> The real problem that I was trying to illuminate is that the fact that > if a particular > interface was implemented as a macro then the AC_CHECK_FUNCS() breaks. Most POSIX functions are required to exist as a linkable function, even if they are also implemented as a macro. There are relatively few exceptions on this front. > Some seem to argue that one should always know if a particular interface > might > be implemented as a macro (or compiler internal?). If so, then you have to > roll your own autoconf macro. My point is that it would be more user > friendly > if AC_CHECK_FUNCS (or a newly named AC macro) were to presume that if > either you could link to the external name or if the external name were > #define-d, > then you #define HAVE_INTERFACE. It already exists - AC_CHECK_DECLS. And for the weirdest of cases, the combination of AC_CHECK_DECLS with AC_CHECK_FUNCS gives you the full check support for all four possible cases: declared and working mistakenly declared but not implemented linkable but undeclared completely missing -- Eric Blake _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
