This is the mail archive of the
kawa@sources.redhat.com
mailing list for the Kawa project.
Re: VerifyError with kawa-1.6.70
- To: "Brian D. Carlstrom" <bdc at ai dot mit dot edu>
- Subject: Re: VerifyError with kawa-1.6.70
- From: Per Bothner <per at bothner dot com>
- Date: 23 Oct 2000 22:06:43 -0700
- Cc: kawa at sourceware dot cygnus dot com
- References: <E13npjU-0003j5-00@nestle.ai.mit.edu>
"Brian D. Carlstrom" <bdc@zurich.ai.mit.edu> writes:
> I'm trying to benchmark Kawa with the qobi Scheme version of the Gabriel
> benchmarks, similar to:
>
> http://sources.redhat.com/kawa/papers/KawaLisp98-html/x1005.html
>
> I just ran into this VerifyError although everything else if fine.
>
> Any ideas? A compiler bug or a picky verifier? any easy way to work
> around this in the later case.
It's a compiler bug. I just checked in the following patch:
2000-10-23 Per Bothner <per@bothner.com>
* StackTarget.java (compileFromStack0): Always call emitCoerceToObject
even when stackType.isSubtype(type). Fixes bug in Stalin's
destruct.sc benchmark, reported by Brian D. Carlstrom <bdc@ai.mit.edu>.
Index: StackTarget.java
===================================================================
RCS file: /cvs/kawa/kawa/gnu/expr/StackTarget.java,v
retrieving revision 1.5
diff -u -r1.5 StackTarget.java
--- StackTarget.java 1999/08/07 19:20:26 1.5
+++ StackTarget.java 2000/10/24 03:01:46
@@ -34,10 +34,8 @@
return true;
}
- if (stackType.isSubtype(type))
- return true;
stackType.emitCoerceToObject(code);
- return false;
+ return stackType.isSubtype(type);
}
public void compileFromStack(Compilation comp, Type stackType)
--
--Per Bothner
per@bothner.com http://www.bothner.com/~per/