To be consistent with the behavior of luci-lib-json, an empty Lua table
should be encoded to an empty JSON list, not an empty JSON object.
To still allow encoding empty JSON objects, the usage of anything other
than a number or a string as a key (for example an empty table or a
function) can be used to force encoding as an object:
json.stringify({}) -- "[]"
json.stringify({[{}] = true}) -- "{}"
Signed-off-by: Jan-Philipp Litza <janphilipp@litza.de>
Previously, the following caused a segmentation fault:
json.stringify({[{}] = true})
This was caused by lua_tostring() returning NULL for anything but
strings and numbers, letting json_object_object_add crash.
This patch makes jsonc ignore all keys which have no string
representation altogether.
Signed-off-by: Jan-Philipp Litza <janphilipp@litza.de>
Both init.d and uci-defaults scripts included in luci-app-statistics
still contained stuff related to lucid, which package was removed by
91b97bc9f6
Changes to scripts:
* init.d: lucid reference removed,
/var/etc creation moved earlier (before first possible use)
* uci-defaults: lucid parameters & restart removed
(but busybox httpd stuff was left intact, although it might be unnecessary)
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Mark luci-app-multiwan @BROKEN, as it depends on the old 'multiwan'
package, located in the deprecated 'oldpackages' feed.
'packages' feed contains both 'mwan3' and 'luci-app-mwan3' apps that
have superseded the old packages.
This should fix#395.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Adjust default settings to match the plugins installed by default
(default plugins: iwinfo, interface, load)
* disable plugins not installed by default to avoid error messages at start
* remove references to Freifunk interfaces, as most users do not have those
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
No functional changes, but the file is reorganised:
* group settings: general settings / output plugins / input plugins
* sort settings inside a group
* remove quotes from option names to match the current uci behaviour
* whitespace corrections
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Clarify the CPU time consumption graph by removing the "idle" data from it.
Especially with light traffic, removing "idle" enables the graph
to scale better and to properly show the CPU load variations.
If "idle" data needs to be seen, it might be added as a second graph below.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
* Clarify the short explanation on the statistics section's front page.
* Mention the possibility of additional collectd plugins to get more stats.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Utilise alt_autoscale_max to make the memory chart y-axis to scale better
for devices with e.g. 128 MB RAM.
Also fix the axis min value to 0.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Implement support for alternative scaling of the y-axis.
By default, rrdtool will autoscale to 1,2,5,10,20,50,100,200,... etc.,
which is not always suitable (e.g. memory charts for device with 128 MB).
Rrdtool 1.0.50 already supports alternative autoscaling that creates
a tighter y-axis. Implement graph-level options in Luci statistics to
support those boolean options as "alt_autoscale" and "alt_autoscale_max".
info at http://oss.oetiker.ch/rrdtool/doc/rrdgraph.en.html
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
CPU plugin in the Luci statistics was missing the label definitions,
so the field labels are like "cpu_system" instead of "System".
Add proper label definitions to CPU (like the other plugins already have).
The statistics graph was also missing softirq and interrupt stats, although colors
for them were defined. Softirq consumes massive amount of CPU especially with
any qos in use, so it is important for the user to see also that data. Add both
softirq and interrupt stats to the graph.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
This is useful in client mode as well, since it allows one to use --route-noexec and --up <cmd> or --route-up <cmd> to create routes manually instead of relying on whatever routes vpn server pushes down to the client.
mode=server dependency in luci was introduced together with script_security by mmunz back in 2011.with no explanation in the commit why mode=server was there.
Earlier update to collectd 5.4.1 changed the field from "ping" to "value",
which was changed in the graph definition here, but the label definition
was forgotten. Field's label now reads "ping_IPaddr_value".
Correct the label definition to show only IPaddr like the other two graphs.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Collectd 5.5.0 introduced new data to conntrack plugin:
In addition to the number of tracked connections there is also
the static max conntrack value and the calculated use percentage.
Luci's conntrack plugin intrepretes "conntrack-max" as a new data instance
and includes it in the graph in addition to the real "conntrack" number.
Eliminate "max" from graph by specifying empty "" instance as data source.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
The setfilehandler() functions used for mime and url encoded message
bodies all operate with a signature of fh(meta, chunk, eof), but for
unhandled encodings, the callback was directly assigned to the sink
function, which has a signature of snk(chunk). Insert a wrapper to
properly generate the EOF flag, and include a stub "meta" block
providing a virtual "name" and also the original client provided
Content-Type header, to possibly help with taking alternative actions in
the file handler.
The sink function created for raw content decoding also used the wrong
signature for the sink function.
Signed-off-by: Karl Palsson <karlp@remake.is>
Changes in 2015e and 2015f:
http://mm.icann.org/pipermail/tz-announce/2015-June/000032.htmlhttp://mm.icann.org/pipermail/tz-announce/2015-August/000033.html
Morocco will suspend DST from 2015-06-14 03:00 through 2015-07-19 02:00,
not 06-13 and 07-18 as we had guessed.
Assume Cayman Islands will observe DST starting next year, using US rules.
Although it isn't guaranteed, it is the most likely.
North Korea switches to +0830 on 2015-08-15.
The abbreviation remains "KST".
Uruguay no longer observes DST.
Moldova starts and ends DST at 00:00 UTC, not at 01:00 UTC.
Signed-off-by: Hannu Nyman <hannu.nyman@iki.fi>
Musl libc requires the length parameter of getnameinfo() to be exactly
`sizeof(struct sockaddr_in)` or `sizeof(struct sockaddr_in6)`, depending on
the passed sockaddr family.
Signed-off-by: Jo-Philipp Wich <jow@openwrt.org>
* minor fix function _list() set to local
* new function compare_version() lua version of opkg compare-version
Signed-off-by: Christian Schoenebeck <christian.schoenebeck@gmail.com>
The regex used to process the iface query string argument in
vnstat.htm was stripping underscores, which caused the graph not to
display for interfaces with names containing underscores. This
patch adds the underscore to that regex so that the interface name
will be correct and the graph will be displayed.
Signed-off-by: Michael Marley <michael@michaelmarley.com>
