GitLab

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

It's assigned multiple times.

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

Fixes another leak.

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

This went unnoticed since the manager forgot to cleanup
this at all (see follow up commit).

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

The current logic was fragile and failed e.g. when the
FEEDBACKD_UDEV_VAL_LED would not match leaking the dev since the
g_object_unref() at the end of the loop was never hit.

Prevent that with automatic cleanup and explicitly ref'ing the
devs we want to use.

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

g_ascii_strdown() creates a copy already.

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

g_enum_to_string() is transfer full.

Guido Gunther authored Apr 22, 2021 and Guido Gunther committed Apr 26, 2021

So far we only unref'ed the device.

When the client goes away we build up a list of events to end feedback
for to make sure the hash table isn't modified in place. tThe alternative
approach would be to disconnect the on_event_feedbacks_ended handler,
duplicate the DBus signal emission and use g_hash_table_foreach_remove
but that looks less robust than using the same code path.

Closes: #25

This allows us to store the DBus sender

This matchs phosh's style.

It is TriggerFeedback and not Feedback (anymore). Also add a note how to
end the feedback in order to prevent loss of hearing :-).

If the trigger is the last thing in sysfs_path/LED_TRIGGER_ATTR, then
the new-line at the end of the output causes g_strv_contains to fail to
find the trigger in the list. This strips the new-line before splitting
the output into a list for searching.

You can see the problem this patch fixes with the librem 5, by:

$ rmmod ledtrig-timer     # or whatever is listed at the end, so 'pattern' is the last thing
$ cat /sys/devices/platform/pwmleds/leds/blue:status/trigger
[none] kbd-scrolllock ... pattern
$ fbd-ledctrl -p /sys/devices/platform/pwmleds/leds/blue:status -t pattern
Failed to set perms of /sys/devices/platform/pwmleds/leds/blue:status/pattern to 101: No such file or directory
$ modprobe ledtrig-timer
$ cat /sys/devices/platform/pwmleds/leds/blue:status/trigger
[none] kbd-scrolllock ... pattern timer
$ fbd-ledctrl -p /sys/devices/platform/pwmleds/leds/blue:status -t pattern
<success>

Don't leak the error values in the success case

This is not an error per se.

tracks the FdbFeedbackSounds in a GHashTable in the FdbDevSound

It was off in several places. Now that there's no outstanding
MRs clean it up via

   uncrustify --replace -c phosh.cfg src/fbd-feedback-manager.c

Reads the compatible list of the device and tries to find a device-specific theme file in all  directories. If no configuration is found, we fall back to a default theme file. The device-specific theme file can always be overriden by specifying the FEEDBACK_THEME environment variable

Run a script as a post install to make sure that all GLib schemas are compiled

The foreach loop was not used to validate all the include theme files

Use a separate group for users that should be able to access
the LED devices.

The names changed in device tree which tickles up into userspace.

The fbd_dev_led_set_brightness function is only getting called with
parameter 0 at the moment so this was never noticed.

This allow us to get debs quicker

This allows pa to cork, duck, etc.

This is not a critical since we prevented submitting initially
but daemon cleanup processes all entries in the theme.