Subject: publishing
From: Vassily Litvinov (vass@cs.washington.edu)
Date: Mon Mar 05 2001 - 20:22:24 PST
Publishing changes to the Vortex typechecker:
- fixed the problem when the "dynamic" type was not interpreted as
"any type you wish"
- enabled compression of signature lists during implementation-side
typechecking
- miscellaneous cleanup changes
- removed "debug" pragmas from my files in the checked-in version
This typechecker typechecks both Vortex and Whirlwind with no type errors.
I am compiling the fully-optimized version of the compiler executable
so it can be installed in the central places.
CAVEATS
(1) when "dynamic" type is used, an overly allowing policy is
currently implemented. Namely, during constraint solving process, if
a fresh type variable is ever unified with "dynamic," it will unify
with anything else after that. For example, when unifying constraint
"signature foo(T,T,T):void" with "signature foo(dynamic,int,string):void",
unification will succeed. It arguably shouldn't succeed as no value
for "T" can match both "int" and "string", even though any value for
"T" will match "dynamic".
I chose this implementation because it is simpler. In order to
support the more precise policy, sofisticated code is needed, leading
to high chance of introducing bugs. If problems are encountered, I'll
look into it again.
(2) apparently there is a bug in my signature-list-compression code,
as it eliminates too many signatures. It is therefore disabled for
client-side typechecking. It is still enabled for implementation-
side typechecking, as the dramatic benefit of speading it up was
preferred over checking a couple of particularly weird signatures
(that's where the bug shows up). This is temporary and I'll debug the
code when I get a chance. Rest assured that plenty of significant ITC
errors to worry about are still detected.
Vass
_______________________________________________
Cecil mailing list
Cecil@cs.washington.edu
http://majordomo.cs.washington.edu/mailman/listinfo/cecil
This archive was generated by hypermail 2b25 : Mon Mar 05 2001 - 20:23:04 PST