Re: libgcj/1516: Method.invoke won't accept boolean parameters

[Bryce McKinlay <bryce at albatross dot co dot nz>]

The following reply was made to PR libgcj/1516; it has been noted by GNATS.

From: Bryce McKinlay <bryce@albatross.co.nz>
To: jeff.sturm@commerceone.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: libgcj/1516: Method.invoke won't accept boolean parameters
Date: Thu, 28 Dec 2000 17:43:10 +1300

 jeff.sturm@commerceone.com wrote:
 
 > >Description:
 > Method.invoke throws java.lang.IllegalArgumentException for
 > methods with boolean (primitive) types in their signatures.
 > >How-To-Repeat:
 > Run the attached test case.  It should print `true' and
 > `false', not throw an exception.
 > >Fix:
 > Add boolean rules to can_widen.  Alternatively, rewrite
 > can_widen to use an array of bitmaps or similar, since the
 > existing rules are not too pretty or efficient.
 
 I've checked in a fix for this:
 
 http://sources.redhat.com/ml/java-patches/2000-q4/msg00284.html
 
 Another option for efficiency improvement here might be to combine the ffi_type
 lookup with the primitives conversion table. However, right now I'd say that
 micro-optimizing the _Jv_CallAnyMethod stuff is a misdirected effort, I'll bet
 that by far the slowest part of invoke() is the _Jv_LookupDeclaredMethod()
 call, which must be done to check for overriding. Per's plan for speeding up
 _Jv_LookupDeclaredMethod involves global String interning.
 
 regards
 
   [ bryce ]