[Numpy-discussion] help from OS X 10.5 users wanted

Russell E. Owen rowen@uw....
Fri Oct 15 16:44:12 CDT 2010


In article 
<AANLkTikgOu-ics2cgNnpRoZcBgLdXnXfd+Mgbok71-fe@mail.gmail.com>,
 Friedrich Romstedt <friedrichromstedt@gmail.com> wrote:

> 2010/10/15 Christopher Barker <Chris.Barker@noaa.gov>:
> > On 10/15/10 10:54 AM, Russell E. Owen wrote:
> >> I have a 10.4 Intel machine I keep around
> >> for the sole purpose of building backward-compatible binary installers
> >> for PIL and matplotlib. If help is still wanted I'm happy to build numpy
> >> as well.
> >
> > I'll let Ralf and Friedrich and Vincent respond, but that sounds like a
> > great option.
> 
> Thanks for waiting for us!  Here is our status:
> 
> Vincent is currently setting up the machine with a crucial
> carbon-copy-clone, reformat, and ccc back.  This is because we had
> only around 50 MB left in the end on the partition, and according to
> Vincent enlarging the partition wasn't possible for some reason be
> both don't know.  Well, technicalities.
> 
> We're having to set up py2.6 & py2.7, what shouldn't be an issue, and
> will have to compile mpl for both of them since the build process
> should not change Python version while its running, i.e., when
> building the py2.6 dmg, the docs will be built using py2.6, and when
> py2.5 dmg, then py2.5, that's what I mean.
> 
> Also LaTeX is needed, Russell, as you might know.  MacTeX-2010 is
> about 1.6 GB download.

Interesting. I had no idea but I can install that.

> Here come now the interesting facts:
> 
> 1)  Some tests of numpy failed in 2.0.0.dev.  When the machine is
> running again I can send the logs.  All some strange-looking typecode
> string tests with dtype('...') iirc.

I have no idea what to make of this.

> 2)  I noticed that the paver at some late point tried to switch from
> py2.5 to py2.6, what is rather strange to me.  I must have a look
> where precisely the build failed for this reason.  py2.6 is the
> DEFAULT_PYTHON_VERSION (iirc) in pavement.py, and changing it to 2.5
> fixed it.  Strange.
> 
> 3)  I found no v1.5.1 tag yet (yesterday).  Will the HEAD become 1.5.1?
> 
> Here a comparison of our two systems (Russell's and our's):
> 
> *  We will have 10.4 (?), 10.5, 10.6 available on the same machine
> with vpn access for everyone who wants a cert.

Mine has 10.4 and 10.5 available (via an old two-partition external hard 
drive dedicated to building python packages). It has no public access 
and I'd prefer to keep it that way. It's also off most of the time and 
there are times when I will not have it at all (since I have it on 
long-term loan).

> *  But we need time to set it up properly.  We're unwilling to do
> half-baken things, so I agree that Vincent it installing 10.6 right
> now (I just got the message), but time is rare this weekend.
> 
> *  So my suggestion would be, Russell, if you could do the build more
> easily then we can, just feel free, I was hoping Vincent and me would
> get the credits though ;-), but first it must succeed on time.
> Vincent, what do you think?
> 
> *  For future, I would prefer Vincent's machine.  We have dyndns and
> the machine can be dedicated for building this stuff.  If we get a
> 10.4, we have 10.4 to 10.6 all together on a single place, and we
> could do with it whatever we feel like.
>...

It sounds like you have things under control. I propose to leave it to 
you. It sounds like you are doing a great and very thorough job. If you 
need a confirming build for some reason I'm happy to do that.

Are you also building matplotlib then? If you are then please install 
ActiveState Tcl/Tk so the matplotlib will be compatible with 3rd party 
Tcl/Tk (as well as Apple's built-in Tcl/Tk). The best version for 
python.org's Python 2.6 is 8.4.19. I'm betting the same is true of the 
32-bit Python 2.7. I'm not sure what version of Tcl/Tk the 64-bit 
version of Python 2.7 was built against, but that's the one to match.

I'm hoping to build PIL and matplotlib for Python 2.7 in the next month 
or so, depending if I can figure out how to do it. (The 32-bit version 
should be easy; it's the 64-bit version I'm worried about).

-- Russell



More information about the NumPy-Discussion mailing list