[MiNT] uname - was sending diffs to Apache ...

[Guido Flohr <gufl0000@stud.uni-sb.de>]

Hi everybody!

This is really an important issue, please read!

Various configuration scripts automatically detect the host system
type by running the "uname" program.  In the MiNT sector there seems
to be confusion about the correct output of "uname -s".  So far I have
come across three different versions:

	(a) MiNT
	(b) mint
	(c) FreeMiNT

In shell scripts you can't say "treat the uname output case-insensitive"
(well, at least it's not absolutely straigh-forward) so spelling is really 
a matter.

(a) As far as I can see it, this is the output of the original
K[GE]MD uname.  I think we should consider this the "reference" version
as it is probably the one that is spread the most.

(b) I don't know the origin of this one.  Is it possible that it came
along with the MiNT kernel?

(c) Hm, this is not really a frequently used uname.  Actually, I think
that I am the only person who uses it.

Background information: The "uname" program (at least the GNU version
of it) internally calls the MiNTlib function "uname()" and as far as 
I can see it, this function has returned "MiNT" for ages now (and
"TOS" for non-MiNT systems).  This means, if you build and
install an update of the GNU shell utilities, you will always end up
with an uname that returns "MiNT" unless the MiNTlib was 
modified.

I personally modified my MiNTlib to return "FreeMiNT" instead of 
"MiNT" for recent (>= 1.14) kernel versions because I considered
"FreeMiNT" and not "MiNT" the official name of our OS.  But this
difference is in fact easy to handle (although I don't care
to much about it, it was just an idea).  If the majority agrees with
my point of view, future versions of the MiNTlib could be adapted
without big hassle.

Remains the problem with the lowercase "mint".  IMHO this is incompatible
with most past versions of "uname" and it should be removed.

The thing is a little bit urgent:  Many GNU software packages use a 
standard script called config.guess (which calls uname) to determine
the host system type.  I want to send my patches for the GNU binutils
to the GNU folks in the next couple of days and this requires that
I also send in a patch to this config.guess script so that future
versions of it will correctly detect a MiNT host.

So, what version should be our "official" one? (a), (b) or (c) or
any combination of that?  This may seem pedantic but it's really
important.  My vote: Allow only "MiNT" and "FreeMiNT".  But if
really many people use a uname that returns "mint" we will have
to allow the lowercase version as well.

A related problem: At least some kernel versions contained a
uname program (is it still included in the kernel?).  I think
this should be removed.  A working MiNT system is based on
GNU utilities.  I don't think that it is a good idea then if the
kernel and the GNU shell utilities install concurrent versions
of uname.  We had better modify the MiNTlib so that future
versions of uname (as part of the GNU sh-utils) meet our ends.

Your votes please! ;-)

Ciao

Guido
-- 
http://stud.uni-sb.de/~gufl0000/
mailto:gufl0000@stud.uni-sb.de