freenode/#clasp - IRC Chatlog
Search
21:16:11
Bike
it's just that because of the way we build we have to worry about it trying to evaluate that ensure constructor cell stuff before static-gfs is actually loaded.
21:18:42
drmeister
Ok, but this is happening in Acclimation and Concrete-Syntax-Tree - AFTER static-gfs is loaded. So it should be using the (load-time-value (ensure-constructor-cell ...)) expansion at after static-gfs are loaded and it should work - right?
21:20:38
drmeister
I didn't think I was looking at the right problem - now I know that there type error is related to the real problem and not some mis-compilation.
21:23:56
drmeister
I'm playing with udb and adding my common lisp object inspection extension to it.
21:24:39
Bike
for context, the idea here is to avoid all the overhead of calling the make-instance, initialize-instance, allocate-instance generic functions and checking keywords and stuff, by computing a function tailored to deal with the particular class and keywords at the call site
21:25:13
Bike
this function can't be permanent since the class can be redefined and stuff, so load-time is about as early as we can compute it
21:25:59
drmeister
Understood - and you intended that the compilation of these optimized functions happen at load time - but only after static-gfs is fully loaded.
21:26:47
drmeister
Ok, so I'll remove the changes I made to Acclimation and Concrete-Syntax-Tree and let them crash the way that they were and try to get some info on why they are crashing.
21:28:52
Bike
if we don't get this figured out by monday i'll probably disable it locally so i can focus on bignums
0:45:49
drmeister
This is frustrating. I've built a version of clasp that should fail with this problem but it does not.