[kaffe] Re: Planning for 1.1.3 release

Dalibor Topic robilad at kaffe.org
Mon Dec 1 08:40:03 PST 2003

Hi Hakon,

Hakon Gunsen wrote:
> Hi!
> I don't know if this is a blocker, but I'm unable to compile Kaffe - both 1.1.2 and cvs, on the new Fedora Core distribution.
> There are some differences, but when compiling cvs it breaks in the make process with this message:
> /bin/sh ./rebuildLib @essential.files
> Compiling classes from  @essential.files  using  /home/hakon/kaffe/kaffe/kaffe/kaffe-bin -verbosegc -mx 256M at.dms.kjc.Main
> ./rebuildLib: line 58: 21577 Segmentation fault      $JAVAC $VERBOSE $JAVAC_FLAGS -d $LIBDIR $CPATH ${1+"$@"}
> make[1]: *** [lib/stamp] Error 139
> make[1]: Leaving directory `/home/hakon/kaffe/libraries/javalib'
> make: *** [all-recursive] Error 1
> Does anyone else have a Fedora to test with?

I've played with Fedora Core 1 a little bit, and I think it's not as 
good as similar efforts at community managed linux distibutions (debian, 
gentoo). But anyway, I've tried to get kaffe to run in the beast, since 
it's got the new shiny NPTL enabled kernel, and other proprietary VMs 
are known to have had some problems with it.

Short story is that kaffe has its problems as well. The crash happens 
quite early, when processClass is calles on String after the 
initialisation of base types. It crashes in sysdepCallMethod for 
<clinit>. I don't have the slightest idea why it does that, I assume 
that method->ncode is a bad address in this particlar case. I haven't 
yet found out where that one comes from. I'm rewriting the macros (ugh, 
macros!) from classMethod.h to inline methods to get somewhat useable 
support from the debugger.

Fedora's gcc seems to miscompile kaffe at -O2, so that the stack trace 
is garbled. Building kaffe with CFLAGS=-O0 -g gives a useable stack 
trace, but as I said, I'm currently at loss where the error might come 
from. Compiler warnings don't seem to offer much of a guidance here, 

I'll check in what I've got refactored as inline functions over the 
week, so that others can chip in.

dalibor topic

More information about the kaffe mailing list