GitLab

If ksgrd changes the monitors then all IDs are invalid.
This can happen when a new device is plugged in and removed.

SensorObjects can remain valid and are re-matched up.

Otherwise things fail to build when it is not available (primarily on
FreeBSD).

As it turns out, DeviceStatistics does not update monotonuously, so we
end up without updates that would reset the update rate to 0 or
otherwise miss data. So instead, use an internal timer that queries the
statistics object regularly and use that data for the upload/download
sensors.

Similar to the old aggregate sensors but now in a different place.

The NetworkManager backend is preferred, let's see if this causes actual
problems first before adding another backend.

It has now been replaced by a new plugin that uses NetworkManager.

Otherwise we end up with all sensor names being prefix with a space.

Otherwise we still get updates for things that should have been removed.

Otherwise you end up with thinking things are running while they are
not.

It is set up to be able to handle different backends, though for the
moment only the NetworkManager backend is functional.

This allows us to filter out duplicate plugins before trying to load
them, avoiding nasty crashes because the system-installed plugin is
abi-incompatible with a newer version of the plugin api. Eventually we
can extend this with additional checks to make sure plugins are
compatible.

The final name consists of "prefix" + " " + "name" so that we can easily
change the prefix without needing to change the name. This helps when
dealing with sensors that should have a device name prefixed so they are
unambiguous.

It's just a user visible string so there seems no harm in this. It is
not yet properly signalled to the client that it changed however.

This allows us to create SensorObject instances in contexts where we
don't have access to a SensorContainer.

This allows us to create SensorObjects in a separate context that does
not have access to a SensorContainer instance.

This allows us to create SensorObjects in a separate context that does
not have access to a SensorContainer instance.

GIT_SILENT

Make the dbusCall function take care of its deletion so the lambda does
not have to.

These are often used for non-partition things, so we should just exclude
them.

Ideally we should just rewrite the entire thing using Solid instead of
playing whack-a-mole.

GIT_SILENT

`/` is used as path separator, so root ends up being empty leading to
weird stuff like partitions//usedlevel . It also cannot be translated so
we end up with an empty category for root. Since "root" can be a valid
mount point, the placeholder adds some extra characters.

finished via the OK button (or some other way of accepting
the dialog). Clicking *cancel* should not apply the
changed settings.

BUG: 423260

The buffer for a sscanf() %s argument with a length needs to be
1 longer than the length, for the terminating NUL.

BUG: 346228

The precision -1 was used in places where it meant "use the default
precision". For KLocale, this went through decimalPlaces() and
usually returned 2. For QLocale, this is invalid, and 6 is normally used.

This meant that 1% was displayed as 1.000000%, which is kind of useless,
just like 24KiB/s was displayed as 24.000000KiB/s -- implying a
network which could conceivably also transmit 1 bit in a little over
2 minutes, and measure that accurately.

Provide convenience methods (static, inline) for providing more
sensible precisions.

The port from KLocale -> QLocale left out the format-specifier
in the call to QLocale::toString(double, char, int), so we
ended up with the "precision" specifier as the char parameter --
speficying the format char. It looks like 0 (NUL) is interpreted
as 'f' on the Qt side, and precision is then 6.

The result was tooltip output like "31,405,564.000000 KiB"
which is nonsense -- there is no reason for six digits
of precision when displaying KiB until such a time as
micro-kibibytes ( 1/122 of a bit) are addressable.

Force format 'f' to avoid scientific notation, and
set precision to 0 for integer sensors, -1 (I assume that becomes
6 again internally in QLocale::toString()) otherwise.
The static_assert() is there as a reminder of which method
in QLocale should be called here.

Using 0x25CF or '#' is a bit of magic constant conjuration,
so factor it into a named constant and a convenience function.