Open Stack

On 09/24/2013 07:27 AM, Sean Dague wrote:
> On 09/24/2013 06:44 AM, Nikola Đipanov wrote:
>> On 24/09/13 10:15, Gary Kotton wrote:
>>> Hi,
>>> Anyone know the root cause of:
>>>
>>> 2013-09-24 06:47:01.670 | Cleaning up...
>>> 2013-09-24 06:47:01.670 | No distributions matching the version for
>>> pyparsing>=2.0.1 (from
>>> cliff>=1.4.3->python-neutronclient>=2.3.0,<3->-r
>>> /home/jenkins/workspace/gate-nova-pep8/requirements.txt (line 25))
>>> 2013-09-24 06:47:01.670 | Traceback (most recent call last):
>>> 2013-09-24 06:47:01.671 |   File ".tox/pep8/bin/pip", line 9, in
>>> <module>
>>> 2013-09-24 06:47:01.671 |     load_entry_point('pip==1.4.1',
>>> 'console_scripts', 'pip')()
>>> 2013-09-24 06:47:01.671 |   File
>>> "/home/jenkins/workspace/gate-nova-pep8/.tox/pep8/local/lib/python2.7/site-packages/pip/__init__.py",
>>> line 148, in main
>>> 2013-09-24 06:47:01.671 |     return command.main(args[1:], options)
>>> 2013-09-24 06:47:01.672 |   File
>>> "/home/jenkins/workspace/gate-nova-pep8/.tox/pep8/local/lib/python2.7/site-packages/pip/basecommand.py",
>>> line 169, in main
>>> 2013-09-24 06:47:01.672 |     text = '\n'.join(complete_log)
>>> 2013-09-24 06:47:01.672 | UnicodeDecodeError: 'ascii' codec can't
>>> decode byte 0xe2 in position 56: ordinal not in range(128)
>>> 2013-09-24 06:47:01.673 |
>>> 2013-09-24 06:47:01.673 | ERROR: could not install deps
>>> [-r/home/jenkins/workspace/gate-nova-pep8/requirements.txt,
>>> -r/home/jenkins/workspace/gate-nova-pep8/test-requi
>>>
>>
>> This seems to be an issue with pip not being able to find the right
>> version of pyparsing. This is strange as the command invocation log does
>> not suggest any alternative indexes are used - and it works for me.
>>
>> Maybe someone from the infra team has some more insights?
>>
>> PS - the unicode error seems to happen when trying to log the failure -
>> not the cause of it.
>>
>> N.
>
> Global requirements caps pyparsing at < 2. If software isn't in global
> requirements, it's not let into the mirrors for the gate (so that shadow
> dependencies don't sneak in requirements we don't support).
>
> Which ... is exactly what happened.
>
> cliff 1.4.5 was released, and in a patch level release bumped up a
> minimum requirement a major version (pyparsing > 2 now required). This
> created a wedge situation.
>
> There are 2 options.
>
> 1) cap cliff
> or
> 2) uncap pyparsing
>
> I went for point 2, though it's not been tested with any of the stack,
> and realistically not something we probably want to do at this stage of
> release. Maybe dhellman would have other opinions once he gets online
> this morning.
>
> https://review.openstack.org/#/c/48014/ is the review in process
>
> And to members of our community who are also maintainers of python
> libraries, I'd ask that you please be really cognizant of when you move
> requirements up major versions, and please us that as criteria to bump
> your version numbers at least a minor level (i.e. second digit) and not
> just a patch level. Integrating 100 python sources into a coherent whole
> is a hard job, and anything that can make it easier would be appreciated.

Hmmm... the issue being the infra pip mirrors take some time to update, 
so here is the cliff capping review which should get the system running 
again:

https://review.openstack.org/#/c/48019/

	-Sean

-- 
Sean Dague
http://dague.net