On 30 Mar 2011, at 23:48, John Wiegley wrote:
I'm not sure yet whether this is a Boost issue, or a g++ 4.6 issue. I've compiled Boost 1.46.1 on Mac OS X using g++ 4.6 and these build options:
sh bootstrap.sh && ./bjam debug \ --toolset=darwin --build-type=complete --layout=versioned \ define=_GLIBCXX_FULLY_DYNAMIC_STRING=1 define=_GLIBCXX_DEBUG=1 \ install
Did you build g++ with _GLIBCXX_FULLY_DYNAMIC_STRING? Or are you sure where it is picking up it's standard library from? My suspicion is that you are hitting the (which you seem to know about) _GLIBCXX_DEBUG and std::string bug in Mac OS X. Unfortunately I don't have a good answer on how to avoid it, I have tried various workarounds, but I always seem to end up getting bitten in the end.
Then I build and execute the following example code:
#include <boost/regex.hpp> #include <string>
int main(int argc, char *argv[]) { boost::regex expr("foo.*bar"); return boost::regex_search(std::string(argv[1]), expr) ? 0 : 1; }
with this compile/link command:
g++-mp-4.6 -D_GLIBCXX_FULLY_DYNAMIC_STRING=1 -D_GLIBCXX_DEBUG=1 \ -g -o regex regex.cc -lboost_regex-xgcc46-d-1_46_1
and the result is:
terminate called after throwing an instance of 'std::bad_cast' what(): std::bad_cast
This is happening in boost::re_detail::cpp_regex_traits_base<char>::imbue, on line cpp_regex_traits.hpp:218.
The same everything, using g++ 4.5, works fine, so maybe this is a g++ 4.6 bug? I seem to recall having seen this problem before, due to mismatching a debug Boost with a non-debug client app, or something similar. But I've have verified that my client app is using the same flags that I build Boost with.
Has anyone run into this? Do I need to take this bug to the g++ folks? Any pointers, ideas?
Thanks, John _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost