[Scons-dev] Python 3 strategy

[William Blevins]

On Mon, Jan 25, 2016 at 9:39 AM, Russel Winder <russel at winder.org.uk> wrote:

> I am having difficulty making a decision…
>
> The earlier Python 3 branch is founded on using six. At the time a good
> decision. Now however we have agreed that 2.7 is the base version and
> thus future rather than six is the better tool for Python 3. This
> brings into doubt the python3-port branch as a good base of operations.
>
> William has done some work changing the python3-port branch, as have I,
> all based on the massive previous work. Despite all this effort I
> wonder if we have the right base to progress further. Maybe we have and
> carry on, but now is the time to stop before much more effort goes in
> if this is not the way forward.
>
> The alternative is to abandon the current python3-port and start again
> from default based solely on future.
>
> Instead of having a separate branch, let us work solely on default,
> turning it first into a Python 2.7 only thing using futurize to prepare
> the ground. This is really an extension of what is currently happening
> of course, but with a turbo charger. Let us insist that the SCons
> default branch is "futurize -1" with all __future__ imports in place,
> so that all Python 2.7 code is using Python 3 semantics where
> possible.
>

I think this is reasonable, but I don't know what effort has been made
prior to now.


>
> This can happen within the current infrastructure effectively
> unchanged. Obviously all CI should be green with each changeset. I
> suspect it will not run under Python 3 in this form.
>
> When this is done, we can add the "futurize -2" with the dependency on
> future, obtained not by vendorizing but by package management install.
> This should leave the Python 2.7 CI still green – assuming all the CI
> servers have future installed. (With Codeship and Drone this is handled
> by using a requirements.txt file in the usual pip-ish way.) Then the
> work of getting a Python 3 CI green can happen, with the Python 2.7 CI
> always green.
>

I'd prefer to make a solid attempt to not have a dependency of future if
possible.

Once it is added, chances are it will never be refactored out...


>
> Of course even though the Python 2.7 tests are red, and the Python 3
> tests do not run, python3-port on Python 3.5 and 2.7 actually works.
> This is the opposite side of the coin, why change strategy now?
>

Does SCons appear to function in the branch? I haven't tried actually
compiling anything?

There aren't that many 2.7 tests failing. It's less than 50 on my Linux
configuration, and I have most of the tools setup. Not sure what the other
platforms look like...


>
>
> --
> Russel.
>
> =============================================================================
> Dr Russel Winder      t: +44 20 7585 2200   voip:
> sip:russel.winder at ekiga.net
> 41 Buckmaster Road    m: +44 7770 465 077   xmpp: russel at winder.org.uk
> London SW11 1EN, UK   w: www.russel.org.uk  skype: russel_winder
>
>
> _______________________________________________
> Scons-dev mailing list
> Scons-dev at scons.org
> https://pairlist2.pair.net/mailman/listinfo/scons-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://pairlist2.pair.net/pipermail/scons-dev/attachments/20160125/6f99a490/attachment.html>