If a port support task was deferred until suggested, do not assume that the
suggested plugin actually supports the port, instead re-launch support check.
This covers the cases where a net port appears in a modem which only supports
AT ports (e.g. Nokia case).
Instead of providing a method to get if a plugin is requesting to get sorted
last, we provide a way to comparing two plugins, compatible with the
GCompareFunc required in g_list_sort().
The plugin manager no longer needs to take care of keeping track of which probes
are being run and to which physical device they correspond, as the upper layer's
`MMDevice' already does this. This simplifies the internal logic quite a lot, as
we can now easily suggest plugins to the remaining port probes of the same
device directly.
We no longer need to query the plugin manager whether we're checking support for
a specific port, as the ports are now grabbed only once in the MMDevice, and
therefore we ensure that supports check is launched only once in the
PluginManager for that specific port.
Also, we no longer need to query the plugin manager whether we're checking
support for a given device, as the logic to check when all ports of the device
are probed is now maintained internally.
The new `mm_plugin_manager_find_device_support()' will be an async method which
waits for ports to be added in the given `MMDevice'. As soon as all added ports
are probed, the device support check is considered finished.
If deferring tasks until result suggested, and we never got a suggestion, the
tasks may have been left deferred in loop forever. We now remove the defer
timeout in the case of waiting for suggestion, and ensure that we cancel tasks
if the last valid probing ended without best plugin found.
We chain up the Generic plugin created MMBroadbandModem objects within the
GDBusObjectManagerServer in MMManager, so that they get properly exported in
DBus.
Support checks are fully asynchronous and result is always reported when the
check is considered ready, so in-progress replies to
`mm_plugin_supports_port_finish()' don't make any sense.
There is no single case where more than one plugin may end up wanting to support
a given port, and therefore there is no need to report the numeric support level
when reporting SUPPORTED.
The new mm_plugin_manager_is_finding_port_support() allows to check whether the
Plugin Manager is looking for support in a specific port.
The new mm_plugin_manager_is_checking_device_support() allows to check whether
the Plugin Manager is looking for support in any port of a given device.
As soon as the first support check of a port in a given device finishes with a
valid best plugin and level > 0, propagate the result to other support tasks
corresponding to ports in the same physical device.
Previously, this propagation of support check results was only done once the
port was grabbed by a plugin, not just when the plugin reported that it
supported it. This change in behaviour isn't probably a big deal, as there
should not be any case where a plugin says it supports a port and then cannot
grab it.
We will keep a reference to each of the support checks currently in progress,
grouped by physical device path. The stored SupportsInfo structs as well as
the support check operations, are guaranteed to be kept valid as long as the
Plugin Manager exists. Or in other words, the Plugin Manager cannot be disposed
if there is still an ongoing supports check operation.
The new `mm_plugin_manager_find_port_support()' method requests the Plugin
Manager to iterate over the list of plugins internally handled, launching
supports task for the given port in each of them.
The method is fully asynchronous, and the result can be retrieved with
`mm_plugin_manager_find_port_support_finish()' once the operation is ready.
Aleksander Morgado
Aleksander Morgado