The previous naming convention was confusing because it did not make
it explicit that the string is not being copied. We had this wrong already
in the Lua bindings and thanks to some miracle it hasn't backfired so far
(it was using the "wrap" behaviour with a string that doesn't stay alive).
In some places we actually need the "copy" behaviour and in some other
places we need the "wrap" behaviour, so let's have both variants available.
The intention is to make checks for enabled log topics faster.
Every topic has its own structure that is statically defined in the file
where the logs are printed from. The structure is initialized transparently
when it is first used and it contains all the log level flags for the levels
that this topic should print messages. It is then checked on the wp_log()
macro before printing the message.
Topics from SPA/PipeWire are also handled natively, so messages are printed
directly without checking if the topic is enabled, since the PipeWire and SPA
macros do the checking themselves.
Messages coming from GLib are checked inside the handler.
An internal WpLogFields object is used to manage the state of each log
message, populating all the fields appropriately from the place they
are coming from (wp_log, spa_log, glib log), formatting the message and
then printing it. For printing to the journald, we still use the glib
message handler, converting all the needed fields to GLogField on demand.
That message handler does not do any checks for the topic or the level, so
we can just call it to send the message.
This change completely refactors the way components are loaded in wireplumber:
- The module_init() function must return a GObject now. This object is either
a WpPlugin or a WpSiFactory in the current modules.
- When the component loader initializes a module, it automatically registers
the WpPlugin or WpSiFactory with their respective methods. There is no need
to register the WpPlugin or WpSiFactory in the module now.
- The wp_core_load_component() API has been refactored to be asynchronows. This
allows the component loader to automatically activate WpPlugin objects, and
therefore allows the application to directly get the WpPlugin without having
to find it. This simplifies a lot of things.
- The 'ifexists' and 'nofail' component flags now work even if the respective
WpPlugin could not be activated.
- The code that loads components in main.c has also been simplified a lot,
and the option to load dangling components has also been removed.
But keep the additional comments and debug statement improvements.
Also, merge the stuff from common.h, which was previously shared
with m-default-nodes.
* Remove entirely the hook priority numbers and use before/after dependencies
* Split the WpEvent code out of WpEventDispatcher
* Add methods on WpEvent to interface with it from the WpEventDispatcher.
As a bonus, we can now also implement tooling to inspect which hooks would
in theory run for an event and write tests around that
* Removed some internal debugging facilities and log calls, will redo it later.
* Using spa_list now for the list of hooks, to reduce the number of allocations
happening in the "hook collection" algorithm
* Switched some internal data to use g_new0 instead of g_slice_new0
* Added g_free to free WpEvent structures... surprisingly, we were leaking them
before
It is better to have type-specific event names to minimize the amount
of constraint string matches we do on hooks, as most hooks (if not all)
are interested on specific types of objects only.
Similarly, use a different object manager for each object type to
minimize the performance impact of iterations and lookups, as all
such actions are interested in only 1 object type every time.
Port all existing hooks to the new event names and the get-object-manager API.
Hooks need to have a priority relative to the event they are executed on,
so it does not make much sense to have all kinds of different priorities
based also on the event type and/or the module where they are defined.
Also, it wouldn't be acceptable to have such an enumeration on the public API.
- Sharpen the hooks, so that they are called only when needed.
- Make settings live, apply them when they are changed.
- Remove the state saver after events hook, call it directly.
- Remove the settings bookkeeping as the gobject properties.
- Remove the scheduling of default-nodes-changed signal via core.
This scheme provides for an orderly execution of hooks as the priorities
are controlled from one single place. Enumeration is defined in such a
way that new items can be added easily.
All the event hooks are changed to get the priorities from this
enumeration.
- Add a new variable "name" in WpEventHook and use it to log all the
hooks(by name) picked up in _push_event(). This gives a clear picture
if hook is registered for a given event.
- Form a name for an event and a chain of events for an event run, log
both of them. This gives a clear picture of the events executed and
order in which they are dispatched.
- Similarly build hooks chain and print it in _source_dispatch(), this
gives a clear picture of the hooks picked and the order in which they
are dispatched.
- Log only the dispatchable(with hooks) events, this de-clutters
the log messages.
Previously a default source had to be Audio/Source and a default sink
had to be Audio/Sink. This did not account for:
- monitor sources (Audio/Sink)
- duplex nodes (Audio/Duplex)
- virtual sources (Audio/Source/Virtual)
With this change the restriction is now on the number of input/output
ports, so any Audio/* node with output ports is a valid target to select
as a default source, while any Audio/* node with input ports is a valid
target for a default sink.
Fixes#60
