It's not used elsewhere, so there is no good reason yet to expose it.
Besides, I expect that GLib will likely add a method in the future
to return the XDG_STATE_HOME (which is relatively new in the xdg spec)
* Make the flags public and give them nicer names
* Pass down the flags from the caller, so the caller can now explicitly
ask for looking into specific directories
* Rename the methods
* Remove and inline the wp_get_xdg_config_dir() method, since it's only
used internally
* Refactor the lookup dirs ordering to get both WIREPLUMBER_*_DIR env
variables to replace all the other directories. Previously, we were looking
for scripts in WIREPLUMBER_DATA_DIR, but we were also looking in /etc at
the same time (with precedence, even), which could result in unexpected
behaviour. Now, if a WIREPLUMBER environment variable is specified,
we only look in there.
This also corrects the logic of loading config files in m-lua-scripting.
Previously, if an error occured within the iteration function,
the error was not properly propagated to the caller because -EINVAL
was being added to nfiles instead of checked.
The previous approach to loading config files was to ask WP for the
directory and then search those for the config files. This patch changes the
approach - a caller now asks WP to search for a specific config file or
iterate over a config file directory.
This allows us to implement a directory lookup order, i.e.
"wireplumber.conf" may be in XDG_CONFIG_DIR, /etc/,
/usr/share and the first one found is used.
For configuration directories, the new method iterates over all matching
entries (files + directories) and invokes a callback for each entry.
This enables distributions to ship default files in /usr/share/wireplumber
but have admins and users override them on a local basis. For lua scripts in
particular, overriding a distribution-provided file with an empty file
effectively disables it, adding a file adds it in the right sort order.
Sometimes, especially when running the wireplumber daemon with valgrind, a new
global with the same Id as another old global owned by a proxy wants to be
exposed before the old proxy is destroyed. If this happens, instead of returning
with an assertion error, we remove the old global and we export the new one.
Globals can be removed by the object manager when their associated proxy is
currently being bound. Adding extra checks in wp_global_proxy_bin() to return
a transition error avoids binding the proxy when that happens.
If we don't remove the socket from epoll when the connection is lost, we risk
having the lost callback triggered more than once by epoll, creating race
condition in the unit test because more than 1 event happened. The epoll thread
and socket will be re-created again when re-connecting.
Note that we cannot destroy the epoll thread before triggering the lost
connection callback because we are running the callback in the epoll thread,
which is why removing the socket from epoll is the only way to make sure the
lost connection callback is only triggered once.
Aim for 8 pointers on each struct, wich exceptions:
- 12 pointers for already crowded/important ones
- 4 pointers for empty ones that are used for proxies; we are
unlikely to add any virtual methods on these classes ever
Until now, object manager could only match pw global properties on
pw global objects, because this is the only available properties set
at the time the registry creates the global.
With this change, the object manager will now bind the proxy
if the type and the pw global properties have matched and will wait
until the proxy is available with all of its properties and tries
the check again.
There is no real use for groups in our API. Just use the name of
the file as the default group and be done with it...
Storing multiple groups with this API is problematic because it
forces flushing the file to disk multiple times, one for each group,
and it's just more performant if we use a prefix in the keys
to implement some form of logical separation.
This commit also makes the GKeyFile a temporary object. As we
always load the file from the file system in _load()
and we always replace its contents with a new dictionary in _save(),
there is no point in keeping the keyfile's internal data structures
stored in memory.
Save errors are now also propagated to adhere to the programming
practices of GObject
