Please don't force a fixed version dependency on pep8

[uiwahm]

Hello,

the use of a fixed version dependency for pep8 makes it difficult to include python-swiftclient (also holds for other packages) in a Linux distribution. The problem is as follows:

python-keystoneclient 0.1.2 depends on pep8==1.2
nosehtmloutput 0.0.4 depends on pep8==1.1
python-swiftclient 1.2.0 depends on pep==1.3

There is no way to build these packages without having a mess.

Counter example for your dependency on pep8: other packages (e.g. keystone 2012.1.2, nova 2012.1.2) don't force any specific version of pep8 in their test-dependencies.

Would you please consider changing your dependency for pep8 to be a ">=" dependency?

Thank you, Markus

[Zeb Palmer (zebpalmer)]

The only requirements for pep8 at this point are via the 'hacking' package which does require a specific version of pep8 but hacking is only a requirement for testing i.e. within the tox or development virtual environments.

[Tim Burke (1-tim-z)]

To give a bit of an update on this, python-swiftclient currently matches the 'hacking' version spec from global-requirements, which should remove the python-keystoneclient conflict.

nosehtmloutput still depends on pep8==1.1, though; maybe we should nudge openstack-infra to update it?