Re: Tracking performance regressions in 1.1.7-pre

On Friday 15 October 2004 16:23, Grzegorz B. Prokopski wrote:
> On Fri, 2004-10-15 at 08:25, Laurent Martelli wrote:
> > >>>>> "Grzegorz" == Grzegorz B Prokopski <[email protected]>
> > >>>>> writes:
> >
> >   Grzegorz> On Thu, 2004-10-14 at 12:53, Laurent Martelli wrote:
> >   >> Revision 2933 is OK but 2967 is an aging turtle.
> >
> >   Grzegorz> Ok, that's something.  Now, before we get into
> > reviewing Grzegorz> 3.5MB(!) diff can you try sablevm-classpath
> > upated to Grzegorz> rev. 2934?  This is where the merge of GNU CP
> > CVS was Grzegorz> done.
> >
> > Revision 2934 of classpath does not compile:
> >
> > laurent@stan:~/src/sablevm-classpath-staging+2934$ make
> > Making all in lib
> > make[1]: Entering directory
> > `/home/laurent/src/sablevm-classpath-staging+2934/lib'
> > top_builddir=.. /bin/bash ./gen-classlist.sh standard
> > /usr/bin/gcj --bootclasspath '' --classpath ..:../vm/current:.:
> > -C -d . @classes
>
> ^^^^^^^^^^^^^^
>
> > ../java/lang/Class.java:1193: error: Duplicate variable
> > declaration: `String name' was `<unknown> name' (<built-in>:0).
> > private String name;
>
> You have to use jikes.  GCJ has never been guaranteed to work.
> There's a configure switch to force which compiler to use.
> If it didn't work comment out GCJ detection from acinclude.m4:
>
> dnl  CLASSPATH_WITH_GCJ
>   CLASSPATH_WITH_JIKES
> dnl  CLASSPATH_WITH_KJC

GCJ 3.3.x and up work nicely. It just needs damn much resources, much 
more then jikes. Configuring with "--with-jikes" is enough when jikes 
is available in PATH. There should be no need to edit any file for 
this.


Michael