TL;DR
=====
Interesting failures to report this week, with data. Developers, pay
attention!
REQUEST: If you are a CouchDB committer and a member of our GitHub
organisation, you have access to re-run Travis jobs. I would ask that
you *NOT* rerun jobs on the following branches: master, 2.1.x.
Rerunning jobs removes evidence of the failure log, making it difficult
to analyse the failure after the fact. This request doesn't apply to
Jenkins, as Jenkins will create a new job with the same configuration
and not hide the result in the UI.
As we have moved to GitHub Issues and just merged a significant change
to our Jenkins setup, I am resetting the list of test failures in this
email. The old query filter will remain up in JIRA; as issues recur I
will port them over from JIRA to GH.
The new open testsuite issue query URL is:
https://s.apache.org/couch-testsuite-issues
We currently have 7 open GH testsuite issues.
New Fixes
=========
Commit 354d3b0: Fix rewrite_js test which used bad pattern of deleteDb
immediately followed by createDb. (Fixes Jenkins build 5)
New Issues
==========
#547: dev/run fails to startup cluster
This was the mem3 race condition fix that didn't yet get cherry-
picked to 2.1.x. That fix has now been applied, commit 9875e8b.
#548: 500 error on getting view output
Getting the results of a view fails due to a gen_server exit.
@davisp thinks something isn't properly monitoring the index, and it
closes as a race condition. A fix has been proposed in the issue.
No PR has been submitted yet.
#551: couchjs segfaults
Moved out of COUCHDB-3352. Note the IRC log pasted into the issue
that may point at a possible workaround.
#552: view monitor failure in couchdb_views_tests/couchdb_1283
This is another eunit test failing on attempting to access an index
where gen_server exits with a {normal, ...} type error. May be
the same root cause as #548.
#553: js restartServer() sometimes fails
JS tests that have calls to restartServer() sometimes fail. Likely
the timeout needs to be increased here - the logfile looks normal.
#554: js stats test has inconsistent results
Sometimes the stats API test has off-by-one errors. I am concerned
that we're actually doing the wrong thing here, but it could just be
a poorly written test. More investigation is required.
#555: Mysterious killed eunit test in couch_replicator_use_checkpoints_tests
This is the first time I've seen an eunit test actually report that
it was *killed*. The remainder of the eunit tests execute after
this, so it's not an OS-level termination as far as I can tell. The
couch.log shows nothing unusual either. Need more eyes on this one.
Other
=====
(INFRA) Sometimes, docker pull times out on Jenkins nodes.
We're seeing nodes simply go-out-to-lunch in Jenkins runs, usually
during the `docker pull` command, but sometimes in the middle of an
actual run. I will be opening an ASF INFRA ticket to try and resolve
this problem tomorrow.
