jikes 1.17 bug (Was: Re: [kaffe] Some patches to use jikes)

[Hervé Roussain]

>
>
>
>With jikes 1.17, everything compiles, and all but one
>regression tests pass on i386-linux-jit3. The failing
>test is SoTimeout, which fails because kaffe can't
>find SocketImpl.setOption(int, Object) method. I have
>to take a closer look at the disassembly of the source
>files to figure out if that's a bug in kaffe or in
>jikes 1.17.
>
>keep up the good work,
>
>dalibor topic
>

It seems to be a jikes bug. But it is also a gcj bug (after a comment in 
SocketImpl.java & DatagramSocketImpl.java in classpath's source) that 
classpath solve by redeclarations of "setOption" and "getOption" methods 
in the 2 mentioned classes.
If you decide to do the same thing than in GNU Classpath, don't forget 
the DatagramSocketImpl (it has the same problem but this is not shown 
with regression tests).

As the problem may be reproduced in other cases, I join a file that can 
be put in test/regression, or at least that shows you the bug. It works 
with kjc, but not with jikes 1.17 (it compiles but throw a 
"NoSubMethodError" exception. I called it "AbstractInterfaceTest" 
because there is a problem with jikes when an abstract class implements 
an interface. It may be renamed or thrown out if you think it's bad idea 
(that is NOT a kaffe bug)

Best regards,
Hervé

PS: I encounter a "bug" when I post something to kaffe's mailing list : 
I receive my post 5 hours after I posted it. I believe this is because I 
live in France and there is a 5 hours shift between Eastern USA & 
Western Europe. How could this be "repaired" ?
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: AbstractInterfaceTest.java
Url: http://kaffe.org/pipermail/kaffe/attachments/20021206/f5785e0b/attachment-0003.txt