[Bug libstdc++/56587] [4.8 regression] libstdc++-abi/abi_check fails for powerpc

jakub at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Sun Mar 10 16:41:00 GMT 2013


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=56587

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> 2013-03-10 16:40:45 UTC ---
The ABI files look correct to me, powerpc-linux-gnu/baseline_symbols.txt
(== powerpc64-linux-gnu/32/baseline_symbols.txt ) are -m32 files, _M_next_bkt
takes std::size_t argument and the baseline_symbols.txt files for that use j
letter (i.e. unsigned int), while for 64-bit
powerpc64-linux-gnu/baseline_symbols.txt _M_next_bkt uses m letter (i.e.
unsigned long).

find config/abi/post/powerpc* -name \*.txt | xargs grep _M_next_bkt
config/abi/post/powerpc64-linux-gnu/baseline_symbols.txt:FUNC:_ZNKSt8__detail20_Prime_rehash_policy11_M_next_bktEm@@GLIBCXX_3.4.18
config/abi/post/powerpc64-linux-gnu/32/baseline_symbols.txt:FUNC:_ZNKSt8__detail20_Prime_rehash_policy11_M_next_bktEj@@GLIBCXX_3.4.18
config/abi/post/powerpc-linux-gnu/baseline_symbols.txt:FUNC:_ZNKSt8__detail20_Prime_rehash_policy11_M_next_bktEj@@GLIBCXX_3.4.18

If you are building powerpc64-linux --with-cpu=default32, you need to shuffle
the baseline_symbols.txt around (64-bit move into 64/ subdirectory, 32/
subdirectory contents up), but that is not in any way a regression, and if you
haven't done this shuffle you'd get hundreds of other failures, not just these.



More information about the Gcc-bugs mailing list