On 07 May 2015 13:52, Thorsten Kukuk wrote:
> On Thu, May 07, Mike Frysinger wrote:
> > Starting with commit d5259e751111cb108c784b044296185f543fc0be (Add header
> > definitions for rpc_gss_*() APIs), the gss headers were pulled in all the
> > time leading to build failures like so:
> > CC libtirpc_la-bindresvport.lo
> > In file included from ../tirpc/rpc/svc_auth.h:44:0,
> > from ../tirpc/rpc/rpc.h:68,
> > from bindresvport.c:46:
> > ../tirpc/rpc/rpcsec_gss.h:38:27: fatal error: gssapi/gssapi.h: No such file or directory
>
> Here is my proof of concept how I think we should solve this.
> But there is one part of your patch I have no solution for:
>
> > --- a/tirpc/rpc/svc_auth.h
> > +++ b/tirpc/rpc/svc_auth.h
> [...]
> > @@ -63,8 +67,10 @@ typedef struct SVCAUTH {
> > int (*svc_ah_destroy)(struct SVCAUTH *);
> > } *svc_ah_ops;
> > caddr_t svc_ah_private;
> > +#ifdef HAVE_RPCSEC_GSS
> > svc_rpc_gss_parms_t svc_gss_params;
> > rpc_gss_rawcred_t raw_cred;
> > +#endif
> > } SVCAUTH;
>
> You are changeing the size of a struct here. I'm not sure
> if this will work, if an application is compiled with headers
> where it is disabled and then runs with a tirpc where it is
> enabled. Or the other way around.
> Does somebody have an idea how to solve that?
i did notice that, but i would point out that this struct in the 0.2.5 lacked
these members. they're new to the 0.3.0 release. so if ABI is a concern, we
already lost that battle ;).
-mike
Attachment:
signature.asc
Description: Digital signature
