* use base SI units for memory consumption
(breaks graph values compared to older version)
* fix quoting for regular expressions
* use common defaults for configuration settings
The HostDenied field is renamed to ssh. To preserve existing data,
# cd /var/lib/munin/example.net
# mv hostname-hostsdeny-HostsDenied-g.rrd hostname-hostsdeny-sshd-g.rrd
This refactor incidentally also fixes a bug where empty or commented-out
lines where also counted.
Signed-off-by: Olivier Mehani <shtrom@ssji.net>
The plugin saw only one substantial commit (its initial addition to the
repository). This very first version of the plugin was already broken
(undefined variable "core1") and was never fixed.
Thus it feels sane to assume, that the plugin was never used.
Many plugins are still failing these tests.
These expected failures are listed in a file
(t/test-exception-wrapper.expected-failures)
A wrapper script is used for running the tests and comparing the result
with the expectation (based on the file being listed in the above file).
This allows to test all new plugins, while ignoring all known failures.
"command -v" should be the preferred way of determining the location (or
existence) of an executable.
Sadly "checkbashisms" interprets "command -v" as a possible bashism,
since "command -v" was introduced in IEEE 1003.1 2013.
But checkbashisms checks for compliance with IEEE 1003.1 2004 - here
"command -v" is listed only as a portability extension.
See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=733511 for the
discussion about checkbashism and "command -v".
See https://unix.stackexchange.com/a/85250 for the discussion about
"command -v" being the most portable check for a program's location.
The workaround itself is bit ugly: 'checkbashisms' is monkey-patched
and fed into the perl interpreter. This makes it harder to re-run the
check command manually (due to complicated quoting).
