On Thu, Apr 16, 2009 at 12:16 AM, Bob Friesenhahn <bfriesen@xxxxxxxxxxxxxxxxxxx> wrote: > On Wed, 15 Apr 2009, Douglas Held wrote: > >> Well, you would just change autoconf so that it fails fast when its >> configuration is not "well formed"... > > How would autoconf know that? Also, automagic things that work in most cases fail in a few... I would not like `fail fast on old packages'. I understand that for people building linux packages, which often may change quickly and be well maintained, this might be desirable. Simple mature packages, however, in the past already needed some changes just because of autoconf improvements (new errors get detected), which might not be good in any case. I use autoconf also to compile libs for embedded devices. On those packages sometimes not even make install works :-) (I know that this is not what autoconf is for and that we use it wrongly etc, but I like autoconf because it can be used in a simple way - in the end configure is just a shell script - and its most flexible build tool I know :)). If you need some "configure && make && make install DESTDIR=" on win32 why not having a wrapper that aligns --prefix of configure? On win32 things are strange anyway. You cannot have a absolute PATH in Makefile (because of colon) etc pp If you like, you can cross compile under linux for win32 (with e.g. gcc-mingw32 or WINE + $CC) - DESTDIR will work :-) I'm not sure about the details, but those annoying drive letters may not be needed on win32 always, sometimes UNC pathes like //myhost/windows/temp and so on may work. On cygwin the path mapping can be really painful. On MinGW, you usually use /c/windows/tmp which works with DESTDIR (in same way as on unix except of course that you get a directory called like a drive letter, which shouldn't be a problem). oki, Steffen _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
