Re: [PATCH] Fix for `submodules' in (ice-9 session) (closes #30062)

[Jose A. Ortega Ruiz]

On Tue, Aug 31 2010, Andy Wingo wrote:

> Hi,
>
> Can you submit a test please, also?

I'm not quite sure what you want the test to be: the procedure
`submodules' is not exported by `(ice-9 session)', and there's no test
suite for the latter, so testing this change properly would mean
creating such a test suite, checking all directly affected exported
procedures. Those are `apropos-fold-exported' and
`apropos-fold-accessible'.

Testing them is a bit involved, due to the cycles in the module tree
that we have discussed in the past, and that imply that any fold that
does not keep track of visited modules and escapes somehow upon
detecting a cycle, will enter an infinite loop. Even when escaping,
one's not guaranteed to visit all modules. IIRC, those cycles cannot be
eliminated for backwards compatibility reasons.

IMHO, given the cycles, the apropos-fold* functions are broken (one
cannot fold over cyclic structures), and maybe should be fixed instead
of applying only this partial fix. A possible solution is making
apropos-fold-* work on a real DAG, by prunning cycles during the
traversal.

IIRC, the modules causing cycles where among those having gensym's as
names: if they were catchable by a predicate somehow, and easy solution
would be to filter them out in `submodules' -- if for some reason a user
needs a gensym module (i don't know what they are or whether they're
part of the public module interface), she can access it via
`module-submodules'.

jao