Date: Sun, 29 Jan 2006 21:40:05 GMT From: Mike Durian <durian@shadetreesoftware.com> To: openoffice@FreeBSD.org Subject: Re: ports/92218: openoffice.org-2.0 build fails with shlibsign core dump Message-ID: <200601292140.k0TLe519042343@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR ports/92218; it has been noted by GNATS. From: Mike Durian <durian@shadetreesoftware.com> To: bug-followup@freebsd.org Cc: Subject: Re: ports/92218: openoffice.org-2.0 build fails with shlibsign core dump Date: Sun, 29 Jan 2006 14:31:18 -0700 Well, removing lang/gcc-ooo so it could be rebuilt did not help. I still get the unresolved __usleep symbol, but I think I see what is really going on. I've been using and upgrading FreeBSD for quite a while now and there were some symbolic links in /lib dated from 2003 that probably should have been removed at some point. In paticular, /lib/libc.so was a link to /lib/libc.so.5 When building with the stock system compiler, ld searches /usr/lib for libraries. The ld from gcc-ooo search /lib before /usr/lib. So when I linked with the system ld it found /usr/lib/libc.so which is a link to /lib/libc.so.6. /lib/libc.so.6 resolves __usleep. The ld from gcc-ooo found /lib/libc.so which is a link to libc.so.5. libc.so.5 does not resolve __usleep and thus causes the test for pthread_create in -lpthread to fail. I have removed the symbolic links in /lib and I'll see how thing go. Conceivably, it might be worth adjusting the gcc-ooo linker to use the same search behavior as the stock system linker to avoid this sort of confusion in the future. mike
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200601292140.k0TLe519042343>