[AstroPy] Deployment and packaging

[Peter Erwin]

On Jun 16, 2011, at 8:00 PM, Lisa M Winter wrote:

> Peter, it looks like you are using Mac OS 10.3?  I hadn't found any binary installations that work with Snowleopard... and now of course we will need to worry about Lion.  Then there was the issue of how this all works on a 64-bit machine.
> 

Ah, I see the problem. The naming scheme for those installer disk images is confusing; the "10.3" 
really means "10.3 and more recent" -- so they're meant to be installed on Macs running OS X 10.3, 
10.4, 10.5, *or* 10.6. They *are* meant to work with Snow Leopard (and earlier versions).
(I'm using 10.6.7 on my two main laptops, with an older laptop that's still running 10.5.)

The 64-bit issue *is* a bit of a problem; for the time being, I'm keeping all my Python 2.x installations
32-bit only, at least on my Macs. (In the Mac OS X world, the 32-bit versions of Python work just fine on 64-bit
*machines*; you only run into problems if you try to, e.g., install a 64-bit version of Numpy into a 32-bit-only
version of Python.)  So I'd steer clear of the "Building 64bit NumPy/SciPy/PyLab on Snow Leopard" idea
unless you *really* need to access > 4 GB of memory within your Python session.

At www.python.org, you have the option of 32-bit-only versions of Python 2.7 -- which can be installed
on Mac OS X 10.3 through 10.6 -- or a "fat" 32-bit/64-bit version, which can only be installed on 10.6.
Personally, I'm sticking with the 32-bit-only version.


> There is a tutorial on: http://blog.hyperjeff.net/?p=160, which I had tried.  However, I ended up with errors in several points going through this.
> 
> I think that the average user is likely to see a website like this and give up at some point through the process if they get too many errors and continue to use whatever package they are used to (e.g., IDL).
> 
> Also, I agree with Brad, matplotlib was difficult to install, until you use a distribution like Enthought that just gets everything working.  I didn't like using Scisoft, however, because it installs multiple copies of some applications I've already installed (like iraf/pyraf).  Plus, I'm not sure all the distributions are up to date.

I've added some text to the astropy-distribution wiki page concerning Scisoft. It's *not* very up-to-date;
the official ESO version (Fedora Linux only) only has Python 2.5.4 (plus numpy 1.3, scipy 0.7, et.c), while the
unofficial Mac OS X version now has Python 2.6 (though I'm not sure how you're supposed to know this
without downloading the entire installer and going through its contents, like I just did...).

I share your dislike of Scisoft; we've had some problems with the Linux version on our local
Linux boxes, having to do with outdated libraries.

   -- Peter

=============================================================
Peter Erwin                   Max-Planck-Insitute for Extraterrestrial 
erwin at mpe.mpg.de              Physics, Giessenbachstrasse
tel. +49 (0)89 30000 3695     85748 Garching, Germany
fax  +49 (0)89 30000 3495     http://www.mpe.mpg.de/~erwin