[Numpy-discussion] help from OS X 10.5 users wanted
Wed Oct 13 14:48:12 CDT 2010
2010/10/13 Christopher Barker <Chris.Barker@noaa.gov>:
> On 10/12/10 2:00 PM, Vincent Davis wrote:
>> I could do 10.4 also. For me doing this is maybe best described as a
>> curiosity. But my understanding was that some where having trouble
>> install on 10.5 with binaries made on 10.6. So if I can learn
>> something and help others that is great.
> Well, I appreciate the effort.
Thanks a lot.
> Ideally, I think the Mac numpy binaries should match the python.org
> python binaries:
> Python 2.5: 32-bit PPC/i386 for Mac OS X 10.3.9 through 10.6
> Python 2.6: 32-bit PPC/i386 for Mac OS X 10.3.9 through 10.6
> Python 2.7 32-bit PPC/i386 for Mac OS X 10.3.9 through 10.6
> Python 2.7 PPC/i386/x86-64 for Mac OS X 10.5 or later
> Python 3 ??? (similar I think)
> Supporting 10.3.9 may be a bit much these days, so it's probably easiest
> to build on a 10.4 machine, except for the 2.7 64bit build on 10.5.
We have to decide about how to install Python on the machines. It
aren't VMs, as one cannot virtualise workstation Mac OS X using
The most straightforward way would be to just install the python.org
binaries available. Note that Python 2.5 *must* be compiled on Mac,
there is no binary on python.org.
I would like to discuss the feasibility of compiling numpy against a
host-built Python built from the sources from python.org. In this
case I would compile including the x86_64 arch, if it is available on
10.5 at all? Could both users of a self-built Python (e.g. 2.6 on
10.6, this would be my case), use this precompiled numpy then, as well
as users using the python.org binaries? It would enable users to use
x86_64 arch without self-compiling.
Is there a check if the python.org installer was run and is registered
in the computer's package database?
Ralf, what is the advantage of hard-coding the destination directory
over running a setup.py script as postflight in the package installer?
(I understood that background-compilation is really not such a bright
idea.) The hidden setup.py would be compatible with all installation
locations of the user's Python.
> In theory, it should all be buildable on 10.6, but I know I've struggled
> with doing that sort of thing -- but maybe the macpython folks could help.
I would like to concentrate first on the 10.5 machine of Vincent.
I had the attitude to set up an openvpn first, because I personally
don't like to use 3rd party servers when transferring passwords. The
VPN is up and running, and the authentication is PKI based.
This are the open questions:
1) Installation scheme of the different Pythons on the packager
machine (binary or host-compiled)
2) Layout of the installer (hardcoded or hidden setup.py)
More information about the NumPy-Discussion