Chris Tinney is trying to get the SSC working on his Intel Mac. I told him
to send along any questions to this list.
However, it seems that he's getting different results out of
cvs.starlink.ac.uk and cvsweb.starlink.ac.uk. Are they out of sync? I
haven't looked myself...
---------- Forwarded message ----------
Date: Tue, 13 Jun 2006 07:14:19 +1000
From: Chris Tinney <[log in to unmask]>
To: Brad Cavanagh <[log in to unmask]>
Subject: Re: Any ideas on configuring CVS Starlink for intel mac
A lot of the old build also worked using the Rosetta system as well.
But I'm trying to get an intel specific system built, so I can link against
the libraries.
Incidentally, an intel machine is not really critical for testing the intel
building process ... you can start doing it on a powerpc machine, but force it
to use gcc4 and g95 or gfortran. They seem to produce most of
the problems, rather than being an intel machine.
What I have noticed is that when I followed the instructions for
doing a checkout from the CVS server, (ie commands like
cvs -d :pserver:[log in to unmask]:/cvs
I seem to get a system
that is significantly out of date compared to that on the
cvsweb interface at
cvsweb.starlink.ac.uk/cvsweb.cgi
I found the reason I couldn't even get started ... the top level
Makefile.dependiencies was essentially empty. Once I
downloaded a Makefile.dependiencies from the cvsweb, I
got something that would start building (at least it
make starconf!). It crashed building TCL, when I found that
the cvsweb server has a whole tree for building TCL on MacOSX,
which was not present at all in the checkout I had.
So I think the cvs.starlink.ac.uk may be stale.
Cheers,
Chris T.
On 13/06/2006, at 6:44 AM, Brad Cavanagh wrote:
> Hi Chris,
>
> On Sun, 11 Jun 2006, Chris Tinney wrote:
>
>> just a short note ... I got an intel Mac recently, and wanted to compile
>> some of the starlink libraries (to use with perl mostly, like the
>> NDFlibraries).
>>
>> I tried checking out the CVS version of starlink, and thought I'd try
>> building it. But it fails miserably at the first hurdle, not even being able
>> to make and install the starlink version of autoconf, configure, etc (ie
>> starconf).
>>
>> I presume this is because the bootstrap process is finding an i386 processor
>> on a Darwin OS, and falling over as it doesn't know what to do. A
>>
>> Do you have any ideas on what my first step would be in getting the
>> autoconf system bootstrapped up?
>
> No idea, actually. I suspect that your suspicions are correct, because we
> haven't had access to an Intel Mac yet and haven't been able to run through
> the whole process ourselves.
>
>> If not, don't worry ... I'm sure its bnot your job to know, and I'll bumble
>> along myself.
>
> You can ask on the stardev mailing list ([log in to unmask], you may have
> to visit http://jiscmail.ac.uk to sign up first, as I don't know if it's open
> to unsubscribers or not). They'll have a much better idea as to how you
> should progress.
>
> I'll be getting an Intel Mac sometime in the near future (before the end of
> July, I'm told) so we'll be able to fix things to get them working properly.
>
> I don't know if it'll help you or not, but I've had reports that the build I
> just made works on Intel Macs. The disk image is available at
> http://www.jach.hawaii.edu/~bradc/starlink/Starlink-OSX-20060502-beta.dmg I
> don't know if libraries can be linked against, but the applications seem to
> work, which suggests some degree of success. :-)
>
> Cheers,
> Brad.
|