feature request for MLton.
Tue, 21 Aug 2001 09:27:07 -0700 (PDT)
> One more thing---I noticed on your web page that you're interested in
> a debugger. How much effort might you have to devote to this project?
> I ask because I have just hired a postdoc to work with me on
> multi-language debugging, and we are interested in finding partners.
Well, we've just kicked around some high level debugging ideas; nothing as
encompassing as your Debugging Everywhere ideas. MLton might make an
interesting test case for multi-language debugging, certainly from the
point of view of manipulated types. Probably the biggest stumbling block
is mapping between source and object locations. Even with additional
front-end support, it's not clear how to maintain information about source
mappings through all of MLton's program transformations
(de-structure/functor-ization, monomorphisation, closure conversion ---
and that's all before "normal" transformations like inlining, dead-code
elimination, etc.). As a first pass, we were considering tying object
debugging info with the last IL that is recognizably ML-ish.
But, like I said, we've just kicked around these ideas, and there hasn't
been a big push to get debugging support into MLton. While we're
certainly willing to answer questions, I don't think that anyone's
planning on devoting a significant amount of time to adding a debugger in
the near future.