This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: GCC 3.4 Release Status (2004-01-05)
> sparc-sun-solaris:
> http://gcc.gnu.org/ml/gcc-testresults/2004-01/msg00289.html
Brad's testsuite results are not representative, because he appears not to
have a standard configuration.
The representative results for sparc-sun-solaris and sparc64-sun-solaris are
those of Kaveh:
SPARC: http://gcc.gnu.org/ml/gcc-testresults/2004-01/msg00128.html
The compat failures are caused by Mark's recent patch, which he has pledged
to fix. So we have 3 FAILs for the C testsuite:
FAIL: gcc.c-torture/execute/20030307-1.c execution, -O2
FAIL: gcc.c-torture/execute/20030307-1.c execution, -Os
Corner-case problem with sibcalls and pure functions. I have a patch, but it
pessimizes too much for mainline. I may eventually put it on the branch.
FAIL: gcc.dg/20031202-1.c execution test
Introduced by Jakub's RTX_UNCHANGING_P patch. He proposed a potential
solution some time ago.
We have 0 FAIL for the C++, Fortran and Objective-C testsuites.
SPARC64: http://gcc.gnu.org/ml/gcc-testresults/2004-01/msg00129.html
Same remark for the compat failures, so we have 0 FAIL for the C and Fortran
testsuites.
We have 2 FAILs for the C++ testsuite:
FAIL: g++.dg/parse/stack1.C (test for excess errors)
Problem with the stack size limit on Solaris.
FAIL: g++.old-deja/g++.other/crash34.C (test for excess errors)
Real problem. I posted a message about it some time ago, but got no answer:
http://gcc.gnu.org/ml/gcc/2003-12/msg01072.html
Could a specialist of the C++ front-end take a look at it?
We have 5 FAILs for the Objective-C testsuite, which are visible on all LP64
platforms.
Summary: C, C++, Objective-C and Fortran are in pretty good shape on both
platforms. However, there are two issues that I'd like to sort out before
the 3.4 release (but they are not regressions):
- ABI conformity on SPARC64: Mark Kettenis recently discovered that GCC
deviates from the ABI for small structures (and unions). See PR 13557,13559.
- FP register alignment on SPARC64. See PR 10904,13058.
On libffi/libjava: they don't get much interest on these platforms. libjava
is in good shape and I suspect the failures for libffi are the symptoms of a
unique problem.
On Ada: currently broken by Mark's patch. Otherwise the results are good (8
FAILs in the testsuite).
--
Eric Botcazou