bug#17650: problem with emacs bzr change 117190

bug#17650: problem with emacs bzr change 117190

[Stefan Monnier]

>   * lisp/subr.el (sit-for): Remove universal-arg dependency.
> Stefan,

> This change is breaking at least one of my programs (flash-paren.el) such
> that whenever it is in a sit-for and I press a key, I get the error

> 	<t> is not defined

> in the echo area.  It seems like something bogus being pushed onto 
> unread-command-events although I haven't narrowed it down yet; it's
> difficult to debug functions run via idle timers.

This might be related to bug#17650 (which I haven't really investigated
yet).  Can you provide a recipe?


        Stefan

bug#17650: problem with emacs bzr change 117190

[Stefan Monnier]

>>>>> "Noah" == Noah Friedman <noah@splode.com> writes:

> Sorry to keep bugging you about this.  I think I know where the problem is
> now.

> In keyboard.c:read_char around line 2847:

>       /* If this has become non-nil here, it has been set by a timer
>          or sentinel or filter.  */
>       if (CONSP (Vunread_command_events))
>         {
>           c = XCAR (Vunread_command_events);
>           Vunread_command_events = XCDR (Vunread_command_events);
>         }

> This second time that unread-command-events is examined, it's not checking
> for the case that c is a cons.

A `cons' is perfectly normal, but indeed it needs to check if it's
a `cons' whose `car' is t, thanks.  The patch below seems to fix
Thierry's case, so hopefully it fixes yours too.


        Stefan

bug#17650: problem with emacs bzr change 117190

[Noah Friedman]

Indeed it does.  Thanks!

In <jwv4mzyygxq.fsf-monnier+emacsbugs@gnu.org>, Stefan Monnier writes:
>>>>>> "Noah" == Noah Friedman <noah@splode.com> writes:
>
>> Sorry to keep bugging you about this.  I think I know where the problem is
>> now.
>
>> In keyboard.c:read_char around line 2847:
>
>>       /* If this has become non-nil here, it has been set by a timer
>>          or sentinel or filter.  */
>>       if (CONSP (Vunread_command_events))
>>         {
>>           c = XCAR (Vunread_command_events);
>>           Vunread_command_events = XCDR (Vunread_command_events);
>>         }
>
>> This second time that unread-command-events is examined, it's not checking
>> for the case that c is a cons.
>
>A `cons' is perfectly normal, but indeed it needs to check if it's
>a `cons' whose `car' is t, thanks.  The patch below seems to fix
>Thierry's case, so hopefully it fixes yours too.
>
>
>        Stefan