RE: [MiNT] /proc

["Julian Reschke" <reschke@muenster.de>]

> > From: Julian Reschke [mailto:reschke@muenster.de]
> > Sent: Thursday, November 18, 1999 9:59 AM
> > To: gryf@hrzpub.tu-darmstadt.de; mint@gfanrend.fishpool.fi
> > Subject: RE: [MiNT] /proc
> >
> > > proposal of using a mint.cnf keyword (and an Ssystem() opcode in
> > > addition) for setting the path sounds very reasonable to me.
> > >
> > > I therefore ask again: Any objections agains this proposal?
> >
> > Yes:
> >
> > a) it is really not compatible enough with the other two
> > implementations,
>
> Please refresh my memory...

MetaDOS's SLB loader and MagiC will retrieve the SLB search list from the
current environment. So every process can have a different search path.

> > b) it doesn't the support the concept of different search paths for
> > different implementations.
>
> Now I'm confused... My understanding was that this was the searchlist for
> the *kernel*, and that it will only be used when the client didn't use an
> absolute path when opening the library. Why would clients need this? Sure,

Right.

> you can read it so you can search it yourself if you can't find
> anything in
> the custom search-path, but why do this if the kernel evaluates SLBPATH
> itself?

Well, it doesn't. And even with the recent proposal, the setting would be
system-wide.

> > b) has anybody thought about a way how to retrieve a list of
> > processes that
> > currently use a particular SLB?
>
> Why not use an Ioctl on /proc? Oh, I forgot, that will probably compromise
> security and we will have to remove the multituser-features
> because they're
> useless anyway :-(

That would be the elegant way (and the MiNT developers should like it as it
wouldn't work under MagiC :-).