Changes committed to git://git.pcp.io/kenj/pcp master
Ken McDonell (5):
qa/common.avahi & qa/946: improve diags for debugging avahi test failures
qa/956: isolate pmcd under test
qa/common.avahi: capture stderr from avahi-browse
qa/188: update known working glibc versions for Gentoo
qa/370: deal with arithmetic uncertainty
qa/188 | 2 +-
qa/370 | 19 ++++++++++++++++++-
qa/370.out.10 | 10 +++++-----
qa/946 | 8 ++++----
qa/956 | 40 +++++++++++++++++++++++++++++++++++++---
qa/956.out | 6 ++++++
qa/common.avahi | 7 ++++---
7 files changed, 75 insertions(+), 17 deletions(-)
Details ...
commit 2d01877fc501bcb9c517c362bc2c2c90cf669129
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Feb 3 12:31:02 2016 +1100
qa/370: deal with arithmetic uncertainty
For reasons that remain a "Mystery, Berryl", there is some rare
differences in arithmetic between pmdumplog and sar2pcp, e.g. for
one metric at one point in time the value is 0.37 vs 0.38 similarly
there is another metric and instance and point in time where the
value maybe 0.00017 or 0.00018.
Add some fuzziness in the filtering to accommodate this.
A real fix would be nice, but that has so far escaped us.
commit 27c46f08ee8b6af1dd0238ac15f7853f68508335
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Feb 3 10:26:45 2016 +1100
qa/188: update known working glibc versions for Gentoo
Works for glibc-2.21-r1::gentoo.
commit 2dbba523e3464783f795dd855ffd07db511840fa
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Feb 3 10:16:17 2016 +1100
qa/common.avahi: capture stderr from avahi-browse
We were seeing failure text in .out.bad files like ...
Failed to resolve service 'PCP pmproxy on foo.localdomain' of type
'_pmproxy._tcp' in domain 'local': Timeout reached
where foo.localdomain was NOT the system under test, but was a system
that had recently been powered down and this seems to confuse avahi and
there is stale state preserved locally.
Since this has nothing to do with the PCP tests in question, capture
stderr from avahi-browse and salt it away in the .full file, but don't
pollute the .out file with messages like this.
commit f3037b2d1b1fd53238eae3f073ff8d26a604f15e
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Feb 3 10:03:41 2016 +1100
qa/956: isolate pmcd under test
This test seems to be exposed to random (non-QA) clients connecting
to the pmcd under test. To combat this ...
1. stop pmcd
2. add a -s $tmp.socket option to pmcd.options
3. get a free tcp port and set PMCD_PORT in the environment
4. start pmcd ... no clients will know how to connect to this one,
except possibly via avahi (but that's not an issue at the moment)
5. use -h unix:$tmp.socket in all tests (not -h unix:) to make sure
we can speak to our pmcd
6. clean up the mess and put pmcd back in _cleanup
commit ba01f4452330b030968b44b22404fe0afc5a785b
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Feb 3 09:46:58 2016 +1100
qa/common.avahi & qa/946: improve diags for debugging avahi test failures
|