On Tue, 10 Jul 2001, Andreas Beck wrote:
> > > should we use <inttypes.h> with a fallback?
> > This should be discussed before making a decission, because changing
> > all the datatypes through the whole code is going to deep in that
> > state. We should address this task _after_ the final release, ok?
>
> Basically I thought autoconf would take care of that. Seems we
> need to tell autoconf about it ?
>
> Another thing to consider: I think we once had ggi_s32, ggi_u32
> and friends.
>
> Does someone remember, why these were abandoned ?
I have never seen them. So they must be abandoned, before I came to
GGI...
> They had the huge advantage of causing no namespace collisions
> ...
>
> > > should the modules be more unifor? Some modules use FOOCONFFILE and
> > > FOOCONFDIR and pull them in in foo/init.c while other modules only use
> > > FOOCONFFILE. Some other differences as well that I have seen. Do we
> > > want the modules to be uniform?
> >
> > That's true. Andy: What is more recent? The only FOOCONFFILE usage or
> > the combination of FOOCONFFILE and FOOCONFDIR?
> >
> > My guess is the last one, because of the possibility of overriding
> > the internal confdir path under Win32 as this lines from
> > (libgii/gii/init.c) shows:
>
> Yes, I suppose this split was done to support that braindead
> stuff for windows:
>
> > const char *giiGetConfDir(void) {
> > #ifdef __WIN32__
> > /* On Win32 we allow overriding of the compiled in path. */
> > const char *envdir = getenv("GGI_CONFDIR");
> > if (envdir) return envdir;
> > #endif
> > return giiconfdir;
> > }
>
> We should probably place a big fat warning near there to _NOT_ allow that
> under OSes that have real user management.
So we should do the split in all other libs, too?
> As libggi is dynamically loading targets, the configuration is a
> security-critical file, if suid-programs are to be made (which
> should be possible, though not recommended).
>
> Otherwise someone could just override the config for such a program to make
> it execute arbitrary code (I envision display-make_me_root :-).
I see.
CU,
Christoph Egger
E-Mail: [EMAIL PROTECTED]
