Subject: Re: Solaris MACHINE_GNU_PLATFORM
To: None <email@example.com>
From: Takahiro Kambe <firstname.lastname@example.org>
Date: 03/24/2003 00:03:07
In message <20030323142749.GA11567@desk06.rd.bbc.co.uk>
on Sun, 23 Mar 2003 14:27:50 +0000,
Jonathan Perkin <email@example.com> wrote:
> > I think that patching config.sub is better.
> Who's going to go around and add CONFIG_SUB_OVERRIDE=YES to every single
> Makefile that needs it then?
Yes, it is a problem. But this arise from GNU configure's convention
for Solaris system.
> > o GNU configure itself generate fuzzy way system platform. I've
> > checked on SunOS 5.8,
> > (1) No argument configure results "sparc-sun-solaris2.8".
> Sure, but we always specify an argument, so this is a moot point.
> > (2) Giving -host=sparc-sun-solaris results "sparc-sun-solaris2".
> > (Yes, this is config.sub's result.)
> That's exactly my point. Changing a single variable to match what every
> package currently does is IMHO a lot simpler and neater than changing
> the vendor script to match our "broken" behaviour, then have to fudge
> that behaviour in a ton of Makefiles.
Another side is consistency for our pkgsrc system. If we use
"solaris2", then we should use "Solaris2" as value LOWER_OPSYS's name.
BTW, is there any package that contain "sparc-sun-solaris2" in
installed files and has problem on GNU configure argument?
> Ideally we'd distinguish between every single OS release so that we can
> preserve binary compatability,
I don't think so. Instead of distinguish those OS release, we
distinguish with functionality. As for Solaris, SunOS 4.X based
system and SunOS 5.X based system are basically differnet system, not
simply release number.
Last, thanks for your comment.
Takahiro Kambe <firstname.lastname@example.org>