[Bf-committers] solving the issues of switching to python 2.4 on Os X for next release

Willian Padovani Germano wgermano at superig.com.br
Thu Aug 25 21:23:42 CEST 2005


Stephen Swaney wrote:
> On Wed, Aug 24, 2005 at 08:15:12PM -0300, Willian Padovani Germano wrote:
>>I guess we should officially stay with 2.3 -- and use it in all 
>>platforms which have issues with 2.4.
> I would respectfully disagree with this.  Most of the world has moved
> to 2.4.

Hi Mr. Stephen :),

As you say this is an issue we will probably have to deal with again and 
again, in transition periods like this (which can take a long time to 
pass). I don't have "the" solution, just chatting to see if we can find 
something.

Clearly we must try hard to keep Blender as small and easy to install as 
possible, but there's no clear winner (Py 2.3 or 2.4) in this front when 
we consider all platforms involved.

As you know, Python is evolving, with more features, more speed, 
important bugs fixed (including security ones, check their site) and 
it's important to support Blender users who are also fond of Python -- 
they write the scripts, after all. But for some systems it's not 
progressing well. Don't know what the problem is: if people help 
reporting these bugs, if there are Py developers to fix them for Irix 
and FreeBSD, etc.

> I think the best way to deal with this is the same way we deal 
> with Windows now - include our own distribution of Python 2.4
> with enough modules to run our distributed scripts on our 'exceptional'
> platforms.

Yes, taking control of the situation like done under Win might be the 
way to go. But this doesn't solve the troubles for systems where 2.4 is 
buggy, as you said.

And under Linux and OS X, for example, a mismatch would force users to 
download new versions of 3rd party extensions (gui toolkits, image libs, 
etc) that they already had installed, to be able to use them with 
Blender. To have these versions only for Blender and not break anything, 
they wouldn't be able to use default installs, they'd need to unpack 
them in Blender's py modules default or user dir.

Freeing each platform manager to decide which version is better for that 
platform, 2.3 or 2.4 (*), seems to me the best way to go now and in 
future updates.

But then we need to define an "official" version. As I wrote, not to 
mean Blender must be compiled with that, but just so that script authors 
know what is safe to aim for when they are coding for the community. 
Core Python doesn't change *that* much (mostly it is upgraded with more 
features), so this is ok.

(*) 2.2 might not compile anymore with our sources, guess Hos mentioned 
he had tried. Maybe we can fix something to support it again, but 
guaranteeing scripts in general will work with 2.2 is not a good idea 
(many will, though).

-- 
Willian


More information about the Bf-committers mailing list