Add a new general config setting allow_origin that is included as
Access-Control-Allow-Origin header in http responses. This allows to make http
request against forked-daapd from browsers with enabled CORS
(https://www.w3.org/TR/cors/) via JavaScript XMLHttpRequest.
Per default the setting is not defined and no Access-Control-Allow-Origin header
is included in the response.
It seems ffmpeg/libav reads a file type like Monkey's audio in large
chunks, so that each packet must be passed to the decoder multiple
times. The decoder will return 0 until the packet is completed. This
change makes forked-daapd able to deal with a return value of 0.
Consider this use case:
1. Playback stops, device switched off
2. Device switched back on, playback started by user
Before this commit, the device would be unselected when it came back on.
I don't think that is desirable, so with this commit the selection will
stay (except if the device comes back on during playback).
Filescanner was broken in FreeBSD. Besides fixing this, using
libinotify instead of kqueue directly should make the code easier
to maintain, since it will be less divergent.
This commit includes these changes:
- Add libinotify to FreeBSD install scripts
- Fix reading multiple events from inotify fd (possible bug in Linux too)
- Deferred scanning since FreeBSD doesn't have IN_CLOSE_WRITE
- Configure search for inotify library
- Removal of kqueue stuff
Apparantly having "LIMIT -1 OFFSET 0" makes (certain versions?) of
sqlite3 unable to do the "ORDER BY". Might be a bug in sqlite. This
is only a partial fix, which will work for queries that don't actually
use an index. Those few that do may still be broken.
stops (default is clearing the queue) and if true do not clear the
queue.
This option is necessary because some dacp clients rely on the queue
being empty if player is stopped. Retune for example does not always
show the queue contents if player state is stopped.
overrun, rendering the device unusable because snd_pcm_writei()
starts blocking. This is just a poor temporary fix until we
either start using SND_PCM_NONBLOCK or put alsa in it's own thread
or implement Pulseaudio...
