Re: [gdm-list] [PATCH] fix gcc warnings

[Brian Cameron <Brian Cameron Sun COM>]

Ludwig:

> [reply moved here because I think you are referring to the patch in
> this thread]
>
> Brian Cameron wrote:
> > Ludwig Nussel wrote:
> > > Brian Cameron wrote:
> > > > Anyway, I did not apply a few of your changes to daemon/verify-pam.c
> > > > since they caused warnings to show up in Solaris.  If you want to
> > > > rework that patch so that the current code is surrounded by
> > > >
> > > > #ifdef __sun
> > > > ...
> > > > #else
> > > > your code
> > > > #endif
> > > >
> > > > Then I would accept it, so it works on Solaris and Linux.
> > > Which parts? I don't have Solaris so I can't check myself.
> > Your previous patch only had five 1-2 line modifications.  The
> > existing code should go in the #ifdef __sun section, your modification
> > should go in the else.  I will test your patch on Solaris if you
> > send it to me.
>
> I assume you are referring to the following patch. If that causes
> warnings on Solaris then the prototypes for conversation function
> and pam_set_item must be different. On Linux it's

Yes, I was referring to the changes in daemon/verify-pam.c and yes
the conversion function definition is different on Solaris and Linux.

> int (*conv)(int num_msg, const struct pam_message **msg, struct pam_response **resp, void *appdata_ptr);
> int pam_get_item(const pam_handle_t *pamh, int item_type, const void **item);

Here are the definitions in Solaris /usr/include/security/pam_appl.h. 
Note the minor differences:

struct pam_conv {
        int (*conv)(int, struct pam_message **,
            struct pam_response **, void *);
        void *appdata_ptr;              /* Application data ptr */
};

extern int
pam_get_item(
        const pam_handle_t *pamh,       /* PAM handle */
        int item_type,                  /* Type of object - see below */
        void ** item                    /* Address of place to put 
pointer */
                                        /*   to object */
);

> I'd rather live with the warnings on either platform instead of
> adding ugly #ifdefs

I'm happy to add #ifdef's if anybody on Linux wants the warnings to go 
away.  Or if someone has a better idea of how to make the code compile
without warnings on all platforms.

I don't think warnings about a missing "const" would cause any
problems on any platform, though.

Brian