Re: [Javapathfinder-devel] Junit test failures?

[Peter C. M. <pcm...@em...>]

Ok, the LDC_W Java 1.5 issue should be resolved, here's my comment

    /*
      * Java 1.5 silently introduced a class file change - LDCs can now 
directly reference class
      * constpool entries. To make it more interesting, BCEL 5.1 chokes 
on this with a hard exception.
      * As of Aug 2004, this was fixed in the BCEL Subversion 
repository, but there is no new
      * release out yet. In order to compile this code with BCEL 5.1, we 
can't even use Type.CLASS.
      * The current hack should compile with both BCEL 5.1 and svn, but 
only runs - when encountering
      * a Java 1.5 class file - if the BCEL svn jar is used
      */

good thing I used the BCEL Subversion repository for our jpf-lib.zip 
snapshot.

But there are remaining Java 1.5 runtime issues: (1) TestAssert still 
hangs (interesting, it doesn't do much else than just raising an 
AssertionError - I guess that also has to do with class const 
handling), (2) some of the threading tests fail for apparently the same 
cause, which I assume is a change in one of the system classes (Thread, 
ThreadGroup etc.) we are depending on. Both need direct access to a box 
running Java 1.5, so it has to wait until next week.

-- Peter



On Apr 28, 2005, at 12:35 PM, Jean-Francois Halleux wrote:
> BCEL 5.1, "standard" binary distribution.
>
> JF
> -----Original Message-----
> From: Peter C. Mehlitz [mailto:pcm...@em...]
> Sent: jeudi 28 avril 2005 21:09
> To: <jav...@li...>
> Cc: hal...@sk...
> Subject: Re: [Javapathfinder-devel] Junit test failures?
>
>
> Thanks. Yes, it's a known BCEL / Java 1.5 problem caused by a changed
> LDC_W bytecode spec (LDCs can now directly reference class constants):
>
> java.lang.RuntimeException: Unknown or invalid constant type at 99
> 	at org.apache.bcel.generic.LDC.getType(LDC.java:148)
> 	at gov.nasa.jpf.jvm.bytecode.LDC_W.setPeer(LDC_W.java:47)
> ...
>
> strange thing is that it should be fixed in the current BCEL version.
> Which one do you use?
>
> -- Peter
>
>