[Pythonmac-SIG] Building 2.5 on SL - configure-2.5.4-for-10.6.patch (1/1)

[Ned Deily]

In article <48C14E26-E790-47C2-BD44-FC65159D8D03 at zope.com>,
 Zvezdan Petkovic <zvezdan at zope.com> wrote:
> On Oct 7, 2009, at 3:32 AM, Ned Deily wrote: 
> > In article
> > <7a9c25c20910062100i25e0bbe0jb2c37e9025812f93 at mail.gmail.com>,
> > Stephen Hansen <apt.shansen at gmail.com> wrote:
> >> I'm trying to build a "private" instance of Python 2.5.4 on Snow  
> >> Leopard,
> >> and am running into a problem. [...]
> >
> > A-hah!  The problem is that Snow Leopard is also Darwin 10.0 and it
> > turns out that an important test in the configure script doesn't  
> > handle
> > the two-digit major version number correctly.
> >
> > The attached patch should fix the problem.  Or simply manually edit
> > configure to replace the line:
> >  FreeBSD/4.* | Darwin/[6789].*)
> >
> > with the three lines:
> >  Darwin/[012345].*)
> >      ;;
> >  Darwin/*)
> >
> > The same problem affects all of the current Pythons which is why you  
> > saw  it on 2.6.3 as well.  I'll see that it gets taken care of.
> 
> First, let me state that your patch is probably in agreement with the  
> intended purpose of that case in the configuration.
> 
> But, just for the record, the setting itself does not cause the  
> problems for Python 2.6 if it's built for Mac OS X 10.6 target.

I didn't say anything about building for a 10.6 target; this was to fix 
building for a 10.4 target on 10.6.  But you are right: no change is 
needed to build on 10.6 for a 10.6 target.  Nor is there a problem with 
building for a 10.5 target.  10.4 target is the problem child.

But, after posting this last night, I realized that the solution in the 
patch is not correct (although it fixes the immediate problem) because 
the original test in configure is flawed.  What it's doing is basing the 
decision about the _XOPEN_SOURCE stuff on the os level of the build 
machine (its darwin version); what it should be doing is basing it on 
the value of the deployment target.  That's probably true for other 
darwin version tests in configure but I didn't see any other glaring 
10.6 gotchas on a quick first glance.  I plan to submit a different 
patch after getting some time to test the various configurations.

-- 
 Ned Deily,
 nad at acm.org