Re: [PATCH 1/1] libtirpc: Add detection for new rpc_gss_sec members

[Petr Vorel <pvorel@suse.cz>]

Hi Steve,

Thanks for having a look.

> Hello,

> On 10/25/23 2:01 PM, Petr Vorel wrote:
> > From: Petr Vorel<petr.vorel@gmail.com>

> > 4b272471 started to use struct rpc_gss_sec member minor_status, which
> > was added in new libtirpc 1.3.4. Add check for the member to prevent
> > failure on older libtirpc headers.

> > Fixes: 4b272471 ("gssd: handle KRB5_AP_ERR_BAD_INTEGRITY for machine credentials")
> > Signed-off-by: Petr Vorel<pvorel@suse.cz>
> > ---
> >   aclocal/libtirpc.m4 | 4 ++++
> >   1 file changed, 4 insertions(+)

> > diff --git a/aclocal/libtirpc.m4 b/aclocal/libtirpc.m4
> > index bddae022..dd351722 100644
> > --- a/aclocal/libtirpc.m4
> > +++ b/aclocal/libtirpc.m4
> > @@ -25,6 +25,10 @@ AC_DEFUN([AC_LIBTIRPC], [
> >                            [AC_DEFINE([HAVE_LIBTIRPC_SET_DEBUG], [1],
> >                                       [Define to 1 if your tirpc library provides libtirpc_set_debug])],,
> >                            [${LIBS}])])
> > +     AS_IF([test "$enable_gss" = "yes"],
> > +           [AC_CHECK_MEMBER(struct rpc_gss_sec.minor_status,,
> > +                         [AC_MSG_ERROR([Missing rpc_gss_sec.minor_status in <rpc/auth_gss.h>, update libtirpc or run with --disable-gss])],
> > +                         [#include <rpc/auth_gss.h>])])
> >     AC_SUBST([AM_CPPFLAGS])
> >     AC_SUBST(LIBTIRPC)
> > -- 2.42.0

> This does not work... since it is looking at that gssrpc/auth_gss.h
> instead of the tirpc/rpc/auth_gss.h so the check fails
Is it? There is no <gssrpc/auth_gss.h>. I suppose you test on some recent
Fedora, I'll retest it.

I tested it on openSUSE Tumbleweed, where libtirpc-devel is
installed into /usr/include/rpc/, thus /usr/include/rpc/auth_gss.h exists.
But on Debian (and likely on RHEL/Fedora as you noticed it) is on
/usr/include/tirpc/rpc/auth_gss.h.

I hoped that this is handled elsewhere via -I/usr/include/tirpc.
So, I'm really confused why would have look at <gssrpc/auth_gss.h>.

Kind regards,
Petr

> I like the idea of having the check, but I'm not sure on
> how to point it in the right direction.

> steved.