RE: [Ikvm-developers] A few fixes and a call for participation

[Jeroen F. <je...@su...>]

Hi Brian,

I'm glad to see the verifier doing its job :-)

I replied to the list so that others can see the resolution of this
issue as well, I hope you don't mind.

For the ClassCircularityError, I don't necessarily need a small repro,
if you don't mind sending me the whole code I can easily work from that
(if you include instructions on how to run it).

Regards,
Jeroen

> -----Original Message-----
> From: Brian Sullivan [mailto:bs...@le...]=20
> Sent: Friday, August 08, 2003 03:28
> To: Jeroen Frijters
> Subject: Re: [Ikvm-developers] A few fixes and a call for=20
> participation
>=20
> Jeroen Frijters wrote:
> > Hi Brian,=20
> >=20
> > Thanks for the feedback!
> >=20
> >>3) We're getting illegal targets message, about 20 of them=20
> in several=20
> >>programs.
> >>
> >>VerifyError: Illegal target of jump or branch
> >>  at MethodAnalyzer.ctor(TypeWrapper wrapper, Code method,=20
> >>ClassLoaderWrapper classLoader)
> >>java.lang.VerifyError: (class: nc231a, method:=20
> >>check_entries_OF_sect_nc231a_001, signature: ()I, offset: 149,=20
> >>instruction: goto) Illegal target of jump or branch
> >=20
> > Would you please send me one of the classes that triggers this,
> > including the corresponding exception?
>=20
> Thanks for helping me get this going.
>=20
> I wanted to reproduce this in straight Java if I could, so I kept on=20
> simplifying it.  Then, making sure, I ran the exact same code with=20
> straight Java.  It _too_ gave a VerifyError.
>=20
> Scratching my head, as this code always works, I realized that I=20
> compiled with Eclipse to run in IKVM, whereas these tests are=20
> normally=20
> compiled with jikes from the command line or javac on some systems.
>=20
> This one turned out to be an Eclipse compiler bug, nothing to do with=20
> IKVM or Classpath, and upgrading Eclipse to 2.1.1 (from=20
> 2.1.0) did fix=20
> the problem.
>=20
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=3D37621
>=20
> I'll do the simplifications of the others down.  I did a bit on the=20
> ClassCircularityError, but hadn't gotten anywhere yet on a simpler=20
> reproduction.
>=20
> If curious, the VerifyError case got down to:
>=20
> for(;;)
> {
> 	if(booleanvar)
> 	{
> 		if(true) return 1;
> 		break;	// comment this out to work properly
> 	}
> 	else
> 	{
> 		if(otherboolean) break;
> 	}
> }
>=20
> Thanks,
> Brian
>=20
>=20