On Fri, Sep 11, 2009 at 10:48:10AM -0400, Stephen Gallagher wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 09/11/2009 10:29 AM, Sumit Bose wrote: > > On Fri, Sep 11, 2009 at 09:07:01AM -0400, Stephen Gallagher wrote: > > On 09/11/2009 06:13 AM, Sumit Bose wrote: > >>>> Hi, > >>>> > >>>> with this patch the client sends its PID to sssd. This is at least > >>>> needed by the krb5 provider if the client PID should be part of the > >>>> credential cache file. > >>>> > >>>> bye, > >>>> Sumit > >>>> > >>>> > >>>> ------------------------------------------------------------------------ > >>>> > >>>> _______________________________________________ > >>>> sssd-devel mailing list > >>>> sssd-devel@lists.fedorahosted.org > >>>> https://fedorahosted.org/mailman/listinfo/sssd-devel > > > > I'm not a huge fan of the assignment in extract_uint32_t. > > > > Isn't > > *var = (uint32_t)body[*c]); > > the same, or does that throw typecast warnings? > > > > > >> no, this is not the same, this takes the byte a body[*c] and cast it to > >> uint32_t, because body is uint8_t. > > > > If the latter, then I'd prefer that it at least be written as: > > *var = *(uint32_t *)&body[*c]; > > > > Using array notation gives an implication I don't like. > > > > Same comment about array notation use in add_uint32_t_item. > > > >> My preference would be > > > >> *var = *(uint32_t *)(body + *c); > > > >> but the other notation was used in the code before I started using it in > >> the PAM related parts. I think it would make sense to agree on a > >> notation and change the whole code. Would you mind filing a bug? > > I agree. I'll file a bug on this. > > > > > > > > > > > Also, don't we need to bump the protocol version if we're adding new > > mandatory arguments? > > > >> Good point. Shall we make it mandatory? Currently it is only needed if > >> you want to have the pid in your ccache file (which is not the default). > > I think it should be mandatory in the protocol, with the option to just > pass 0 if it's not supported by the client (for compatibility) >
Most of items are not mandatory at the protocol level. If e.g. the remote host is not known to the client it is not sent to the server and the server complains if he really needs it, e.g. the user name. I haven't put a check like 'if cli_pid==0 do not send to the server' because as getpid(2) says "These functions are always successful.". On the server side cli_pid is 0 if the client does not send a PID item. I think the way it currently works is the way your are expecting it to work. bye, Sumit > > > > > > > -----BEGIN PGP SIGNATURE----- > Version: GnuPG v1.4.9 (GNU/Linux) > Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/ > > iEYEARECAAYFAkqqYyMACgkQeiVVYja6o6ONIACgnyqn3oDuS0UtvBJQKzF7CSM2 > pKUAoJ8jDGhZqFfY8neejo++x80uH0Q/ > =4+35 > -----END PGP SIGNATURE----- > _______________________________________________ > sssd-devel mailing list > sssd-devel@lists.fedorahosted.org > https://fedorahosted.org/mailman/listinfo/sssd-devel _______________________________________________ sssd-devel mailing list sssd-devel@lists.fedorahosted.org https://fedorahosted.org/mailman/listinfo/sssd-devel
