This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: target/2669: [Solaris 8/Intel] shared libgcc isn't built
> A make check is just underway, I'll submit the results to gcc-testresults
> and post a comparison once this is done. If all goes well and appropriate
> names for the two t-* files can be suggested, this should be ready for
> check-in.
The first set of testresults are at
http://gcc.gnu.org/ml/gcc-testresults/2001-04/msg00562.html
the second set (with a shared libgcc enabled) is
http://gcc.gnu.org/ml/gcc-testresults/2001-04/msg00563.html
There's only one difference:
+FAIL: gcc.c-torture/compile/20001226-1.c, -Os
in the shared libgcc case. I find
WARNING: program timed out.
compiler exited with status 1
in gcc.log. But this is due to the fact that cc1 needs excessive amounts
of memory here:
PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND
13080 ro 1 60 0 133M 94M sleep 0:12 12.84% cc1
on a machine with 128 MB physical, so it's paging to death ;-( This may
well by due to having run make -j2 check this time, but not in the first
run. Manually running only this test completed in 3 min 39 sec, well
withing dejagnu's 300 sec timeout. So this shouldn't be considered as a
problem with the shared libgcc patch, which didn't cause any regressions,
except pseudo-regressions for the fortran tests:
What is highly confusing is the way the testsuite deals with a missing
language: as I said, I've configured the second bootstrap without f77
(i.e. with --enable-languages=c++,objc), still all f77 tests are run, but
obviously fail. The test framework should be able to cope with this. It
happens since gcc/Makefile.in unconditionally contains check-g77. This
variable needs to be constructed at configure time like LANGUAGES is.
Rainer
-----------------------------------------------------------------------------
Rainer Orth, Faculty of Technology, Bielefeld University
Email: ro@TechFak.Uni-Bielefeld.DE