I've rebuilt 5118 rawhide-20071220 src.rpms on x86_64 in mock
buildroots which
contained rawhide-20071220 except {gcc,lib}*-4.1.2-36.*.rpm, with additional
gcc-4.3.0-0.4 (available from koji, dist-f9-gcc43 component),
compat-libgfortran-41 (available from
http://people.redhat.com/jakub/gcc/compat-libgfortran-41-4.1.2-36.src.rpm )
and later on also with gettext subpackages just rebuilt with gcc-4.3.0-0.4.
Out of those 5118 src.rpms 1054 were failed builds. For those that failed
to build, I have retried with stock rawhide-20071220 mock buildroots (i.e.
gcc-4.1.2-36). 547 failed builds failed even with 4.1.2-36 (this count
includes even ExclusiveArched rpms etc.), logs for those are at
http://sunsite.mff.cuni.cz/rawhide20071220-gcc43/fails-even-with-41/
and generally don't interest me, as this is not a regression introduced by
4.3.0-0.4.
The remaining 507 failures only fail with gcc-4.3.0-0.4 and not with
gcc-4.1.2-36, though most of them are just C++ being stricter, something
that ought to be fixed in the packages.
I've tried to quickly grep through the failed logs and categorize them:
[...]
88
http://sunsite.mff.cuni.cz/rawhide20071220-gcc43/unsorted/
Unsorted, package maintainers please check this out, if you
believe there is a GCC bug rather than package bug, get in
touch with me with additional details.
both fail with
"checking how to run the C++ preprocessor... /lib/cpp
configure: error: C++ preprocessor "/lib/cpp" fails sanity check
See `config.log' for more details."
But I don't have access to the config.log to see whats going on.
I talked with upstream on IRC and they told me that for them it builds with
"gcc (SUSE Linux) 4.3.0 20071129 (experimental) [trunk revision 130511]"
So this might be a gcc bug.
Any ideas?