[racket] Understanding error in use of submodule in my language

Thanks, Matthew.

I fixed it by not expanding the module* form in the local-expand,
i.e. I used (list #'module*) as the third argument to local-expand.

Cheers,
Marco


On Fri, Aug 22, 2014 at 4:38 PM, Matthew Flatt <mflatt at cs.utah.edu> wrote:

> I think this may be a problem in macro expansion with submodules.
>
> The `#%module-begin` from `racket/base` adds a submodule declaration
> for `configure-runtime` when it doesn't see an existing submodule
> declaration for `configure-runtime` --- but it looks for an existing
> declaration as an immediate one, before macro-expanding the module
> body.
>
> In your code, the `mute` wrapper hides the declaration of
> `configure-runtime`, so a conflicting one gets added. The tricky part
> is that this happens the second time that the `test` submodule is
> expanded:
>
>  * The overall module's body is forced via `local-expand`.
>
>  * As a part of that local expansion, the `test` submodule is expanded'
>    its expansion includes a declaration of a `configure-runtime`
>    submodule.
>
>  * The overall module's expanded body is wrapped back in
>    `#%module-begin`, which will re-expand all of the content, including
>    the `test` submodule.
>
>  * The `test` submodule this second time around has a `#%module-begin`
>    that is supposed to be bound to `#%plain-module-begin`, but it ends
>    up being captured[*] by `#%module-begin` from "bootstrap-lang.rkt",
>    so that the submodule's body is again local-expanded and again send
>    to `#%module-begin` from `racket/base` with a `mute` wrapper around
>    the `configure-runtime` sub-submodule.
>
> The [*] step seems wrong to me. I think this may be an example of a
> problem deep in the macro expander, and solving that is my next
> project.
>
>
> Since the repair (assuming that I have correctly diagnosed the cause)
> may be a while, do you have a workaround already? The macro name `mute`
> suggests that you may want to suppress the printing of expression
> results at the top-level of a module, and that might be as easy as
> expanding `#%module-begin` to `#%plain-module-begin`. Or maybe
> `syntax/wrap-modbeg` can help, or maybe the strategy used by
> `syntax/wrap-modbeg` would work better for your macro.
>
>
> At Thu, 21 Aug 2014 17:22:29 +0100, Marco Monteiro wrote:
> > Hello!
> >
> > I'm using Racket 6.1
> >
> > I'm trying to understand this error. I'm using submodules in a language I
> > defined. The language file is here:
> >
> > http://pastebin.com/7RiHHRmi
> >
> > Notice the commented out line 28.
> >
> > I use that language in the following code
> >
> > http://pastebin.com/Qa6RD9wF
> >
> > raco expanding this last file fails with
> >
> > module: duplicate submodule definition
> >   at: configure-runtime
> >   in: module
> >   context...:
> >
> /usr/share/racket/pkgs/compiler-lib/compiler/commands/expand.rkt:29:11:
> > loop
> >    /usr/share/racket/pkgs/compiler-lib/compiler/commands/expand.rkt:
> > [running body]
> >    /usr/share/racket/collects/raco/raco.rkt: [running body]
> >    /usr/share/racket/collects/raco/main.rkt: [running body]
> >
> > If I uncomment line 28 it no longer fails. If I use (module ...) instead
> of
> > (module* ...) it no longer fails independently of line 28 being commented
> > out or not.
> >
> > Trying to macro expand that in drracket gives me "Internal error: car:
> > contract violation ." error.
> >
> > What is happending here?
> >
> > Thanks,
> > Marco
> > ____________________
> >   Racket Users list:
> >   http://lists.racket-lang.org/users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.racket-lang.org/users/archive/attachments/20140905/97a3894f/attachment-0001.html>