[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Re: 2.0-alpha and Solaris 7
On Sat, 14 Aug 1999, David J N Begley wrote:
> Prima facie, this looks cool. Environ is Sol7/SPARC and gcc 2.8.1. I'd
> better wait for the release of "alpha2" before pushing this too much harder,
> to ensure that I'm chasing a genuine problem as opposed to one caused by
> mixing 'n' matching revs.
I managed to build CVS and pull down a copy of OpenLDAP-devel. As before,
"configure" now works for POSIX threads. The "inet_aton" issue hasn't gone
away, though.
- configure claims inet_aton is present, but it only appears to compile
a test programme, it doesn't actually *link* anything (at which point
it would fail)
- AC_LIBS needs "-lresolv" added for inet_aton to be included, as used by:
libraries/libldap
libraries/libldap_r
which in turn then affects everything linked with these libraries:
clients/tools
clients/ud
clients/finger
clients/gopher
clients/mail500
clients/fax500
clients/rcpt500
servers/slapd
servers/slapd/tools
servers/slurpd
tests/progs
- currently AC_LIBS on Sol7 contains "-lgen -lnsl -lsocket", so I was
thinking of modifying the configure.in tests (circa line 625):
AC_CHECK_LIB(socket, main)
[...]
AC_CHECK_LIB(nsl, main)
[...]
AC_CHECK_LIB(gen, main)
AC_CHECK_LIB(resolv, inet_aton)
the issue being, of course, that these tests are meant to find socket();
given that socket() is in -lsocket and not -lgen or -lnsl (yet they're
included anyway), it may not be such a bad thing after all...
The "main" check in -lsocket, -lgen and -lnsl appear to match the symbol
"_thr_main" in each library - there's no equiv in -lresolv so a "main"
check there wouldn't work
Alternatively, could the current check for inet_aton actually do a link
in addition to compile operation (which would fail), followed by a check
for inet_aton in -lresolv (compile/link would pass, add -lresolv to
AC_LIBS)?
- "make distclean" doesn't remove these files:
contrib/Makefile
contrib/saucer/Makefile
contrib/web_ldap/Makefile
include/lber_types.h
After all this, "make test" still results in a core dump anyway so now I need
to dig a little further...
dave