Marc Fournier [Tue, 12 May 2015 20:14:23 +0000 (22:14 +0200)]
curl_xml: fix 3 small memory leaks
Discovered while testing the previous 2 commits. NB: valgrind
already complained about these before 0afea606 was applied, so this
isn't related to issue #935.
Jan Andres [Sat, 31 Jan 2015 08:43:30 +0000 (09:43 +0100)]
processes: Fix build for Solaris 32-bit binary
Work around a #error in <sys/procfs.h> that occurs when building a
32-bit binary with _FILE_OFFSET_BITS=64. This uses a hack similar
to the one in swap.c.
Jan Andres [Wed, 28 Jan 2015 14:06:04 +0000 (15:06 +0100)]
zfs_arc: Don't spam the log if we fail to read the "l2_size" kstat value.
The value got lost with some patch to Solaris 10 and has only reappeared
recently in 11.2. If we ever fail to read this value, stop trying, so we
don't report an error on every interval.
Marc Fournier [Tue, 12 May 2015 20:40:27 +0000 (22:40 +0200)]
oconfig: fix oconfig_free to free all elements
The recursive nature of this function made it difficult to free the root
node of the config tree. Splitting it in 2 allows to work around this
problem.
Marc Fournier [Tue, 12 May 2015 20:40:27 +0000 (22:40 +0200)]
oconfig: fix oconfig_free to free all elements
The recursive nature of this function made it difficult to free the root
node of the config tree. Splitting it in 2 allows to work around this
problem.
Marc Fournier [Tue, 5 May 2015 20:10:16 +0000 (22:10 +0200)]
Turbostat: reorganise metric names
The `plugin_instance` now always is the hardware element
(cpuNN/coreNN/pkgNN) the data belongs to. Any optional identifier string
is moved to the `type_instance`.
Also remove the "_W" unit suffix as the "power" `type` implies we're
dealing with Watts.
The idea is to have the naming scheme in line with the way most other
plugins usually organise data.
Marc Fournier [Fri, 1 May 2015 18:40:59 +0000 (20:40 +0200)]
Turbostat: confine AC_COMPILE_IFELSE() inside a AC_CACHE_CHECK block
The previous way of doing worked as expected, but didn't print out
anything in the summary, making it difficult to know why the plugin
wasn't built on older systems.
Marc Fournier [Wed, 29 Apr 2015 22:15:19 +0000 (00:15 +0200)]
write_sensu: fix format-string portability problem
Casting counters, derives and abolutes to int64_t was incorrect, as they
are respectively `unsigned long long`, `int64_t` and `uint64_t`.
Apart from potentially loosing precision, the `%ld` format-string made
clang choke on the 32bit architecture (follow-up to 78340212).
Marc Fournier [Tue, 28 Apr 2015 05:39:02 +0000 (07:39 +0200)]
make java components discovery follow symlinks
When `--with-java` points to a symlink, `find` should resolve it, making
the configure script work seamlessly with symlinks pointing to JDK
installations.
This fixes the confusing discrepancy between `--with-java=/path/to/java`
failing and `--with-java=/path/to/java/` working.
Ruben Kerkhof [Sun, 18 Nov 2012 10:38:24 +0000 (11:38 +0100)]
Fallback to /usr/lib/jvm if JAVA_HOME is not set
This makes the java plugin build out of the box
on systems with a JDK installed.
/usr/lib/jvm is the default location for the JDK
on at least Fedora, Red Hat and Debian.
In the function meta_data_get_string(), when the type mismatchs,the ERROR statement should be
ERROR ("meta_data_get_string: Type mismatch for key `%s'", e->key);
not
ERROR ("meta_data_get_signed_int: Type mismatch for key `%s'", e->key);
Marc Fournier [Tue, 14 Apr 2015 16:48:07 +0000 (18:48 +0200)]
write_sensu: fix 2 format-related details
The default CC on freebsd-10 (clang 3.4) spotted these issues:
write_sensu.c:494:38: error: format specifies type 'long long' but the argument has type 'int64_t' (aka 'long') [-Werror,-Wformat]
write_sensu.c:494:45: error: data argument not used by format string [-Werror,-Wformat-extra-args]
Tim Laszlo [Fri, 20 Jun 2014 14:00:43 +0000 (09:00 -0500)]
network: set ttl before opening server sockets
If TimeToLive is defined in collectd.conf after Server options, the
net_config_ttl value is updated after network_config_add_server and
the desired ttl is not used.
openvpn plugin: Don't signal an error when no clients are connected.
In the multi1_read() function, an error (zero) was returned when no
clients were currently connected to the OpenVPN server, because the
"read" variable was initialized to zero and the while loop exited before
it was set to one. This is not the intended behavior.