Florian Forster [Mon, 15 May 2017 12:40:26 +0000 (14:40 +0200)]
curl_json plugin: Fix array index and key handling.
Previously, the "key" was loaded by calling cj_cb_map_key() from
cj_cb_inc_array_index(). That means that the key for the previous element
was loaded as the array index was updated for the next element, resulting
in an off-by-one error. Also the key was not unset in time, resulting in
two metrics with the same identifier being created.
This patch fixes this with the following changes:
* cj_advance_array() (nee cj_cb_inc_array_index()) now loads the key for
the new index position instead of the previous one.
* The initial "0" key is loaded from cj_cb_start_array().
* cj_advance_array() always updates the key. The "update_key" argument
has been removed.
* Refactoring: key loading has been moved out of cj_cb_map_key() and
into its own function, cj_load_key().
Unit tests are in a separate commit for easier cherry-picking.
Florian Forster [Sun, 14 May 2017 06:24:51 +0000 (08:24 +0200)]
perl plugin: Fix potential double-free.
The av2notification_meta() function updated it's parameter by doing
(*m) = malloc();
Afterwards, if an error occurred, it would call `free(*m);` and (potentially)
return, leaving an invalid pointer stored in `m`. The caller would then try to
free the returned pointer because it was non-NULL.
This patch fixes this by doing several code cleanups:
* The meta argument is not updated unnecessarily. Instead, a local variable
is allocated and used within the loop and only on success is the return
pointer updated.
* Introduce additional local variables instead of re-using the `tmp` variable
multiple times.
* Name the variable pointing to the end of the linked list appropriately
`tail` and remove one level of indirection. Reading code that is using
pointers to pointers is unnecessarily hard.
Florian Forster [Fri, 12 May 2017 10:08:55 +0000 (12:08 +0200)]
src/utils_cmd_putval.c: Fix multi-value PUTVAL.
vl.values was allocated outside of the loop and then copied into each
ret_putval->vl[]. This means that later values overwrote the values
stored in previous ret_putval->vl[].
Florian Forster [Sun, 7 May 2017 18:34:09 +0000 (20:34 +0200)]
bind plugin: Use timegm() to convert to time_t if available.
FreeBSD doesn't provide the "extern long timezone", presumably because
it's "only" an XSI extension, leading to a portability issue with the
previous approach.
timegm() is a non-standard function available in GNU's and BSD's libc
which is doing exactly what we need. The previous code is left as a
fallback.
Pavel Rochnyack [Mon, 3 Apr 2017 05:57:09 +0000 (11:57 +0600)]
network plugin: Fix endless loop DOS in parse_packet()
When correct 'Signature part' is received by Collectd, configured without
AuthFile option, condition for endless loop occurs due to missing increase
of pointer to next unprocessed part.
Pavel Rochnyack [Thu, 9 Mar 2017 17:57:41 +0000 (23:57 +0600)]
processes plugin: Fix broken compilation on AIX
After a bb0000acfc57e230fc1c26099bc7acb913680f1a an extra pse.cpu_system
statement was left in the AIX / elif HAVE_PROCINFO_H section, which caused
the compilation to fail.
Ruben Kerkhof [Sun, 5 Mar 2017 17:30:23 +0000 (18:30 +0100)]
collectd.spec: Don't enable XFS support on RHEL6
It is missing for i386.
If there's strong interest in having support for xfs in the df plugin in
RHEL6 we can do some macro magic to only BuildRequire xfsprogs-devel if
were not on rhel6 i386, but this was the easiest fix and unblocks our
package builds.
Ruben Kerkhof [Sun, 5 Mar 2017 15:46:47 +0000 (16:46 +0100)]
Check for the CAP_IS_SUPPORTED macro
On EPEL6:
src/daemon/common.c: In function 'check_capability':
src/daemon/common.c:1571: error: implicit declaration of function 'CAP_IS_SUPPORTED'
make[1]: *** [src/daemon/common.lo] Error 1
This patch fixes parsing of metrics' types. Some statistics, such as
rx_q0errors, rx_q0bytes or rx_q0packet, don't contain suffix _errors, _bytes or
_packets. This has caused them to use generic type 'derive', rather than
more specific one, for instance 'if_rx_errors'. Additionally, some
statistics contain string 'error', rather than 'errors'.
Maryam Tahhan [Wed, 8 Feb 2017 13:48:12 +0000 (13:48 +0000)]
intel_rdt: bugfix configuration
Fix the configuration to stop collectd bombing out and deal with configuration
errors gracefully. Otherwise collectd will fail at the configuration phase
with: "Error: Reading the config file failed!".
Sebastian Harl [Sun, 11 Dec 2016 11:06:58 +0000 (12:06 +0100)]
configfile: Reintroduce support for boolean string config values.
For the network plugin, this was changed in ac73c75aed7 (which landed in 5.6)
which was a backward incompatible change breaking user configuration. Adding
support back in a central location ensures a more consistent behavior across
plugins. At the same time, we issue a warning message that this behavior is
deprecated.