This file documents recent user-visible changes to DisOrder.
IMPORTANT: you should read README.upgrades before upgrading.
PulseAudio is now supported, if the support library is available at build time. It is the default for client applications, reflecting its widespread use in desktop environments.
Volume setting is not currently supported when PulseAudio is in use.
Disobedience now has a menu to select the network playback API.
disorder-playrtp now selects playback API using
--api option. The old API-selection options remain
available but are deprecated and will be removed in the future.
RTP playback can now transmit to multiple unicast destinations, added
and removed on demand. This is controlled by the new
option. The motivating use case is a bridged network with high- and
low-capacity components. However it is may also be useful in a routed
network without multicast routing support.
If it's detected at configure time, a new decoder program disorder-gstdecode is built. It uses the GStreamer library for audio decoding and sample-rate conversion, among other features.
The decoder has a number of options for controlling the sample-rate and audio format conversion, allowing various tradeoffs between quality and processor time. (High-quality conversion can consume more processor time than decoding, so this is potentially useful on very a small device such as a Raspberry Pi.
The GStreamer decoder can also use ReplayGain information stored in audio files' metadata to adjust playback volume, so that tracks with very wide dynamic range don't sound really quiet in comparison. This information can be added using tools such as mp3gain, vorbisgain, or metaflac. This is turned on by default if you use the decoder; you can choose between per-album or per-track settings, or turn it off entirely, using command-line options.
The GStreamer decoder is not used by default, but it can be enabled easily enough by dropping something like the following into the server configuration file:
player player *.flac execraw /usr/sbin/disorder-gstdecode -ralbum -f-8.0 player *.mp2 execraw /usr/sbin/disorder-gstdecode -ralbum -f-8.0 player *.mp3 execraw /usr/sbin/disorder-gstdecode -ralbum -f-8.0 player *.ogg execraw /usr/sbin/disorder-gstdecode -ralbum -f-8.0 player *.wav execraw /usr/sbin/disorder-gstdecode -ralbum -f-8.0
tracklengthcan now be used without arguments to clear the lists, as per the documentation.
Thanks to Mark Wooding and Joe Birr-Pixton.
The server will now automatically initiate a rescan when a filesystem is mounted or unmounted. (Use the mount_rescan option if you want to suppress this behavior.)
The server takes care not to hold audio files open unnecessarily, so that devices can be unmounted even if tracks from them are currently being buffered.
You can now edit the required-tags and prohibited-tags global preferences in Disobedience (issue #29).
The “Manage users” option is now more reliably greyed out if it will not work. This depends on a server change too, so it may still be wrong if a new Disobedience is used against an old server.
A .desktop file is now shipped, making Disobedience visible as an application in freedesktop.org-compliant systems.
The Debian package now includes a cronjob that backs up the database daily. See /etc/cron.d/disorder. You can put settings in /etc/default/disorder to turn this off or to control how long the backups are kept for.
The disorder-dump -r option was broken.
disorder-dump now gets permissions right automatically (issue #56).
Unicode support has been upgraded to Unicode 6.0.0.
Client programs no longer depend on libdb.
The following have been removed:
Various minor bug fixes.
Security: Local connections can no longer create and delete users unless they are properly authorized.
The login form now indicates that a cookie will be set.
The sense of the Control → Playing menu item was fixed.
The device configuration option now works under OS X. Devices may be specified either by UID or name. Fixes Issue 27.
Gapless play should be more reliable, and playback latency over RTP should be a bit lower. Note though that all the sound output code has been reorganized and in some cases completely rewritten, so it's possible that bugs may have been (re-)introduced. Decoding of scratches is also initiated ahead of time, giving more reliable playback.
The command backend now (optionally) sends silence instead of suspending writes when a pause occurs or no track is playing.
The libao plugin has been removed, because the plugin API is not usable in libao 1.0.0.
Playlists are now supported. These allow a collection of tracks to be prepared offline and played as a unit.
Multiple tracks can now be dragged in the queue in a single operation. Furthermore, it is now possible to drag tracks from the “Recent”, “Added” and “Choose” tabs to the queue.
Disobedience now supports playlist editing and has a compact mode, available from the Control menu.
Disobedience has a new manual.
Confirmation URLs should be cleaner (and in particular not end with punctuation). (Please see README.upgrades for more about this.)
There is a new --command option which allows the RTP player to send audio data to a user-chosen command instead of an audio API. See the man page for details.
The --device option to disorder-playrtp now works under OS X (as above).
IPv6 should now work. Network address specifications (listen, connect etc) can now be prefixed with -4 or -6 to force use of IPv4 or IPv6, though normally this should not be needed.
Unicode support has been upgraded to Unicode 5.1.0.
