Prabhjeet Singh Grover wrote: > I have set CC to 'gcc' and CFLAGS to '-lsocket'. Tried setting CC to > 'gcc -ansi' also but without success. I am attaching the config.log file > as I was unable to figure out the issue. CFLAGS="-lsocket" seems quite wrong. That belongs in LDFLAGS not CFLAGS (and ideally you would not have to specify it at all as the configure script should figure out the proper flags for using sockets, but that's another issue.) I'm not sure if this is responsible for the failure at hand, which is: configure:4863: checking for gcc option to accept ANSI C configure:4933: gcc -c -lsocket -I/usr/local/include -I/usr/include -I/usr/local/BerkeleyDB.4.4/include -I/usr/local/ssl/include conftest.c >&5 conftest.c: In function `f': conftest.c:33: `__builtin_va_alist' undeclared (first use in this function) conftest.c:33: (Each undeclared identifier is reported only once conftest.c:33: for each function it appears in.) where line 33 is: | va_start (v,p); That to me implies a broken compiler installation. Perhaps there is some header being erroneously included from somewhere that is defining va_start() to expand to __builtin_va_alist, maybe it's happening internally to the compiler. Or it could be the fact that -lsocket is present since it's in CFLAGS, but -l makes no sense in the context of compiling (-c) since libraries are only involved in linking. This is why it should go in LDFLAGS instead, but I'm not sure if that actually matters here. You need to first figure out why va_start() is broken. Try compiling the conftest.c testcase from config.log as simply "gcc -c conftest.c" and see if that succeeds. If it does, the the compiler is fine and it's one of the options being passed to gcc that is causing the failure; otherwise it's a broken installation of gcc and you'll need to track that down. I'm not all that familiar with Solaris so I can't be of much help there. Brian _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf
