit the up arrow key, and
instead of moving up one line, it moved up three lines.  C-h l shows:
4 g <up> <f12> l

Any ideas?  Any other kind of debugging I can do next time this happens? 
Thanks,
-- 
  Ami Fischman
  usenet@fischman.org

Re: doubling of input events on focus (X, 21.1 and later)

[Ami Fischman]

Ami Fischman <usenet@fischman.org> writes:

> I've done this before, and the lossage buffer indeed shows the doubled
> character (as if I'd typed it in twice).  I used to think that maybe this

Grrp!  Sorry.  Meant to say that the lossage buffer DOES NOT show the
doubled char (as the actual kill/yank lower in the post shows) -- the
keystroke shows up only once but is "acted upon" (executed? followed?)
multiple times.

Sorry for the brain-o.
-- 
  Ami Fischman
  usenet@fischman.org

Re: doubling of input events on focus (X, 21.1 and later)

[Richard Stallman]

    Grrp!  Sorry.  Meant to say that the lossage buffer DOES NOT show the
    doubled char (as the actual kill/yank lower in the post shows) -- the
    keystroke shows up only once but is "acted upon" (executed? followed?)
    multiple times.

In that case, I suspect that the character got put on
unread-command-events and re-executed.

Can you find anything in common between the commands that
preceded the commands that got duplicated?

Re: doubling of input events on focus (X, 21.1 and later)

[Ami Fischman]

Richard Stallman <rms@gnu.org> writes:

[...]

> In that case, I suspect that the character got put on
> unread-command-events and re-executed.

Hmm.  Didn't know about that variable.  

> Can you find anything in common between the commands that
> preceded the commands that got duplicated?

Well, the time I most often notice this behaviour is right after I get new
news from foreign nntp servers in gnus (b/c it takes 30-40 seconds to do, I
often switch to another window and then return to the emacs window).

I am fairly certain this behaviour has also happened following some
activity other than getting new news, but I can't recall a specific
instance.

Thanks,
-- 
  Ami Fischman
  usenet@fischman.org

Re: doubling of input events on focus (X, 21.1 and later)

[Richard Stallman]

    Well, the time I most often notice this behaviour is right after I get new
    news from foreign nntp servers in gnus (b/c it takes 30-40 seconds to do, I
    often switch to another window and then return to the emacs window).

If you do this and do not switch windows, is the next command ever
duplicated?

    I am fairly certain this behaviour has also happened following some
    activity other than getting new news, but I can't recall a specific
    instance.

Can you be on the lookout in regard to this?

Re: doubling of input events on focus (X, 21.1 and later)

[Ami Fischman]

Richard Stallman <rms@gnu.org> writes:

>   Well, the time I most often notice this behaviour is right after I get new
>   news from foreign nntp servers in gnus (b/c it takes 30-40 seconds to do, I
>   often switch to another window and then return to the emacs window).
>
> If you do this and do not switch windows, is the next command ever
> duplicated?

No.

>
>     I am fairly certain this behaviour has also happened following some
>     activity other than getting new news, but I can't recall a specific
>     instance.
>
> Can you be on the lookout in regard to this?

Yep.  In fact, I've been looking for this for a while (a week), but have
yet to catch it happening.

-- 
  Ami Fischman
    usenet@fischman.org

Re: doubling of input events on focus (X, 21.1 and later)

[Richard Stallman]

    > If you do this and do not switch windows, is the next command ever
    > duplicated?

    No.

In that case, I can speculate that something in the handling of
switch-frame events might be responsible.

Re: doubling of input events on focus (X, 21.1 and later)

[Ami Fischman]

Richard Stallman <rms@gnu.org> writes:

[...]

> In that case, I can speculate that something in the handling of
> switch-frame events might be responsible.

Not sure if you mean the gnus handling of switch-frame or emacs.  I can
only find one reference to switch-frame in the gnus lisp code, and I don't
believe that that comes into play (it is in gnus-salt.el, in the function
handling mouse-clicks, which I'm not using at all to produce this
behaviour, and the function is not bound to anything else).  

Thanks for following up on this.  Any ideas on how to debug the
switch-frame event handling?

-- 
  Ami Fischman
  usenet@fischman.org

Re: doubling of input events on focus (X, 21.1 and later)

[Richard Stallman]

    Thanks for following up on this.  Any ideas on how to debug the
    switch-frame event handling?

It is not easy to debug unreproducible problems like this.  The only I
have is to add debugging code to record histories of various actions,
so you can examine them after you see the problem.  That way you can
learn about what happened inside Emacs at various points where the
history was recorded.