Re: release Autoconf?

[Akim Demaille]

>>>>> "Mo" == Mo DeJong <address@hidden> writes:

Mo> Some systems don't even have a /lib/cpp so picking it by default
Mo> seems wrong. 

Well, this is not what 2.49 does:

   if test -z "$CPP"; then
     AC_CACHE_VAL([ac_cv_prog_CPP],
     [dnl
       # Double quotes because CPP needs to be expanded
       for CPP in "$CC -E" "$CC -E -traditional-cpp" "/lib/cpp"
       do
         # break 2 since there is a loop in there.
         _AC_PROG_PREPROC_WORKS_IFELSE([break 2])
       done
       ac_cv_prog_CPP=$CPP
     ])dnl
     CPP=$ac_cv_prog_CPP
   else
     ac_cv_prog_CPP=$CPP
   fi
   AC_MSG_RESULT([$CPP])
   _AC_PROG_PREPROC_WORKS_IFELSE([],
                 [AC_MSG_ERROR([C preprocessor "$CPP" fails sanity check])])
   AC_SUBST(CPP)dnl

It is really tested, and at the end of the loop, there is a last check
to make sure we don't use something broken.  Maybe 2.13 was not like this.


Mo> For example, compiling with VC++ cross under Cygwin does not work
Mo> at all. It tries $CC -E and fails because VC++ prints a file name
Mo> to stderr.

Do you still face this failure?  What can we do to have VC++'s -E
used.  If the $? is correct, then it should work fine including if
VC++ dumps Microsoft propaganda.