On Wed, Jul 29, 2009 at 10:29, Steffen
Dettmer<steffen.dettmer@xxxxxxxxxxxxxx> wrote:
> you would have flag list redundant and it could happen someone will
> forget to change one place when adding next flag...
> so maybe just call a function?
I was thinking along those lines myself.
> the withval isn't used? Wouldn't it be --enable-FEATURE then?
probably, I've recently taken over maintenance of this code, which was
originally written over 8 years ago, and this change seems like it is
a good opportunity to use AC_ARG_ENABLE instead of AC_ARG_WITH as it
will minimise any disruption.
> dnl surely my example is wrong (quoting, ...) and not portable, I
> dnl don't know the details.
> dnl
> dnl --with-gcc_flags? maybe better configure CFLAGS=xyz?
> AC_DEFUN([LAL_WITH_GCC_FLAGS],
> [AC_ARG_WITH([gcc_flags],AC_HELP_STRING([--x],[y]),
> [case $withval in
> yes)
> DO_ENABLE_LAL_WITH_GCC_FLAGS;;
> no)
> ;;
> *)
> DO_ENABLE_LAL_WITH_GCC_FLAGS($withval);;
> esac ],
> [ DO_ENABLE_LAL_WITH_GCC_FLAGS; ] )
> ])
> dnl Sets several specific flags to variable lal_gcc_flags
> AC_DEFUN([DO_ENABLE_LAL_WITH_GCC_FLAGS],
> [
> lal_gcc_flags="-g3 -O4 -Wall -W -Wmissing-prototypes
> -Wstrict-prototypes -Wshadow -Wpointer-arith -Wcast-qual -Wcast-align
> -Wwrite-strings -fno-common -Wnested-externs -Wno-format-zero-length"
> case $host_cpu-$host_os in
> *i386-darwin*) lal_gcc_flags="${lal_gcc_flags} -pedantic" ;;
> *) lal_gcc_flags="${lal_gcc_flags} -pedantic-errors" ;;
> esac
> dnl setting CFLAGS is not indended by autoconf. Users should
> dnl set it if they want this (however, then they must set them all
> dnl and know compiler details etc).
> dnl In "else" at least an "compiler not supported" or "option
> dnl not supported" should be added when lal_gcc_flags is set.
> if test "$GCC" = yes; then
> CFLAGS="$CFLAGS $lal_gcc_flags"
> fi
> ])
Thanks, that's very helpful.
Cheers
Adam
_______________________________________________
Autoconf mailing list
Autoconf@xxxxxxx
http://lists.gnu.org/mailman/listinfo/autoconf
