A plea for Python 2.3

[Matt Mackall]

On Wed, 2009-05-27 at 20:38 -0400, Jeff Squyres wrote:
> I was testing the fix for http://www.selenic.com/mercurial/bts/ 
> issue967 today and discovered that the next release of Mercurial will  
> not support Python 2.3.
> 
> As a lurker on this list, I recognize that I may have missed prior  
> discussions and/or rationale for why 2.3 support is being dropped.   
> Have I missed the boat to plea for continued support for Python 2.3?

I'm afraid so. We've done a fairly extensive housecleaning of 2.3ism.
Going back to 2.3 at this point would be too much work, given that we'd
drop it before the bitter end of the RHEL4 lifecycle (2012) regardless.

> I ask because many of us enterprise types are using RHEL4 ("it ain't  
> broke, so they aren't fixing it") which ships with Python  
> 2.3.something.  And I will need 967 fix (support for pushing to https  
> through a proxy), which means upgrading to Mercurial <next> when it is  
> released. 

Well the trade-off inherent in getting seven years of stability from a
distro is that you don't get new features. 

But if you're already installing Mercurial on such machines, it should
be fairly straightforward to package a Python 2.6 install as well. The
easiest way is to make a build Python 2.6 that installs into /usr/local/
as python26. Then running python26 setup.py will adjust hg to use that
Python. Then you can tar up Mercurial 1.3 and Python 2.6 in one tarball
that you unpack in /.

> I, for one, might well be happy with reduced functionality with a  
> Mercurial >v1.3 running on Python 2.3.  For example, if I try "hg  
> some_cool_new_feature ...", hg may respond, "Sorry, I can't do this  
> because Python 2.3 doesn't support it.  Upgrade your Python to be able  
> to use this feature."  I don't know how feasible this is -- or if it  
> would cause unbearable ugliness in the hg code base -- but I thought  
> I'd at least throw out the idea.

The biggest reason to throw out 2.3 was to throw out compatibility
hacks. Further, no desktop operating system has had Python 2.3 for a
while (it's not even available in the Debian 4.0, released in 2007), so
testing against 2.3 means maintaining your own Python build. We were
regularly slipping things that broke 2.3 into the codebase because we
were unable to test against it. There aren't any new Mercurial features
that are dependent on 2.4, it's just about maintenance overhead.

-- 
http://selenic.com : development and support for Mercurial and Linux