Re: Toio little error information from widgets

[Lennart Borgman (gmail)]

Lennart Borgman (gmail) wrote:
> Daniel Brockman wrote:
> > "Lennart Borgman (gmail)" <address@hidden> writes:
> >
> > >                ;;(or (eq value 'fundamental-mode)
> > >                    (intern-soft (concat sym-name "-hook"))
> > >                    (boundp (intern-soft (concat sym-name "-hook"))))
> > >                ;;)
> >
> > This part does not make sense:
> >
> >     (or (intern-soft (concat sym-name "-hook"))
> >         (boundp (intern-soft (boncat sym-name "-hook"))))
> >
> > In general, nothing that looks like this can make sense,
> >
> >     (or FOO (boundp FOO))
> >
> > because `boundp' will only ever be called if FOO is nil.
>
> Thamnks. Sorry, I just added the "fundamental-mode" line to make it more 
> clear when I posted. I  missed to put and "add" around the following two 
> lines.
>
> It should be
>                ;;(or (eq value 'fundamental-mode)
>                ;; (and
>                    (intern-soft (concat sym-name "-hook"))
>                    (boundp (intern-soft (concat sym-name "-hook"))))
>                ;;))
>
> Both the "fundamental-line" and the "and" line must be uncommented to 
> get it working.


I looked a little bit further. The error (which a bit surprisingly is 
"End of file during parsing") happens in `widget-apply' which is called 
from `widget-insert-button-action'.

`widget-apply' internally calls Fapply, which is `apply'.

The message buffer contains

  Debug on Error enabled globally
  End of file during parsing
  Back to top level.

There is no problem just calling a function which gets an error during 
apply, something like this

 (defun temp-bad-div(x)
   (interactive "nNumber: ")
   (message "%s" (apply '/ 100 (list x))))

This gives the trace-back I expect if called with x=0.

`widget-apply' is very short and the crucial part looks like

  GCPRO2 (newargs[0], newargs[2]);
  result = Fapply (3, newargs);
  UNGCPRO;

However i have not much idea about what GCPRO2 does so maybe someone who 
knows that can look a bit further? Why does not the trace-back appear here?