Changes committed to git://oss.sgi.com/pcp/pcp.git dev
debian/pcp.postinst.tail | 12 +++++++++++-
qa/119 | 11 ++++++-----
qa/187 | 2 +-
qa/272 | 6 +-----
qa/280 | 3 ++-
qa/294 | 10 +++++-----
qa/297 | 16 ++++++++--------
qa/322 | 3 ++-
qa/326 | 4 ++--
qa/374 | 18 ++++++++----------
qa/430 | 6 +++---
qa/536 | 10 +++++-----
qa/536.out | 1 -
qa/544 | 10 +++++-----
qa/544.out | 1 -
src/libpcp/src/logutil.c | 14 ++++++--------
src/pmie/rc_pmie | 1 +
src/pmlogger/rc_pmlogger | 2 ++
src/pmlogger/src/ports.c | 2 --
19 files changed, 68 insertions(+), 64 deletions(-)
commit 2a69d1057501ae4a0210da652c82f781f29159f5
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Wed Nov 27 17:49:15 2013 +1100
Remove PATH dependence in the collect2pcp tests
The collectl2pcp tests included a full path to the collectl2pcp
binary used in their output, which usually is just fine. I've
come across a host today where /bin is a symlink to /usr/bin so
the PATH evaluation order then determines the output produced
by the tests - murphys law kicks in - and tests fail.
Resolve this by putting the path used in a seq.full file instead.
commit e9ae3e7a901adcb92ed83cd3eff86926602e5565
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Wed Nov 27 15:30:30 2013 +1100
Tackle further qa fallout from recent hostname updates
One other side-effect of no longer associating hostnames
with anything related to DNS is that the location of the
pmlogger archives for a host may have moved. This causes
a handful of other tests to fail in non-obvious ways on a
test host that contains dots in its (real) hostname.
commit 95f21c50a8be6d442acd42f4adc88fe791aab9d7
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue Nov 26 15:11:51 2013 +1100
Remove non-deterministic output during pcp (pmcd/pmlogger) restart
Dave reported numerous QA failures as a result of this issue, and I
too am seeing some. A workaround existed in qa/187 so seems others
also have hit this. There's a race condition in the restart script
for pcp which calls the pmcd and pmlogger variants which can result
in pmcd stopping but not yet pmlogger, before the pmlogger start is
performed. This results in the get_running_pid check there finding
a pmlogger and reporting that its waiting, but there's no guarantee
being provided that the primary logger has reacted to the fact that
pmcd stopped before we get to that point. Hence, a spurious output
message can occur. Resolve this by making the scripts honour the
"quietly" first parameter in this section of the script (note that
a regular "start" (not "restart") will still report a need to delay
as a result of pmlogger presence.
commit ab69d85af325f99bf4ad98e30e344225669760a1
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue Nov 26 14:56:42 2013 +1100
Resolve further (remote) QA fallout from recent tempdir changes
commit ef226d340c6bc97bafaa8ebd3f07c1e6d0511f00
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Nov 25 13:56:54 2013 +1100
Resolve installer errors around mmv files in deb packaging
commit 417013320bd20a710604139655a9365269374e49
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Nov 25 12:11:39 2013 +1100
Fix permissions on pcp private tmp directories in deb packages
commit 57dcbf04b2ddf2b4c7ecc6cdd86274ec07b425f6
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Nov 25 11:49:53 2013 +1100
Remove recently-unused variable from pmlogger GetPorts function
commit 534aaeb91d27869b53dab64b509bd283a02023da
Merge: 0cfa876 bbbc024
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon Nov 25 07:12:55 2013 +1100
Merge branch 'dev' of git://oss.sgi.com/kenj/pcp into dev
commit bbbc0244d59ada586ac71370ed6bde5732a68261
Author: Ken McDonell <kenj@xxxxxxxxxxxxxxxx>
Date: Wed Nov 20 06:40:35 2013 +1100
libpcp/logutil.c - better cleanup of uncompressed temporary files
Noticed on a system that $PCP_TMP_DIR contained a number of files
like this ...
0DF26i: PCP archive (V.2) log volume #0 host: <somehost>
1RG2GO: PCP archive (V.2) log volume #0 host: <someotherhost>
1WFJrH: PCP archive (V.2) log volume #0 host: <somehost>
These files come from uncompressing the data volume of a PCP archive.
There was a gap between the creation of the temporary file and the
uncompress command finishing where SIGINT (or similar interruption)
would have left the partially uncompressed temporary file in
$PCP_TMP_DIR.
Moved the unlink() to just after the create+open and before the
execution of the uncompress command.
|