This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

libjava status on Solaris 8/Intel and IRIX 6.5


I've finally completed bootstraps and tests on i386-pc-solaris2.8 (with
Tom's patch for PR bootstrap/5948) and mips-sgi-irix6.5.  Both sets of
results look good:

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	ok with fix for PR bootstrap/5948
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00405.html

		=== libjava Summary ===

# of expected passes		1993
# of unexpected failures	25
# of expected failures		19
# of untested testcases		36

mips-sgi-irix6.5	CC=cc
			--disable-nls --enable-libgcj

	3.0.5 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00349.html

	3.1 20020313	ok
	http://gcc.gnu.org/ml/gcc-testresults/2002-03/msg00406.html

		=== libjava Summary for unix ===

# of expected passes		1910
# of unexpected failures	29
# of expected failures		18
# of untested testcases		116

		=== libjava Summary for unix/-mabi=64 ===

# of expected passes		581
# of unexpected failures	531
# of expected failures		12
# of untested testcases		910

The -mabi=64 tests are much worse due to one common error:

ld64: FATAL   9  : I/O error (-ldl): No such file or directory

This happens because there's no N64 libdl.so on IRIX 6, only an N32 one.
This is correctly recognized when generating libgcj.spec: the N32 one (in
mips-sgi-irix6.5/libjava) has

*lib: -lgcj -lm  -lgcjgc  -lzgcj   -ldl %(libgcc) %(liborig)

whereas mips-sgi-irix6.5/mabi=64/libjava/libgcj.spec has only

*lib: -lgcj -lm  -lgcjgc  -lzgcj   %(libgcc) %(liborig)

but make check doesn't use the different spec file.  While this should
probably fixed since this problem might affect other platforms as well,
there seems to be a different solution: the N32 libdl.so is only a dummy
and doesn't define anything, so both ABIs could do without.  If we can test
for this (probably by having a test program with dlopen and trying to build
and run it without -ldl first, falling back to -ldl only if that fails),
N64 testsuite results should be fine as well.

I'm now rerunning the -mabi=64 testsuite with the N32 libgcj.spec to
provide results unaffected by this bug.

	Rainer


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]