------- Comment #2 from paolo dot carlini at oracle dot com 2009-04-03 17:27 ------- Note, most if not all, those fails seem trivial: we are just instantiating with a second non-type template argument (__w) which is too big compared to the size of the type of the third argument, a 32-bit unsigned long. We could either change the latter to unsigned long long or reduce the value of the second. If Benjamin doesn't come to this I'll do it, to my taste, just to quickly shut-up the noise.
Also, I'm noticing the use of types like uint_fast64_t in the testcases, that at the moment is still not safe (will be when Joseph' work on PR 448 will be complete for all the OSes), because some targets may lack the stdint.h header completely. For now, instead of guarding the test with dg-require-cstdint I think we can safely change the types to some normal C++98 types. Or in fact maybe we could guard, it's easier to keep the test unchanged and remove the guards when 448 will be closed. Again, up to Benjamin of course, or I will shut the noise to my taste. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39629
