Re: GSOC | Extending Common Lisp support

[Jamison Hope <jrh at theptrgroup dot com>]

On Jun 4, 2012, at 6:56 PM, Charles Turner wrote:

> Jamison, that patch fixed the function space lookup issues I was
> having, thanks!

Excellent!

> But I've been puzzling over a follow-up namespace
> related problem. Hopefully with all that information loaded into your
> head, you might be able to throw some clues my way :-)
>
> (defun dummy () 'top-level)
> (funcall #'dummy) ;=> top-level
> (flet ((dummy () 'shadow))
>  (funcall #'dummy)) ;=> shadow
> (funcall #'dummy) ;=> shadow
>
> Line 1 does a
>
> lexical.push: Declaration[dummy/2]
>
> and when I do a decl.setProcedureDecl(true) on the flet'ed function
> line 3, I'm fiddling with a Declaration[dummy/39], so definitely a
> different object, and it's in a new scope (a new LetExp), which gets
> popped after FLET's extent.
>
> The call to FUNCTION inside the FLET correctly looks up dummy/39, and
> outside the FLET, correctly looks up dummy/2, so how is dummy/2's
> function binding getting changed? Any clues on where I might find a
> smoking gun?

The past few days I've been doing a lot of running with --debug-print- 
final-expr,
as well as inserting a bunch of Thread.dumpStack()s to see what is  
calling what.
And of course lots of print statements for variables that might be of  
interest.

I just whipped up part of a flet implementation and see the same thing  
you do
when running interactively in the REPL or with -f, but I see correct  
behavior
when running a file without -f:

> $ cat /tmp/flet-test.lisp

> (defun funcall (fn &rest args) (display (apply fn args)) (newline))
> (defun dummy () 'top-level)
> (funcall #'dummy)
> (flet ((dummy (lambda () 'shadow)))
>   (funcall #'dummy)
> (funcall #'dummy)

(Yes, I know that's not standard flet syntax. It does call  
setProcedureDecl()
though.)

REPL test:

> $ kawa --lisp
> #|kawa:1|# (load "/tmp/flet-test.lisp")
> top-level
> shadow
> shadow

kawa -f:

> $ kawa -f /tmp/flet-test.lisp
> top-level
> shadow
> shadow

kawa without -f:

> $ kawa /tmp/flet-test.lisp
> top-level
> shadow
> top-level

Is that consistent with what you're seeing?

Alas, nothing's jumping out at me just yet, but I'll keep looking.

> > Good thing we've got Charles around to figure everything out for us  
> > now!
>
> Can I have his email address? I could use someone like that! :-)

:-)

--
Jamison Hope
The PTR Group
www.theptrgroup.com