From: Ruben V. B. <van...@gm...> - 2010-07-28 16:39:09
|
> > > environment variable MSYSTEM. So if MSYSTEM is changed to say mingw64 > > then the config.guess and config.sub scripts could be changed to > > recognize the uname string to produce i686-pc-mingw64, x86_64-mingw64 or > > even x86_64-w64-mingw32 which ever is preferred. The uname string from > > MSYS would read MINGW64_NT-5.1 if MSYSTEM=mingw64. > > > > I believe MSYSTEM should be set uppercase, as setting mingw32 > lowercase do not identify properly. > > This change should be send upstream? > > > My personal preference for a config.guess return string would be > > x86_64-w64-mingw. > > > > I believe this was already established as "x86_64-w64-mingw32" ? > > http://sourceforge.net/apps/trac/mingw-w64/wiki/MSYS > Great, someone actually found the wiki! MSYS itself is 32-bit, so having uname return something else is not quite correct. There are other things bad about this suggestion: 1. What if you have installed x86_64-w64-mingw32-gcc, i686-w64-mingw32-gcc, and i686-pc-mingw32-gcc, and added all three paths to MSYS's path? uname isn't at all a correct way to determine the compiler's target, especially when you think about cross-compilation, and (future/current) multilib. 2. The wiki states the workaround/"The Right Way" to use mingw-w64 under MSYS. I believe passing -host=... to configure is a small price to pay (if any) for the functionality you get from using MSYS+mingw-w64. The MSYS package provided by mingw-w64 (aka me in this context) is only the complete binary set of packages conveniently packed into one. I don't feel I have the right/will to change internal settings (of which I do not know how far these reach) or to keep any changes up to date. I hope you can understand my reasoning, and I'll be happy to read any arguments against what I stated above and consider revising my opinion, but MSYS works fine as is. Ruben |