On 30/03/12 13:36, niXman wrote: > I created the bug-report: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52787 > > I think there is still hope to make Transactional Memory work in MinGW. Sure, it shouldn't be too hard, but it obviously needs work. Kai Tietz said in the bug that "You have to make sure that pthread-API is present, as libitm depends on pthread." You can try luck using some windows pthread implementation. Although I find a bit strange that to be the only issue, as I'd expect it to produce an error about missing the function/library not a segmentation fault (note that windows doesn't do lazy loading). Alternatively, you could easily implement a basic implementation of __transaction_atomic which simply uses a global mutex, and disable the wrapper when mingw support gets added.
