Re: [PATCH 0/5] python2-variants

[Leo Famulari]

On Wed, Feb 10, 2016 at 10:21:17PM +0100, Ludovic Courtès wrote:
> Leo Famulari <address@hidden> skribis:
> 
> > These patches apply the new python2-variant system introduced in
> > 1be83341 to python2-{cryptography, jsonschema, wheel, requests}, and
> > then split acme into python-acme and python2-acme.
> 
> Neat!  Let’s see how many lines of code that saves us.  :-)

If we must measure something, that's a good metric ;)

> > As usual, I tested the changes to acme / letsencrypt by renewing the
> > certificates of my webserver [0].
> 
> OK.  I’m reviewing this series mostly superficially under the assumption
> that you’ve checked that things still build.

This exercise has made me wish I had a continuous integration system to
test the changes. Is it very difficult to set up a Hydra instance?

Thanks to Ludo and Efraim for reviewing this accidentally sloppy patch
series (I guess I was more tired than I realized!).

I took Efraim's suggestions to propagate the "simplifications" up the
graph of problematic python-2 variants, and I folded these commits into
the commit that fixed the package root of the problem of that part of
the graph, as in 57b7b8cd60. These are the "Take the result of
'package-with-python2' directly' log lines.

There were a few cases where I did not propagate the simplification
because the package in question would still be using the "old" system
afterwards [0]. Those packages still work as before [1].

I removed the changes to python2-oauthlib. There were unnecessary for
this patch set and will require the conversion of python2-unittest2.

[0] There are some of these in openstack.scm.

[1] python2-tempest-lib continues to fail to build from source, as
before.