scm_c_make_gsubr parameters

scm_c_make_gsubr parameters

[guign]

Hello all :)

I just found that guile does not like primitive procedures with a lot of
arguments.
	ERROR in scm_c_make_gsubr: too many args (11) func

The (procedure-properties) show correct arities, e.g.
	scm_c_define_gsubr(...,  9, 0, 0, ...)		(arity  9 0 #f)
	scm_c_define_gsubr(..., 10, 0, 0, ...)		(arity 10 0 #f)

For most values I get a meaningful error message:
	scm_c_define_gsubr(..., 11, 0, 0, ...)		too many args (11)
	...
	scm_c_define_gsubr(..., 15, 0, 0, ...)		too many args (15)

Interestingly the following happens at 16 (my actual case):
	scm_c_define_gsubr(..., 16, 0, 0, ...)		(arity 0 1 #f)
	scm_c_define_gsubr(..., 17, 0, 0, ...)		(arity 1 1 #f)

From 18 on everything is back to normal
	scm_c_define_gsubr(..., 18, 0, 0, ...)		too many args (18)
	...

I think the case with 16 is clear, but I neither found the limit of 10,
nor the special case of 16 documented in the guile 1.8.5 manual (the
version I'm using). So this might be something to help confused people
around section 5.8.2 "Primitive Procedures".  :)

In my case of 16 arguments I got the error "Invalid Number of Arguments"
when calling my procedure. I counted a lot today... ;)

And yes, 16 parameters is really insane, I know. I was just hacking a
little piece of test code and didn't want to look the C list functions
up. If i would have written clean code I might have never stumbled
across it :)

Have a nice time :)
	Kai.

Re: scm_c_make_gsubr parameters

[Ludovic Courtès]

Hi,

guign@mails.selgrad.org writes:

> I think the case with 16 is clear, but I neither found the limit of 10,
> nor the special case of 16 documented in the guile 1.8.5 manual (the
> version I'm using). So this might be something to help confused people
> around section 5.8.2 "Primitive Procedures".  :)

10 is the real limit, and I agree it should be documented.  Would you
like to propose a patch?

Thanks,
Ludo’.

Re: scm_c_make_gsubr parameters

[guign]

On 00:38 Sun 03 Jan     , Ludovic Courtès wrote:
> 10 is the real limit, and I agree it should be documented.  Would you
> like to propose a patch?
Yes, I'll see to it :)