Changes committed to git://git.pcp.io/kenj/pcp master
Ken McDonell (7):
scripts/pcp-push: debugging
scripts/pcp-push: fix git update for mirror repo logic
src/include/pcp.env: rework _get_pids_by_name()
qa/870: additional diags
qa/051 & 589: filter changes for "off net" testing environments
src/pmlogger: fix logic errors in disconnect() handling
qa/389: additional diagnostics
qa/051 | 15 ++++++
qa/389 | 22 ++++++---
qa/589 | 4 +
qa/870 | 3 +
scripts/pcp-push | 31 ++++++++-----
src/include/pcp.env | 99 +++++++++++++++++++++++++++++++++++---------
src/pmlogger/src/callback.c | 10 +++-
src/pmlogger/src/fetch.c | 4 +
src/pmlogger/src/pmlogger.c | 39 ++++++++++++-----
9 files changed, 170 insertions(+), 57 deletions(-)
Details ...
commit 3a8898227326790d74d88b709ffadecd7516674a
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Sep 21 14:37:14 2016 +1000
qa/389: additional diagnostics
commit 993b5c82aba0e4d6587828fd83bc38ab307c3a4d
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Sep 21 14:33:13 2016 +1000
src/pmlogger: fix logic errors in disconnect() handling
From within the callback handler, there was a disconnect() followed
by a fallthrough to code that assumed the preceding myFetch() had
worked ... on vm31 (x86_64 Fedora 23) this was triggering a SEGV
and qa/389 failure.
On inspection there were cases in which disconnect() was being
called multiple times even though no pmReconnectContext() had
succeeded (triggering unnecessary work and repeated and misleading
diagnostic messages), and a bit of looseness in the unlocking
of the c_lock context lock.
commit 5008f08f5aee5486c5b088609c56ddf882fe4ab4
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Sep 21 11:18:09 2016 +1000
qa/051 & 589: filter changes for "off net" testing environments
The Melbourne QA Farm is (for the next 6 months) off the Internet,
so some extra cases need to be handled in the filtering.
commit fc9874f76ff83ac02b9084c9839366776bba178e
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Sep 21 11:17:02 2016 +1000
qa/870: additional diags
Turns out the real problem was in _get_pids_by_name() from pcp.env,
but these diags helped.
commit 9a8916e4b8391e909f9e2ad119ede856a37bbd82
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Sep 21 11:13:00 2016 +1000
src/include/pcp.env: rework _get_pids_by_name()
Additional work to Red Hat BZ #1365424 fix to generalize the logic
to work correctly ...
From the code ...
Find PIDs by matching the name ($1)
Executable alias name avoidance algorithm ...
- if name contains a / use it literally and consider nothing else
- if name matches an executable in $PCP_BIN_DIR, use that full pathname
- if name matches an executable in $PCP_BINADM_DIR, use that full
pathname
- if find(1) can locate name on this week's $PATH, use that full
pathname
- if name matches an executable in $PCP_BIN_DIR and/or $PCP_BINADM_DIR
and/or find(1) matches name, use _all_ of the pathnames in the order
above (needed e.g. for pmlogger in $PCP_BIN_DIR and symlink in
$PCP_BINADM_DIR)
- otherwise use name as is
commit f2ccc4543cc402ffb7cb9acf76a0fe364ff8b7b0
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Mon Sep 19 09:14:30 2016 +1000
scripts/pcp-push: fix git update for mirror repo logic
commit 060aecef381887ef282f2e1e052483c314f8f202
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Mon Sep 19 07:18:01 2016 +1000
scripts/pcp-push: debugging
Need commits to test this out ...
|