On Thu, Jul 06, 2006 at 05:27:53PM -0600, Michael Fuhr wrote: > On Solaris 9, gcc 3.4.2 with -ansi doesn't define int64_t (even > when including <inttypes.h>), and with -pedantic it warns about > long long. With -ansi -pedantic -Wno-long-long it recognizes long > long without complaint.
I just noticed that the conftest program is compiled without the options that will be used for the build. I had edited acsite.m4 and changed <stdint.h> to <inttypes.h>, which resulted in configure finding int64_t because the conftest compile didn't use -ansi; but then when the build used -ansi the compiler failed, complaining that int64_t isn't a type. Setting CFLAGS before the PGAC_TYPE_64BIT_INT tests in configure.in fixed that problem (configure then found long long int as the 64-bit type). Here's a summary of what works on Solaris 9/sparc with gcc 3.4.2: * In acsite.m4 include <inttypes.h> instead of <stdint.h>. * In configure.in, before the PGAC_TYPE_64BIT_INT tests, set CFLAGS to have the same options that will be used for the actual build. With these changes (and after replacing a buggy libtool) GEOS builds and passes all tests on Solaris 9/sparc 32-bit. I haven't done a 64-bit build yet. -- Michael Fuhr _______________________________________________ geos-devel mailing list geos-devel@geos.refractions.net http://geos.refractions.net/mailman/listinfo/geos-devel
