Created on 2017-03-19.14:05:18 by jeff.allen, last changed 2017-03-21.03:53:11 by zyasoft.
|msg11247 (view)||Author: Jeff Allen (jeff.allen)||Date: 2017-03-19.14:05:18|
These tests currently fail on the Travis & Circle CI build bots: test_select_new test_socket test_ssl with the exception of the last two on just one Travis configuration. I'm going to exclude these from the travis build target: if we have green bots, we'll be able to tell when comething breaks that wasn't broken before, e.g. by a commit, and so will potential contributors of PRs. I'm assigning myself to do that bit. But the idea of this ticket is to track fixing these failures, so I'll unassign myself in case someone else knows how.
|msg11249 (view)||Author: Jeff Allen (jeff.allen)||Date: 2017-03-20.07:33:33|
I have achieved two green bots by excluding the networking tests. https://hg.python.org/jython/rev/44bf97c4eb3e At the same time, I added code to regrtest.py, pinched from Python 3.6, that tells us what version of Java we're really running on. Does anybody care to tackle the failing network tests themselves?
|msg11259 (view)||Author: Jim Baker (zyasoft)||Date: 2017-03-21.03:53:11|
FWIW, these socket tests were never 100% reliable on CPython 2.7 in the past. But I haven't looked at CPython 2.7 builds recently to see if this still holds. In any event, more work has been done in CPython 3.x to improve test reliability from a build bot perspective. So given all of this: probably best to exclude and run manually until more investigation can be done. But it might be more worthwhile to do this work in Jython 3.x, and take advantage of the better testing done in CPython's network stdlib.
messages: + msg11259
|2017-03-20 07:33:34||jeff.allen||set||assignee: jeff.allen -> |
messages: + msg11249