I agree, it seems like a bug. Here is a slightly more minimal example:
```
$ cat rectangular.scm
(library (rectangular)
(export real-part)
(import (guile))
(define real-part car))
$ cat solution.scm
(use-modules ((rectangular) #:prefix rect:))
(display (rect:real-part '(1 . 2)))
$ guile3.0 --fresh-auto-compile -L . solution.scm
[...]
ERROR: In procedure real-part:
In procedure real-part: Wrong type argument in position 1: (1 . 2)
```
On the other hand, this doesn't happen if you replace rectangular.scm with
```
(define-module (rectangular)
#:export (real-part))
(define real-part car)
```
I think the implementation of the R6RS library form is to blame here: look at
this.
```
scheme@(guile-user)> ,expand (library (rectangular) (export real-part) (import
(guile)) (define real-part car))
$1 = (begin
(let ((m ((@@ (guile) define-module*)
'(rectangular)
#:filename
#f
#:pure
#t
#:version
'()
#:declarative?
#t)))
((@@ (guile) set-current-module) m)
m)
(let ((iface ((@@ (guile) resolve-r6rs-interface) '(guile))))
((@@ (guile) call-with-deferred-observers)
(lambda ()
((@@ (guile) module-use-interfaces!)
((@@ (guile) current-module))
((@@ (guile) list) iface)))))
(if #f #f)
((@@ (guile) call-with-deferred-observers)
(lambda ()
((@@ (guile) module-export!)
((@@ (guile) current-module))
'())))
((@@ (guile) call-with-deferred-observers)
(lambda ()
((@@ (guile) module-re-export!)
((@@ (guile) current-module))
'(real-part))))
((@@ (guile) call-with-deferred-observers)
(lambda ()
((@@ (guile) module-replace!)
((@@ (guile) current-module))
'())))
(define real-part car))
```
As you can see, the library macro has turned the (export real-part) part into a
re-export from (guile).
signature.asc
Description: This is a digitally signed message part
