Release stuff

Thu Mar 14 12:35:00 GMT 2002

Rainer Orth writes:
> ok: I'm now running bootstraps off the 3.0 and 3.1 branches for Tru64 UNIX
> V4.0F and V5.1 and will submit testsuite results and a comparison when I'm
> done.

I've now run a couple of bootstraps and the results are somewhat
depressing:
alpha-dec-osf4.0f	CONFIG_SHELL=/usr/local/bin/bash
			--disable-nls --enable-libgcj alpha-dec-osf4.0f
	3.0.5 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00290.html
	3.1 20020313	bootstrap failure, PR libgcj/5944
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00308.html
alpha-dec-osf5.1	CC=cc
			--disable-nls --enable-libgcj alpha-dec-osf5.1
	3.0.5 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00291.html
	3.1 20020313	bootstrap failure, PR target/5949
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00309.html
i386-pc-solaris2.8	--disable-nls --enable-libgcj
	3.0.5 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00292.html
	3.1 20020313	bootstrap failure, PR bootstrap/5948
	doesn't even get to completing the 3-stage
mips-sgi-irix6.5	CC=cc
			--disable-nls --enable-libgcj
	3.0.5 20020313	ok
	3.1 20020313	bootstrap in progress
	still in progress, this machine is dead slow ;-(
sparc-sun-solaris2.5.1	CONFIG_SHELL=/bin/ksh
			--disable-nls
	3.0.5 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00293.html
	3.1 20020313	bootstrap failure, PR libgcj/5944
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00315.html
sparc-sun-solaris2.8	CONFIG_SHELL=/bin/ksh
			--disable-nls --enable-libgcj
	3.0.5 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00294.html
	3.1 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00316.html
I've posted the bootstrap results for those cases where bootstraps failed
nonetheless, since (except for sparc-sun-solaris2.5.1) --enable-libgcj is
off by default and the remaining testsuite results are still valid.
So libgcj support regressed from its 3.0 state on alpha-dec-osf4.0f and
sparc-sun-solaris2.5.1 due to a libgcj bug (PR libgcj/5944), on
alpha-dec-osf5.1 due to a gcj bug (PR target/5949) and i386-pc-solaris2.8.
Only sparc-sun-solaris2.8 is sort-of-ok:
The -m32 testresults look reasonable, the -m64 testresults are terrible,
but probably due to known problems: 
* The header generation tests fail since -m64 is passed to gcjh, which
 doesn't recognize that option:
FAIL: calls header generation
/vol/gcc/obj/gcc-3.1-20020313/8-gcc-java/gcc/gcjh: unrecognized option `-m64'
Try `gcjh --help' for more information.
 I've just found this is a known problem: PR other/5874.
* All the execution tests seem to fail because some of the required shared
 libraries cannot be located:
FAIL: Array_1 execution from source compiled test
 There's no trace in libjava.log why this is happening, but if I manually
 run Array_1 with LD_LIBRARY_PATH set correctly, it passes. If I recall
 correctly, a similar problem exists on IRIX 6, which has a bi-arch
 compiler as well, so this is most likely a generic problem in the libjava
 testsuite.
	Rainer