This is the mail archive of the
libstdc++@gcc.gnu.org
mailing list for the libstdc++ project.
How seriously to take abi-check
- From: Jim Osborn <jimo at eskimo dot com>
- To: libstdc++ at gcc dot gnu dot org
- Date: Thu, 2 Mar 2006 20:03:36 -0800
- Subject: How seriously to take abi-check
I built gcc-3.4.5 with the --disable-nls option. Configure and
"make bootstrap" made no complaints, but abi_check fails with this
summary:
=== libstdc++-v3 check-abi Summary ===
# of added symbols: 58
# of missing symbols: 46
# of incompatible symbols: 46
using: baseline_symbols.txt
FAIL: abi_check
Many of the symbols mentioned in testsuite/libstdc++.log seem to
be locale related, so I could believe they're not in current_symbols.txt
because of the --disable-nls option, and that abi_check isn't clever
enough to ignore them from baseline_symbols.txt, but I don't know.
Would that option break this test?
Is there any documentation that would tell me if this:
_ZNSt12__basic_fileIcE4fileEv
std::__basic_file<char>::file()
GLIBCXX_3.4.1
function
unknown
or this:
_ZN9__gnu_cxx17__pool_alloc_base12_M_get_mutexEv
__gnu_cxx::__pool_alloc_base::_M_get_mutex()
GLIBCXX_3.4.2
function
unknown
is also nls related?
Is there a way to get a "baseline" that corresponds to the config
options supplied to the build? If not, is this test generally
taken very seriously? It seems to stop "make check" when it
fails, so I'm tempted to try to make it pass, but maybe I'm
wasting my time.
Any suggestions appreciated,
Jim