Changes committed to git://git.pcp.io/pcp.git master
Nathan Scott (4):
Revert "PR1099 (compressed archive) mitigation in pmwebd: skip them in
graphite mode"
docs: fix formatting botch at end of pmwebd man page
build: update gitignore files for some generated dirt
build: correct logic bug in builddefs INSTALL_MAN macro
Lukas Berk (3):
Remove unused for loop from interp.c
Adjust size variable to be signed size_t
Drop unused condition check
debian/.gitignore | 1
debian/libpcp3-dev.install | 103 +++
man/man1/pmwebd.1 | 2
man/man3/pmwebapi.3 | 2
qa/661 | 5
qa/661.out | 1032 +++++++++++++++++++++++++++++++++++++-
src/include/builddefs.in | 9
src/include/pcp/.gitignore | 1
src/libpcp/src/interp.c | 22
src/pmdas/linux/proc_partitions.c | 2
src/pmdas/mmv/src/mmv.c | 2
src/pmwebapi/pmgraphite.cxx | 16
12 files changed, 1156 insertions(+), 41 deletions(-)
commit cc3281fb34316f27bb535e4eeb9c9aced00f00dc
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Tue May 10 12:33:53 2016 +1000
build: correct logic bug in builddefs INSTALL_MAN macro
The case-insensitive filesystem handling (for MacOSX) was
being triggered for every build due to a logic error on a
conditional statement.
Correcting this means we now install both case flavours of
man pages once again. I think this is also the underlying
issue in https://github.com/performancecopilot/pcp/issues/8
... but there is some other issue at play there too, where
sort order of the mixed case vs lower case variants doesn't
always behave as expected (possibly environmental). That's
proving very difficult to reliably reproduce, however, so I
cannot confirm its fixed yet.
commit 2398ee8b05c242a6e9e9f58e33dbe4809b733e72
Author: Lukas Berk <lberk@xxxxxxxxxx>
Date: Mon May 9 16:35:23 2016 -0400
Drop unused condition check
The 'item' variable in a __pmID_int struct is unsigned, so checking that
it's greater or equal to zero will never be false. Drop the check
commit f377687f170051feaa55e67ec9f4f3968638116a
Author: Lukas Berk <lberk@xxxxxxxxxx>
Date: Mon May 9 16:34:18 2016 -0400
Adjust size variable to be signed size_t
'size' was being compared to a negative value for error checking. This
would never trigger being used as a 'size_t' as it's unsigned. Changing
to ssize_t accordingly.
commit 299c3c08f66b4c87fdeb8e11821efc14fca4aaf1
Author: Lukas Berk <lberk@xxxxxxxxxx>
Date: Mon May 9 16:33:14 2016 -0400
Remove unused for loop from interp.c
The infinite for loop was no longer being used, either one of the two
conditions were hit, or we'd goto all_done. Clean up.
commit a4e17eff10bd4959acb6e497276f4ee40974262c
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon May 9 09:34:25 2016 +1000
build: update gitignore files for some generated dirt
commit 3bc602c79079766fa562685a03ce01788190f458
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon May 9 09:33:27 2016 +1000
docs: fix formatting botch at end of pmwebd man page
commit f1fdb217a59506ee1b16ed113727bee3fa170d68
Author: Nathan Scott <nathans@xxxxxxxxxx>
Date: Mon May 9 09:33:12 2016 +1000
Revert "PR1099 (compressed archive) mitigation in pmwebd: skip them in
graphite mode"
This reverts commit 738ccc07e942d295ec516814e7b52f9c910555d8.
( Thanks Marko for the reminder about this forgotten issue. )
Commit 12da325fa449 properly fixed the archive de-compression
issues, in libpcp, so the pmwebd workarounds for the observed
slowdown can now be removed.
This removes a reduction in available data being presented via
the graphite webapp when the pmlogger_daily scripts are in use,
which compress archives by default after 3 days (pmmgr does not
offer compression so was not affected).
Finally, its worth noting that the reduction of synchronous read
I/O means certain situations will see improved performance using
compressed archives. Modern filesystem optimisations even allow
for avoiding the temporary file async write I/O, as well, under
favourable conditions ... so nothing is as simple as it might at
first seem.
|