Various elements of the source code have been rationalized.
|#22||Background decoders interact badly with server reload|
|#27||Mac DisOrder uses wrong sound device|
|#30||mini disobedience interface|
|#32||Excessively verbose log chatter on shutdown|
|#33||(Some) plugins need -lm.|
|#39||Double bind() non-multicast AF_INET|
|#40||Missing stub function|
|#41||Missing includes for timeval|
|#42||syntax error in empeg_host section|
|#43||decoder segfault with FLAC 1.2.1|
|#44||gcc 4.3.2-1ubuntu12 SUYB patch|
|#45||disobedience doesn't configure its back end|
|#46||Sort search results in web interface|
|#48||build-time dependency on oggdec removed|
|#49||Disobedience's 'When' column gets out of date|
|#51||Improved speaker process robustness|
|(none)||“found track in no collection” messages for scratches are now suppressed|
|(none)||Disobedience would sometimes fail to notice when a track started, leading to its display getting out of date.|
A bug was fixed in the calculation of how much of the track had played. This would lead to the counter showing incorrect values in various situations.
Disobedience's icons have been changed to larger, more colorful ones. The SVG source is included if you want to fiddle with them.
“Select all” is now no longer available in the choose tab. Instead there is a new “Select children” option which selects the file children of a single subdirectory.
The selection bias for newly added tracks was reduced by half. You can put it back to the old very neophilic value with “new_bias 900000” in the config file.
It is now possible to ‘adopt’ randomly picked tracks. Disobedience and the command-line client support this but the web interface does not.
The default track name parsing was modified to handle filenames generated by iTunes.
disorderfm now preserves directory permissions.
The build system will now automatically cope with Fink's broken GTK+ packages. There are a number of improvements to the Debian packaging.
DisOrder now builds on 64-bit Linux systems.
There are various minor fixes.
The login window now has a 'remote' switch. If this is turned off then Disobedience will connect to a local server. Formerly it would always use TCP/IP.
The Delete and backspace keys will now remove tracks from the queue. Escape will now dismiss the properties, users or login windows and return now works in the login and properties windows.
Track and global preferences are now described in a new disorder_preferences(5) man page.
|#21||CGI should use PATH_INFO more sensibly|
Disobedience's “Login” window now works when you are logged in.
Disobedience has been largely rewritten:
Disobedience attempts to cope with servers from older versions, up to a point, but this is not well tested and it's best to keep the server fully up to date.
When a track shares a directory with its alias, the real track name is now returned instead of the alias (the opposite way round to the previous behaviour).
Corrected web browser linked from Disobedience.
Libtool and Automake now install the CGI correctly. As part of this, cgidir has been renamed to cgiexecdir. The configure script will report an error if you try to use the old name.
The gap directive will no longer work. It could be restored if there is real demand.
It is now possible to schedule events to occur in the future. Currently the supported actions are playing a specific track, and changing a global preference (thus allowing e.g. random play to be turned on or off). See the schedule-* commands described in disorder(1).
This has been completely rewritten to support new features:
This has been largely rewritten. The most immediate benefits are:
Customizers should find their lives easier: the syntax is less onerous, it is possible to define macros to avoid repetition, and the documentation is less monolithic (see disorder.cgi(8) as a starting point).
Mail is now sent via the system sendmail program, though it remains possible to use TCP to connect to an SMTP server. See sendmail and smtp_server in disorder_config(5).
The web interface is now installed automatically. If you don't want it, use ./configure --without-cgi. If you want it in a non-default location, or no location for it is detected, set cgidir and httpdir on the ./configure command line.
There is now a new user management window. From here you can add and remove users or modify their settings.
Relatedly, the server will only allow remote user management if you set remote_userman to yes.
scripts/setup now honors command line options, and can set up network play as well as the local default sound device.
|#2||Search results should link to directories|
|#6||Schedule tracks for a particular time|
|#10||Non-uniform track selection|
|#11||Bias random selection to newly added tracks|
|#13||Default encoding for debian setup scripts|
|#16||Cookie expiry causes user to be silently logged out and not subsequently redirected to login page|
|#20||Broken aliasing rules|
Builds --without-server should work again.
The web interface is a bit more liberal in the cookie value syntax it will accept.
Clients fail more gracefully if no password is available.
Debian upgrades from 2.0.x should now work better.
Important! See README.upgrades when upgrading.
Mac OS X and FreeBSD are somewhat supported. There is now a bash script in scripts/setup which will automate the setup after make install.
Users are now stored in the database rather than a configuration file.
The server now has a built-in list of stopwords and players, so only additions to these need be mentioned in the configuration file.
The default inter-track gap is now 0s.
How sound is played is now controlled via the new api configuration command. This also controls how the volume is set, which now works with ALSA as well as OSS.
A bug in the MP3 decoder was fixed (also in 2.0.4).
It is possible to allow users to register via the web interface.
The web interface's browser support has been improved. It has been tested with Firefox 2, Safari 3, Konqueror 3, Internet Explorer 7 and Opera 9.
A bug which would cause a crash if you attempt to rearrange the queue while no track was playing has been fixed. There is a new 'deselect all tracks' option, mirroring 'select all tracks.