[rfc] time to remove LANG=C run from make check

[Vincent Ladeuil]

>>>>> "martin" == Martin Pool <mbp at canonical.com> writes:

    martin> I think it's time to bite the bullet and close bug
    martin> 386180 by not running the test suite a second time
    martin> with LANG=C from 'make check'.  This would cut in
    martin> half (almost) the pqm latency.  Now that we have a
    martin> working buildbot and vila's reporting the results in
    martin> the release process, we can do a test for the ascii
    martin> locale (or any other locale) asynchronously with
    martin> merges.

And successfully and quickly caught two minor regressions in the
suite and one in the windows installer builds (related to tbzr).

    martin> If it turns out we get more encoding-related bugs we
    martin> may roll it back.

None of the above regressions was encoding related.

    martin> Eventually it would be nice to synchronously but in
    martin> parallel test across all platforms and environments.

I can't handle the load alone :) I can give details if you want,
but in summary, I can't run more than once a day. And bisecting
seems fully doable so far.

    martin> If no one screams I'll land this to trunk (but not
    martin> 2.0 of course) in a couple of days.

I scream.


...



I mean, I scream yes :)

  Vincent