Re: scm_wrong_num_args

[Dirk Herrmann]

On 29 Mar 2001, Gary Houston wrote:

> > From: Dirk Herrmann <address@hidden>
> > Date: Sun, 25 Mar 2001 18:17:37 +0200 (MEST)
> > 
> > One aspect is (as you have suggested) to make the error reporting
> > functions create continuations to allow to continue the execution after an
> > error has occured.  This idea can be extended, for example to use the
> > continuation to extract backtrace information (as a replacement or in
> > addition to the current explicit stack creation mechanism).  This would
> > allow to show a mixture of C level and scheme level stack frames.
> 
> Restartable exceptions were actually rejected when the current system
> was evolved, see:
> 
> http://www.red-bean.com/guile/guile/old/0455.html
> 
> Creating a continuation for every exception seems too slow to be used
> in non-debugging sessions.

Thanks for the link.  Hmmm.  I think that we do not necessarily have a
contradition here:  First, we don't need to pass continuations for _every_
exception, only for errors.  Further, we can make the creation of a
continuation depend on whether guile is run in debug mode or not.

I don't see the aspect of 'restartability' as the main reason to create a
continuation in case of an error, but rather the aspect that for debugging
a continuation provides all information about the stack you would like to
have (well, not really _all_ information: in case of tail calls you have
to add a little bit extra data).

The major concern that I have about this issue is, that adding such a
feature (i. e. the possibility to scan a C level stack and separate it
into several stack frames) will be quite complicated and require us to
include a lot of platform dependent code.  The best solution would be if
some of gdb's code was available as a library, but it seems that such a
project (although once planned) was abandoned.

Best regards,
Dirk