On some systems (eg: ARM) the supported frequencies of the CPU are not
properly reported by /proc/cpuinfo so if that fails try to retrieve them
from /sys/devices/system/cpu/cpufreq/policy[0-9]/cpuinfo_[max|min]_freq.
Moving rfkill to the main event loop had unexpected side-effects.
Notably, the network module mutex can block all the main thread events
for several seconds while the network worker thread is sleeping.
Instead of waiting for the mutex let's hope that the worker thread
succeeds and schedule timer thread wakeup just in case.
Open rfkill device only once per module.
Remove rfkill threads and use `Glib::signal_io` as a more efficient way
to poll the rfkill device.
Handle runtime errors from rfkill and stop polling of the device instead
of crashing waybar.
Kernel 5.11 added one more field to the `struct rfkill_event` and broke
unnecessarily strict check in `rfkill.cpp`. According to `linux/rfkill.h`,
we must accept events at least as large as v1 event size and should be
prepared to get additional fields at the end of a v1 event structure.
Multiple .done events may arrive in batch. In this case libwayland would
queue xdg_output.destroy and dispatch all pending events, triggering
this callback several times for the same output.
Delete xdg_output pointer immediately on the first event and use the
value as a guard for reentering.
At this point we're not awaiting any protocol events and flushing
wayland queue makes little sense. As #1019 shows, it may be even harmful
as an extra roundtrip could process wl_output disappearance and delete
output object right from under our code.
Ignore any further xdg_output events. Name and description are constant
for the lifetime of wl_output in xdg-output-unstable-v1 version 2 and we
don't need other properties.
Fixes#990.
