Re: [Chicken-users] How to avoid predefined cond-expand features in csc?

[Peter Bex]

On Mon, Dec 28, 2015 at 09:32:24AM +0100, Sven Hartrumpf wrote:
> PB wrote on 2015-12-24 11:25:
> > These features are a bit weird: when you require a core library, it will
> > register a feature with the same name (I think that's to prevent it from
> > being reloaded again).  I'm not sure if this is in general for units or
> > only for core units, but I think the problem here is that the compiler
> > loads all of them.  So you're really cond-expanding against what's
> > been loaded by the compiler, which is of course undocumented and subject
> > to change.
> 
> And hopefully is changed ... in chicken 5? :-)

There are plans to overhaul the whole unit/module/library loading stuff,
but development on CHICKEN 5 seems to have stalled a bit.  We need more
people to work on it, even if just to flesh out the desired API.

> >> Example:
> >>  > echo "(cond-expand (srfi-1 (print "srfi-1")) (else))" > bugchicken2.scm
> >>  > csi -e '(load "bugchicken2.scm")'
> >>  >            -> as expected
> >>  > csc bugchicken2.scm && ./bugchicken2
> >> srfi-1        -> unexpected; csc should behave like csi.
> > 
> > If you don't mind doing this at runtime, that's a solution:
> > 
> > (when (feature? srfi-1:) (print "srfi-1"))
> 
> Thanks for the idea.
> But I need cond-expand mainly to write programs that are portable
> across different Scheme implementation. So I cannot replace the
> most basic SRFI-0 by some implementation-specific variant.

You could still do something like

(cond-expand
  (chicken (when (feature? srfi-1:) (print "srfi-1")))
  (gauche ...)
  (chibi ...)
  (else ...))

It's not pretty but it should do the trick.  You could even write a
macro that expands to the desired cond-expand form, but that's a bit
tricky to do if you can't rely on syntax-rules being present.

Cheers,
Peter

signature.asc
Description: Digital signature