JDK 1.6.0 patchset 1 "Brumby"

[Jung-uk Kim]

On Friday 10 August 2007 08:50 am, Matt wrote:
> On 8/9/07, Rene Ladan <r.c.ladan at gmail.com> wrote:
> > Jung-uk Kim schreef:
> > > On Sunday 05 August 2007 11:34 am, Rene Ladan wrote:
> > >> I tried installing this on my i386 FreeBSD-CURRENT SMP box
> > >> (2007-08-04 06:32 UTC) using diablo-jdk15 as bootstrap.  The
> > >> build first failed because of GCC 4.2 warnings-as-errors about
> > >> deprecated string conversions.  After commenting out a -Werror
> > >> in work/hotspot/build/bsd/makefiles/gcc.make , the
> > >> hotspot-fastdebug server built fine, but it failed in
> > >> test_gamma.  I've attached the hs_err log.
> > >
> > > The attached patch should fix the build problem on -CURRENT.
> >
> > Got it.
> >
> > > Your error log shows that you set JAVA_HOME wrong:
> > >
> > > Environment Variables:
> > > JAVA_HOME=/usr/local/diablo-jdk1.5.0
> >
> > Building with diablo-jdk15 is supported?
> >
> > After moving the patch to files/ and removing work/ , rebuilding
> > the port now lets the gamma tests succeed.  But now it fails with
> > an AssertionError when zipping up rt.jar.  The build log
> > including some additional information is available at
> > http://home.tiscali.nl/rladan/freebsd/AssertionError.bz2 (35kB).
>
> See suggestion noted in
>
> http://lists.freebsd.org/pipermail/freebsd-java/2007-August/006579.
>html
>
> The optimization reduction noted there fixed my build issues on
> current for both jdk16 and jdk15.

It is very strange but I can believe that.  GCC 4.2.x has many 
optimization bugs.  Some problems were fixed in 4.2.1 but there are 
some serious regressions like these:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32328
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32716
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32992

Complex, C++ intensive, and optimization flag sensitive applications, 
e.g., JDK and OpenOffice.org, may be affected by them without being 
noticed. :-(

Jung-uk Kim