[PKG-Openstack-devel] Accepted python-eventlet 0.20.0-1 (source all) into unstable

[Thomas Goirand]

Hi Ondrej and Daniel,

I'd like to explain what happened first (ie: the course of event that
lead me to this), and say sorry, then a bit more below it.

First, I'm trying to build Neutron in Sid, and I get this issue:

   debian/rules override_dh_auto_test
make[1]: Entering directory
'/home/zigo/sources/openstack/pike/exp/neutron/build-area/neutron-11.0.1'
pyversions: missing X(S)-Python-Version in control file, fall back to
debian/pyversions
pyversions: missing debian/pyversions file, fall back to supported versions
py3versions: no X-Python3-Version in control file, using supported versions
PYTHONPATH=build/* ./run_tests.sh
NOTE: this script (run_tests.sh) is deprecated and will be removed in
the Pike release.

Running ` python setup.py testr --slowest --testr-args='--subunit  '`
running testr
Guru meditation now registers SIGUSR1 and SIGUSR2 by default for
backward compatibility. SIGUSR1 will no longer be registered in a future
release, so please use SIGUSR2 to generate reports.
Could not update option "parse_argv"
Exception in thread Thread-1:
Traceback (most recent call last):
  File "/usr/lib/python2.7/threading.py", line 801, in __bootstrap_inner
    self.run()
  File "/usr/lib/python2.7/threading.py", line 754, in run
    self.__target(*self.__args, **self.__kwargs)
  File "/usr/lib/python2.7/multiprocessing/pool.py", line 327, in
_handle_workers
    while thread._state == RUN or (pool._cache and thread._state !=
TERMINATE):
AttributeError: '_MainThread' object has no attribute '_state'

This happens running testr directly, using /usr/bin/ostestr, or even
using ./run_tests.sh. Because of the above, I thought this was Eventlet
related, so I thought the issue was Eventlet not being up-to-date. Then
I looked into the upper-constraints.txt, and saw that we're supposed to
run with Eventlet 0.20.0. Then I had a look on pypi, and saw only
0.20.1. Then I had a look at the (build-)depends for Eventlet, and saw
it was only used by 38 OpenStack packages, both runtime and build time.
I therefore thought it was fine to move back the package to OpenStack,
since it wasn't used anywhere.

Ondrej, if you feel bad about it, I'm sorry. I don't want to let you go,
and I'm ok to "revert" the upload. Unfortunately, IMO, "revert" here
can't mean using another version than 0.20.0. That's the version used by
upstream Pike, and we should keep this version. If you don't agree,
please let me know why.

Since 0.20.0 is already uploaded to the archive, if we're to use
pristine-tar, then what should happen is importing the currently
uploaded orig.tar.gz in the Git. I don't think that's really a problem.

If you're ok with such a plan, Ondrej, then please do it. The thing is,
there's currently an RC bug opened because Eventlet is slowly moving to
Python 3, and the currently uploaded version can't even compile. I've
pushed patches to fix the issue in the OpenStack git. I'm ok to upload
it, though I would very much prefer to have peer review of these
patches. Especially the one modifying the regexp. I haven't used much
regex in Python, though my understanding is that the r'' form of strings
prevent the interpretation of the escape sequences. Though rb'' is
needed for Python 3, it crashes the Python compiler in Python 2. If
setting only b'', it compile on both, but then unit tests are failing in
Python 3, because the regex becomes bad. My patch therefore removes the
r, and add more slashes to avoid interpretation by Python. That's the
part which I would like peer review on. Lamby wrote that he's not sure
about this patch...

On 11/13/2017 09:51 AM, Daniel Baumann wrote:
> On 11/12/17 22:40, Ondrej Novy wrote:
>> ALL non OS upstream python packages should be moved to DPMT.
> 
> right, that was the consensus, with no ambiguity.
> 
> thomas:
> 
>   * ondrej just did what was on the todo (see gobby): moving those
>     packages (in that case, python-eventlet) from pkg-OS to DPMT. i see
>     nothing wrong with that. also the pybuild thing we have discussed
>     clearly.

I actually see in that list of packages some that have OpenStack as
upstream, as per what I wrote on IRC:

- migrate
- python-cliff
- python-cloudfiles
- python-django-openstack-auth
- python-dogpile.cache
- python-tenacity
- python-pyghmi
- python-rtslib-fb
- python-wrapt
- python-yaql

Are you guys ok to see the above moved to /git/openstack/libs ?

>   * do you disagree with having python-eventlet in DPMT?

I don't think it's a good idea, as it's used so much in OpenStack. I'm
ok with everything else, if we consider the above list of packages though.

>> You need to communicate FIRST.

The message on IRC was an attempt to discuss the mater with you, with
the thoughts that things could be reverted anyway, so it wouldn't mater
much. The thing is, I'm really closed to have everything finished, and I
want to get this done ASAP.

> while I appreciate the "getting stuff done"-mode thomas is in, this
> really needs fixing if this should become a team, hence the emphasis: if
> you change *anything* that is deviating from what we decided (and
> therefore is noted down in the gobby stuff), *please* write a mail/talk
> on irc *first* and give *reasonable* time to answer.

Yeah. I just want to do things too fast, sorry for that. Please don't
make it such a drama, uploads can be reverted.

> there's much stuff that needs working, certainly a little "waiting"
> doesn't hurt working on other parts in the meanwhile.

As I'm mostly the only one working on this, yes it does hurt. I thought
that Eventlet was blocking, and wanted to try fast.

>> Sorry, but I'm really pissed off
> 
> understandably, however, can we all go back to "shit happens, let's do
> it better next time"?

That's exactly my thinking too, and it was my thinking that, when doing
it, if anyone wasn't happy of what I was doing, things could be reverted
anyway.

Now, I've been a bit frustrated that everyone promised to do some work,
and that I've done mostly all again. Ondrej, I would have loved to see
more of your uploads. Are you swamped with work, just like Daniel is?

Cheers,

Thomas Goirand (zigo)