From: Keith M. <kei...@us...> - 2015-06-09 11:27:16
|
On 08/06/15 17:56, Eli Zaretskii wrote: > Sorry, I think MinGW runtime V4 is a dead end, I agree. > because the incompatible way it treated time_t makes it impossible to > build programs that run on old and new versions of Windows. Not sure in what way it was "incompatible", but it was just plain "wrong"; a bit of experimentation suggests that time_t is *always* 32-bit, when using MSVCRT.DLL, so _USE_32BIT_TIME_T is implicit, and unnecessary, (unless you jump through the hoops to substitute any of the brain damaged MSVCR80.DLL descendants). > So I'm sticking with 3.21.1 for the time being, which has all the > niceties of v4, but none of the problems (now that Keith fixed the > last couple of them). Actually, I overlooked one of them that I know of; the inline definition of hypotf(), in math.h, is broken in a way which causes a G++ compile which includes it, *and* has __STRICT_ANSI__ defined, to abort. I didn't have time to fix it properly, for 3.21.1, but I could have mitigated the problem in the interim, by returning the cast of hypot()'s return value, rather than _hypot()'s. -- Regards, Keith. |