From: Bruce Hobbs Date: 2002-07-04T10:04:37+09:00 Subject: Re: Make fails with missing header file (OS X) At 8:56 AM +0900 on 7/4/2002, Chris Gehlker wrote: >Well, most folks don't bother to rebuild configure from configure.in so the >version of autoconf doesn't matter. If memory serves, you *can* still build >configure for Ruby 1.6.7 with autoconf 2.13. > I only tried autoconf after I encountered the missing file. And, autoconf 2.13 did not work with Ruby 1.6.7, at least not for me. >There are only a couple of places where the build might try to read dlfcn.h. >One is in dln.c where the condition is: #ifdef USE_DLN_DLOPEN and the other >is in stubs.c where the condition is: #elif defined HAVE_DLOPEN. Now this is >all from memory of someone else who had this problem, but, IIRC, make >searches for some file and if it finds it, it sets -DHAVE_DLOPEN. > The error pointed at dln.c. >I assume that it is finding this key file in /usr/local/pgsql/bin but I'll >let you track it down if you care. > There seem to be two references to HAVE_DLOPEN in dln.c: #ifdef __CHECKER__ #undef HAVE_DLOPEN #undef USE_DLN_A_OUT #undef USE_DLN_DLOPEN #endif #if defined(HAVE_DLOPEN) && !defined(USE_DLN_A_OUT) && !defined(_AIX) /* dynamic load with dlopen() */ # define USE_DLN_DLOPEN #endif But the error appears to be coming from here: #ifdef USE_DLN_DLOPEN # ifdef __NetBSD__ # include # include # else # include # endif #endif Since, thanks to you, I was able to workaround the problem I think I'll just let it drop for now. My head hurts just looking at the definitions, much less trying to figure out exactly where things went wrong. >The irony is that the real dlfcn.h only >prototypes some functions for AIX. It can be an empty file everywhere else. > Given that's the case it wouldn't seem all that difficult to "fix" things so this wouldn't happen. Easy for me to say. ;-) -- Bruce Hobbs, CDP 856 N Monterey St Phone: 626-570-8028 Partner Alhambra, CA 91801-1574 Cell: 626-278-0273 Engineered Software brucehobbs@engineeredsw.com FAX: 208-474-0732