• Ports
• Configure

Configure Quirks

Zero exit status doesn't guarantee you that configure script did not stumble upon any problems. When making a port, you should carefully examine its output, looking for any problems and trying to understand every "... no" the script encounters.

ac_compile vs ac_cpp

Synopsis

checking vorbis/codec.h usability... yes
checking vorbis/codec.h presence... no
configure: WARNING: vorbis/codec.h: accepted by the compiler, rejected by the preprocessor!
configure: WARNING: vorbis/codec.h: proceeding with the compiler's result

Under the hood

ac_cpp='$CPP $CPPFLAGS'
ac_compile='$CC -c $CFLAGS $CPPFLAGS conftest.$ac_ext >&5'

Workaround

CFLAGS+=        -I${LOCALBASE}/include
LDFLAGS+=       -L${LOCALBASE}/lib
CONFIGURE_ENV+= CPPFLAGS="${CFLAGS}" LDFLAGS="${LDFLAGS}"

Notice CPPFLAGS instead of CFLAGS in CONFIGURE_ENV

Comparisons

Synopsis

checking for netcdf library... test: x/usr/local/lib: unexpected operator
test: x/usr/local/lib: unexpected operator
test: x-L/usr/local/lib -lnetcdf: unexpected operator
-L/usr/local/lib -lnetcdf

Under the hood

if test x"${netcdf_lib_dir}" == x ;

Workaround

Here are some hardcore substitutions:

@${REINPLACE_CMD} -E \
        -e 's/test x(\".*\") ==? x([^:_."[:alnum:]-]|$$)/[ -z \1 ]\2/' \
        -e 's/test x(\".*\") != x([^:_."[:alnum:]-]|$$)/[ -n \1 ]\2/' \
        -e 's/test \"x(.*\") ==? x([^:_."[:alnum:]-]|$$)/[ -z \"\1 ]\2/' \
        -e 's/test \"x(.*\") != x([^:_."[:alnum:]-]|$$)/[ -n \"\1 ]\2/' \
        -e 's/test (\".*)\+set(.*)\" ==? set([^:_."[:alnum:]-]|$$)/[ -n \1\2\" ]\3/' \
        -e 's/test (\".*)\+set(.*)\" != set([^:_."[:alnum:]-]|$$)/[ -z \1\2\" ]\3/' \
        -e 's/test (.*) ==/test \1 =/' ${WRKSRC}/configure

In fact a simple '/test/s|==|=|g' is sufficient in many cases. Other ones are beautifiers.