Subject: Re: Error: shouldn't still have this class as a context
From: Craig Chambers (chambers@cs.washington.edu)
Date: Tue Nov 07 2000 - 10:31:53 PST
The type checking/inference for unary and binary operators is pretty fragile.
It's tough to resolve all the issues in computing what the right representations
should be for arguments of operators like + and =, and to compute the result
representation from them, given all the heavy overloading of different numeric
sizes, formats, and signedness, and the attempt to use enclosing context to help
resolve the overloading. I've gotten it to work on the cases I've tested, but
you could easily have hit a case that still doesn't work right. Or you might
have introduced some sort of bug. Let me know when I can come by and browse
your stack to help figure out what's going on.
-- Craig
Sorin Lerner wrote:
>
> After running my constant folding phase, I hit the error on line 245 in
> rep-classes.cecil. Can anybody tell me what that error means? Here's
> part of the stack trace:
>
> # 7 check_is_subtype (node: <anon/BinaryOpNode/: 0xd05101>,
> lhs: <anon/ConIntRep/: 0x10220d1>,
> rhs: FromContextRepClass,
> error_msg: <anon/i_vstring/: "second operand">),
> rep-check.cecil:992
> # 6 check_is_subtype (node: <anon/BinaryOpNode/: 0xd05101>,
> lhs: <anon/ConIntRep/: 0x10220d1>,
> rhs: FromContextRepClass,
> error_msg: <anon/closure: 0xefff9129>),
> rep-check.cecil:997
> # 5 is_subtype (lhs: <anon/ConIntRep/: 0x10220d1>,
> rhs: FromContextRepClass), rep-classes.cecil:40
> # 3 eval (_anon_closure: <closure>) in is_subtype, rep-classes.cecil:40
> # 2 is_subtype_core_rhs (rep: <anon/ConIntRep/: 0x10220d1>,
> rc: FromContextRepClass), rep-classes.cecil:245
>
> # 1 error (msg: <anon/i_vstring/: "shouldn\'t still have this class as a
> context">), general.cecil:123
> # 0 exit (error_code: -1), system.cecil:7
This archive was generated by hypermail 2b25 : Tue Nov 07 2000 - 10:31:59 PST