http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51212
--- Comment #3 from Richard Henderson <rth at gcc dot gnu.org> 2011-12-21 17:18:11 UTC --- As far as the compiler goes that all sounds plausible. The main problem is that we would have to make libitm exception safe. I'm 100% sure that won't work at the moment. I'm also sure that it would be exceedingly awkward to take an exception in the middle of the actual transaction. Unless Torvald has a brilliant idea there, I'm inclined to simply emit a sorry for combining -fgnu-tm and -fnon-call-exceptions. Unless we redefine that combination such that we only support the exception case for NULL, but non-null faulting memory references just crash. In that case we can simply test for NULL at the start of the accessors and explicitly throw the exception.