bug#24286: Guile-2.2: srfi-1 + #:duplicates (merge-generics...) + for-each raises an exception

[Andy Wingo <wingo@pobox.com>]

On Mon 22 Aug 2016 03:03, David Pirotte <david@altosw.be> writes:

> Hello Guilers,
> Hi Andy,
>
> Here below a mini module [1], which raises an exception [2] using 2.1.3.104-8f2f8 (it
> compiles fine using 2.0.12).  It looks a bit of a 'silly' module but the real one
> here do need this combination, I just minimized it for tracking purposes...
>
> For info, if I either comment the use of srfi-1, or merge-generics, _or_ the for-each
> expression, it compiles fine. I also tried foo to just do a (filter-map string?
> words) and no for-each, to circumvent the problem as far as I could, and it
> compiles fie as well: it appears there is a side effect of the for-each version from
> srfi-1 over the core version, when using goops and declaring #:duplicates
> (merge-generics ...).
>
> Cheers,
> David
>
> [1]
>
> (define-module (foo)
>   #:use-module (srfi srfi-1)
>   #:use-module (oop goops)
>   
>   #:duplicates (merge-generics
> 		replace
> 		warn-override-core
> 		warn
> 		last)

So here you are importing two definitions of for-each: one from (guile)
and one from (srfi srfi-1).  They are both normal procedures.  What
should happen?  I guess merge-generics should fail to do anything.  I
suppose that is the intention of this method in (oop goops):

  (define-method (merge-generics (module <module>)
                                 (name <symbol>)
                                 (int1 <module>)
                                 (val1 <top>)
                                 (int2 <module>)
                                 (val2 <top>)
                                 (var <top>)
                                 (val <top>))
    #f)

And yet you get:

  ERROR: No applicable method for #<<generic> merge-generics (3)> in call (merge-generics #<directory (foo) 204c480> for-each #<interface (guile) 20f1ea0> #<procedure for-each (f l) | (f l1 l2) | (f l1 . rest)> #<interface (srfi srfi-1) 21322d0> #<procedure for-each (f l) | (f l1 l2) | (f l1 . rest)> #f #f)

This sounds like a bug to me, surely the method should match.

On the other hand this looks bogus to me:

  scheme@(foo)> (class-of (resolve-interface '(guile)))
  $4 = #<<class> <module> 244bbd0>
  scheme@(foo)> (class-of (resolve-interface '(guile)))
  $5 = #<<class> <module> 244b900>

Why are we getting two distinct values for the class?  Further
investigation necessary.

Andy