Older blog entries for tbmoore (starting at number 11)

CMU Common Lisp foreign linkage 2, running on
ariel.bricoworks.com
Send questions and bug reports to your local CMU CL
maintainer,
or to cmucl-help@cons.org. and cmucl-imp@cons.org.
respectively.
Loaded subsystems:
    Python 1.0, target Intel x86
    CLOS based on PCL version:  September 16 92 PCL (f)
*

This means that the guts of my shared library changes to CMUCL work; the system has successfully recompiled itself. Yay! That's been a long time coming. I've had to unravel mysterious parts of CMUCL as well as fix my own silly bugs.

My new code can't actually load shared libraries yet, but the hardest part, including low-level compiler changes, is done.

I struggled for a good part of the week with rebuilding CMUCL with my invasive changes to better support shared libraries. I took a pretty big diversion into fixing defmacro, whose compile-time/load-time semantics were broken. Now it's time to start all over again on the original task!

Extended conses aren't going to happen. A two word data structure that's basic to the language would be transformed into a multiword structure... besides, compilers and codewalkers and such quickly transform the list-based source into a structure-based representation. If you really want properties associated with cons cells, you can do it with a hash table.

Made the first net release of cparse, a Common Lisp program for parsing C header files and generating FFI definitions. Lucky Advogato readers get first dibs. I need to make an entry for it at Cliki. Now, however, I'm going to sleep.

I've been doing lots of good hacking on cparse, a Lisp program for generating foreign function interface declarations from C header files. In my last entry I breezily stated that handling sizeof() would be easy. Hah!! Anyway, getting that all right turned up parts of the C grammer that I had "forgotten" to handle. So, after much ripping apart and putting things back together, the parser is in much better shape. It now processes all of Xlib.h except for stdarg "..." function arguments, which aren't supported in CMUCL anyway. Things really are looking good for a public release later this week.

Celebrated my 35th birthday yesterday. When I was a teenager I could never imagine being so old.

Hacked some more on cparse, my Lisp program for parsing C headers and generating FFI declarations. I've been feeding it Xlib.h, which is a good, complicated test case. Turns out I forgot to parse sizeof(). Doh! It won't be hard to add as I already have the support in the C types code to figure out the size of everything.

Saw the best, funniest special on VH1 ever last night: a biography of Iggy Pop. Usually those things are so pompous that they suck incredibly, but in this case it was impossible for VH1 to overwhelm the wit and genius of Iggy himself. The ad for the rap compilation "Monster Booty" was an added bonus.

I've set the garbage collector work aside for a week and have been hacking on another home project. I've written a C header file parser for Common Lisp. This code digests header files that have been run through the C preprocessor and parses the subset of C that is likely to be in a header file i.e., most things other than function bodies. Constant expressions are supported. A parse tree, with each element of C grammer represented as a CLOS object, is produced. I've also written some code that walks that parse tree and emits the def-alien declarations for CMU CommonLisp. So, maybe we're a little bit closer to automatically interfacing with C? Maybe. I hope to make an initial release by the end of the week.

I'd like to get all the ports of CMUCL to use this collector; there's no advantage of the old stop/copy collector that I can see, other than simplicity and possible lack of support for write protecting pages on some platforms.