GCC 3.0 first pass
The results from the first pass at GCC 3.0 (or maybe it's 3.1), using the
3.0 build package, but replacing the tarball with one taken from the CVS
HEAD as of 2002-02-10 (after removing lots of patches that got fixed in the
upstream...):
=== libstdc++-v3 Summary ===
# of expected passes 24
# of unexpected failures 173
# of unsupported tests 6
(followed by a make error for check-DEJAGNU which concerns me)
=== gcc Summary ===
# of expected passes 17617
# of unexpected failures 25
# of expected failures 54
# of unresolved testcases 8
# of unsupported tests 37
=== g++ Summary ===
# of expected passes 4933
# of unexpected failures 1036
# of unexpected successes 3
# of expected failures 974
# of untested testcases 15
# of unsupported tests 1
=== g77 Summary ===
# of expected passes 1452
# of unsupported tests 8
=== objc Summary ===
# of expected passes 1036
# of unexpected failures 6
As is fairly evident, this is far better than the GCC 2.95.4 results; but
libstdc++-v3 is still very broken (and, I suspect, is causing test failures
in g++, though I can't easily prove it). Many of the libstdc++-v3 failures
are "failed to produce binary", and that is the only piece with a Dejagnu
failure as well. So, *my* summary:
g77 passes with flying colors (0/1452 failed)
objc probably passes (6/1042 failed)
gcc probably passes (25/17696 failed)
g++ fails (1039/6946 failed)
libstdc++-v3 fails miserably (173/197 failed)
*IF* I can figure out why the C++ stuff is failing so badly, and get it
beaten into shape, I think this is a strong recommendation in favor of 3.0
as the default compiler on NetBSD; the raw C compiler certainly appears to
be in good shape.
--
***************************************************************************
Joel Baker System Administrator - lightbearer.com
lucifer@lightbearer.com http://users.lightbearer.com/lucifer/
Reply to: