Re: Patch for better multiple values support

[Martin Grabmueller]

> From: Matthias Koeppe <address@hidden>
> Date: 10 May 2001 11:24:16 +0200
> 
> This patch does NOT make Guile signal an error when multiple values
> are passed to ordinary (single-value) continuations.  I don't know how
> to do this; probably the evaluator needs to be changed.  Maybe someone
> with a deeper understanding of the evaluator can help here.

I don't think Guile should test for this, at least not in the normal
evaluator.  In the debugging evaluator, it might be an advantage or
not, I'm not sure.

Properly supporting this would require a check for _all_ evaluated
subexpressions, and if they evaluate to multiple values and we are not
inside a call-with-values, an error must be signalled.  This would slow
down the evaluator dramatically.

As a compromise, we could insert check in some strategical chosen
places, such as in define or set!, which are not so critical
performance-wise.

What do others think?

Regards,
  'martin