On Wed, Apr 20, 2011 at 10:42:45PM -0700, Russ Allbery wrote: > Antoine Verheijen <[email protected]> writes: > > > Hi, all. I'm looking for a quick bit of advice. > > > First, the problem: OpenBSD has built-in Heimdal support but, in their > > infinite "wisdom", the developers elected to remove all trace of the > > roken library which they deemed unnecessary. As a result, the configure > > check for GSSAPI (via gssapi.m4) fails, even though -lgssapi exists, > > because the check hard-codes -lroken in its library list. > > Thanks for the report! We should use the same technique that was used to > solve the same problem in the krb5 library check for consistency, I > think. There, I probed for roken first, similar to your option 1, since > that fits the model used for all other optional dependency libraries. I > can see the appeal in option two, but it requires fiddling with internal > Autoconf cache variables, and I don't trust that to continue working in > later versions since it's not part of the documented Autoconf interface.
Thanks a lot for the response, Russ. I don't disagree with any of this. I was torn on the approach to use, particularly after Simon pointed out that this is part of a larger project. That's primarily why I elected to send this email as opposed to simply revamping my previous patch attempt. > > I'm on vacation at the moment so can't easily work on this, but I'll add a > fix to rra-c-util (which is upstream for this file) as soon as I get back. Sound great! No big hurry. I can work-around locally until it is changed. Based on this, I'll abandon my patch devoted to this and will await your mods. Many thanks. > > > --- a/src/cf/gssapi.m4 > > +++ b/src/cf/gssapi.m4 > > @@ -77,8 +77,11 @@ AC_DEFUN([_RRA_LIB_GSSAPI_MANUAL], > > AC_SEARCH_LIBS([crypt], [crypt]) > > rra_gssapi_extra="$LIBS" > > LIBS="$rra_gssapi_save_LIBS" > > + AC_CHECK_LIB([roken], [ct_memcmp], > > + [rra_gssapi_libroken=-lroken], > > + [rra_gssapi_libroken=]) > > AC_CHECK_LIB([gssapi], [gss_import_name], > > - [GSSAPI_LIBS="-lgssapi -lkrb5 -lasn1 -lroken -lcrypto -lcom_err" > > + [GSSAPI_LIBS="-lgssapi -lkrb5 -lasn1 $rra_gssapi_libroken -lcrypto > > -lcom_err" > > GSSAPI_LIBS="$GSSAPI_LIBS $rra_gssapi_extra"], > > [AC_CHECK_LIB([krb5support], [krb5int_getspecific], > > [rra_gssapi_extra="-lkrb5support $rra_gssapi_extra"], > > @@ -101,7 +104,7 @@ AC_DEFUN([_RRA_LIB_GSSAPI_MANUAL], > > [GSSAPI_LIBS="-lgss"], > > [AC_MSG_ERROR([cannot find usable GSS-API library])])], > > [$rra_gssapi_extra])], > > - [-lkrb5 -lasn1 -lroken -lcrypto -lcom_err $rra_gssapi_extra]) > > + [-lkrb5 -lasn1 $rra_gssapi_libroken -lcrypto -lcom_err > > $rra_gssapi_extra]) > > RRA_LIB_GSSAPI_RESTORE]) > > This is the right basic idea for how I'd fix this, except libroken can be > added to rra_gssapi_extra (that's what it's for) and hence the probe > should be done before setting rra_gssapi_extra and can be simpler. The > krb5 probe uses: > > AC_SEARCH_LIBS([rk_simple_execve], [roken]) > > before setting rra_gssapi_extra, and then the remaining changes aren't > required. > > libroken thankfully doesn't ever depend on libcrypto or libcom_err, so one > doesn't have to be careful about preserving that part of the library > ordering. I wasn't aware of this (since I didn't have a libroken handy). Good to know. It definitely makes it simpler. > > -- > Russ Allbery ([email protected]) <http://www.eyrie.org/~eagle/> > _______________________________________________ > OpenAFS-devel mailing list > [email protected] > https://lists.openafs.org/mailman/listinfo/openafs-devel Thanks again. :-) ------------------------------------------------------------------------ Antoine Verheijen Email: [email protected] AICT (formerly CNS) Phone: (780) 492-9312 University of Alberta Fax: (780) 492-1729 _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
