On Sun, Jun 18, 2017 at 12:19:25PM -0400, Paul Wouters wrote: > On Sun, 18 Jun 2017, D. Hugh Redelmeier wrote: > > > After a pause of a few months, I ran the test suite last night. > > I tested HEAD, as of ce5d67b98214746e8e55a2a1c401343117dba1aa. > > > > A *lot* of tests seem to have failed. I looked at a couple of failures > > and got bored. They were just random changes in XFRM logging and I had no > > way of knowing if they actually mattered. > > > > Are these results expected? > > You caught me in the middle of a big update. > > The problem was that the "ip xfrm pol" output had two issues: > > 1) pseudo random order of in/out/fwd entries in "ip xfrm pol" output > 2) Spurious extra 0.0.0.0/0 socket options line.
this sounds like an iprovement. Is the state order is important? Isn't how how packet processing will find the state. > To fix these, I updated ipsec look to use the one line ip xfrm pol > output using -o, then sort these, and then expand them again to > multiline for readability. I also filter the 0.0.0.0 socket lines out > before expanding. > > So while 2) was easy to predict and edit before the run, I hadn't done > that for issue 1), so once the current testrun with these changes > has finished, I will grab the diffs related to ip xfrm pol, and update > them. This test run is still going at: I wish "ipsec look" output change and the test output change was pushed together. For example you could run tests on a branch and merge them together. It could have avoided others stumbling on lots of failed test results on a Sunday morning and waiting for you to push "fix":) Just a wishful thinking. We will probably break something else. _______________________________________________ Swan-dev mailing list [email protected] https://lists.libreswan.org/mailman/listinfo/swan-dev
