On Sun, 30 Nov 2003, Matze Braun wrote: > > There's no universally-acceptable default. Generally speaking, the > > old UNIX tradition (namely, no debugging and no optimization) is less > > suitable these days than the autoconf default (debugging and a > > reasonable level of optimization). > You're missing the point of my last mail. I agree that -g -O2 is a better > default. But I don't think it's a good idea that autoconf enforces this > default. The point here is usability: > > As a new user I don't expect autoconf to set CFLAGS in a macro whose main > purpose is to detect the name of the compiler. Why is this macro more > complicated than it could be? Why mess with such things as compiler > defaults here? Perhaps because different compilers accept different options. If the configure script supplies unsupported options, configure will fail entirely. > AC_PROG_CC > CFLAGS=-O3 -g > AC_ARG_ENABLE([debug], [AC_HELP_STRING([--enable-debug], > [build with debugging information (default NO)])], > [test "$enableval" = "yes" && CFLAGS="-O0 -g3"]) Apparently this configure script only works with one compiler. Too bad. Hopefully it doesn't assume just one operating system as well. > However if autoconf wouldn't set default, the example just wouldn't > need the extra workaround, while the simple case would still be be easy to > understand at the first place: > > AC_PROG_CC > CFLAGS=-O2 -g A configure script like this will go "poof" as soon as it is used with a compiler that doesn't support -O2, or -O simultaneous with -g. Bob ====================================== Bob Friesenhahn bfriesen@xxxxxxxxxxxxxxxxxxx http://www.simplesystems.org/users/bfriesen
