👍 I probably broke it with my merge.

We can also change pep8 back to installing the default non specified version but that comes with the risk of braking when a new version of pep8 is released as seen before.

I think fixing the PEP version is a good idea -- don't have strong opinions about to which version, though, as long as our PEP8 test is passing.

I only fixed the version of pep8 on the maintenance branch so it would pass without having to back-port changes to it.

I am in favor of letting the pep8 version be free on master. It causes great annoyance every so often, but every time it has version bumped it started to pick of more issues that really should be fixed.

Yes the pep8 version was fixed on master in ad55f34

Ok I unfixed pep8. We can always revisit fixing the version if breaking happens to often.

Not really sure what happens with Travis. Some of the jobs have been running for more than 40 minutes after the log file indicate that they finished

Oops - the long running jobs probably come from the 'travis-after-all' thing I put in. The first job will wait for all the others before ending.

The trailing 'fi' and a problem with the results file upload were my fault from earlier edits - proposed fixes here: #4151

Following the merge of #4155 this only unfixes the pep8 version. If no one else beats me to it I will merge it to get Travis fixed on master again as soon as it goes green.

Ok merging since this gets master passing and @tacaswell confirmed that pep8 version should not be fixed on master