CHANGELOG

1.19.1
 - mad was using new decoder structure but librariy was not updated
 
1.19.0
 - alac.c did not set blocksize to NULL
 - faad should shift samles by 8 bits for mono channel
 - add LOOPBACK option
 - update Windows dll to remove msvc dependencies
 - (P) add -static version first

1.18.3
 - (P) socket string correction
 - loopack option
 - free_ssl_symbol() does not do anything when LINAKLL set to avoid dlclose() link error
 
1.18.2
 - do nothing in sighandler when already stopping threads
 
1.18.1
 - Windows 10 does has renamed libeay32.dll to libcrypto.dll and ssleay32.dll to libssl.dll (and does not even include it)
 - do not crash when SSL not found
 
1.18.0
 - align with UPnP version
 - minor correction in basic.html (RAOP identifier)
 - (P) return if query is missing client
 
1.17.0
 - openSSL loaded manually for compatibility 
 
1.16.2
 - partial-reponse (206) include range with /* except for Sonos (instead of ONLY for Chromecast)
 
1.16.1
 - support SNI for SSL
 - use strcasestr instead of stristr
 - add help for + and - codecs syntax
 
1.16.0
 - device specific <codecs> and <raw_audio_format> cna use + or - in front of option to add or remove compared to default
 
1.15.0
 - add SSL support

1.14.1
 - remove some unsued variables in main context

1.14.0
 - add 'E' tag parameter to get the browse url
  
1.13.0
 - (1) if track ends up earlier than expected and next track is short, ShortTrack was not set
 - (1) crossfade all we can at every step (don't need the whole fade buffer) and make room for process
 
1.12.0
 - Bi-directionnal volume filtering
 
1.11.2
 - do not use imageproxy for LMS < 7.8
 
1.11.1
 - short track shall workaround shall not be activated for 0 duration tracks
 
1.11.0
 - (1) refuse tracks less than 2s and gap tracks less than 10s
 - (1) cli_tag compare includes leading space
 - (1) exit output thread on 0 bytes only when player closes connection on failure (not on HEAD)
 - (2) accept to re-open socket even when draining (obuf management now allows it) - needed for CC and short tracks
 - (2) send track URL only in PLAYING state otherwise short track will create state change in LOAD state
 - (2) send STMd if STMu arrives while decoder is COMPLETE but not STOPPED (case where remote server closes the connection early)
 - (3) add repeatingstream option
 - (3) when url is missing fqdn, test "http" instead of IMAGE_PROXY and do not replace .png by .jpg
 - (3) remove superfluous send_icy_metadata
 - (3) if STMu is received when STMd not sent yet but should have, flush player and ignore STMu (this should trigger restart hopefully)
 - (3) undo right above mod as it's player's duty to restart
 - (4) do not use the "repeating" metadata gathering for icy update (add -1 as offest option)
 - (4) remove SQ_SEEK in both domain as it's not used
 - (4) HTTP 2 sockets backport from UPnP version (should have no impact here)
 
1.10.0
 - move all strings from basic.html to strings<.txt
 - remove sleep timers for conditions
 - removing device was not waking up main thread
 
1.9.1
 - backport UPnPBridge modifications for ICY (no impact)
 - use translation from strings.txt for Yes/No - 
 
1.9.0
 - unload receiver after 30s of stopped and on power off (if already stopped) to release player fro other controllers
 - log refreshing renderer only when parameters actually changed
 
1.8.3
 - remove range-request 
 
1.8.2
 - (1) various UPnP backporting (see UPnP CHANGELOG)
 - (2) move Roon Mode to squeezelite config size as Roon does report LMS version 7.7.5 but does not have tha AIFF header issue
 - (2) set range-request to none, although some client still require ranges
 - (2) hidden 'c' parameter
 - (2) add <stream_length> -2 which sets length only for PCM (when known)
 
1.8.1
  - QueueFlush correction
 
1.8.0
 - add a timeout option to remove renderers
  
1.7.7
- Search for available HTTP port must be 2x MAX_PLAYERS

1.7.6
 - Group update could free NULL pointer
 
1.7.5
 - Move mutex and cli_mutex creation/deletion out of the slimproto thread as Volume detection to make sure cli_send_cmd (volume) is not called before
 - cli_sock initialization must be done beofre slimproto thread starts (running set) and reset when switching/losing server must be mutex_protected
 
1.7.4
 - merge FriendlyName and Name in main context (FriendlyName is not a config item as such)

1.7.3
 - output_close was calling dlclose for each player termination

1.7.2
 - mad and flac correction, following squeezelite
 - encode mode was disabled in Windows version (compiled without CODECS flag)

1.7.1
 - sparc and ppc compiler required kernel > 3.2.0 due to glibc being 2.23+ ==> moved back cross-compiler to use glibc 2.19 (need OBSOLETE enabled)
 
1.7.0 
 - backport squeeze2upnp battery option
 - if ICY send is partial, do not try to send audio data
 
1.6.2
 - do not send any volume unless a real command has been sent by LMS
 - set SO_REUSEPORT only if option supported
 - full cross-compiling
  
1.6.1
 - CODECS flags to allow compilation in thru mode only
 - add SO_REUSEPORT except for Windows
 
1.6.0
 - add m4a to adts decoding in 'thru' mode
 - few alac issue correction
 - exit output_thread on streaming error only when HTTP connection is closed, no bytes sent and decoding completed (not in the select/accept waiting loop)
 
1.5.1
 - accept announces from other hosts as long as they are not a CC device themselves (mDNS relay)

1.5.0
 - add alac support
 
1.4.0
 - tail mode moved to obuf
 - obuf size fixed to 3MB with 2MB for tail

1.3.2
 - add duration to media JSON payload - does not always work, though
 - renderer index could be captured from a previous track in case of fast play/stop/play sessions
 - when using StopReceiver, the sessionId must be stopped, not the mediaSessionId
 
1.3.1
 - flac compression level was not used from mode parameter
 
1.3.0
 - get cli port from server

1.2.2
 - make sure to default to thru if mode mispelled
 - eliminate unloaded codecs in full processing mode
 
1.2.1
 - mode was still comapred to "flac" and not "flc" in slimproto.c!
 
1.2.0
 - CC firmware 1.34.130671 has a problematic GET behavior that requires a tail buffer like Sonos
 
1.1.2
 - re-encocde name for flac is named 'flc', not 'flac'
 
1.1.1
 - mp3 bitrate selection was using one level below minimum   
 - N/A: send_icy was not zeroing len in case of error (non blocking sock)
 - N/A: when building pcm 24 bits, at least 2 frames must be available in destination buf
 
1.1.0
 - add MP3 re-encoding
 - icy artwork removes ' - ' when no artist
 - reduce icy interval to 16384
 - fade-in/out was not working properly for short tracks in flow mode
 - shorter wait in select() if obuf has space and there is data to proceed
 - check mp3 sample rate & bitrate consitency
 - Improve webUI parameters handling
 
  
1.0.0
 - full processing mode with flow
 - send STMd in case of streaming error (0 bytes)
 - many bug fixes
 
0.3.0.9
 - squeeze2upnp backporting
 
0.3.0.8
 - set STREAM_DELAY to 15s
 - fix 24 bits replay gain 
 - codec list is "aac,ogg,flc,aif,pcm,mp3"
 
0.3.0.7
 - flac repositionning issue was not handled correctly
 - set AccuratePlayPoints to false
 
0.3.0.6
 - 32 bits comparaison correction (volume could be stuck)
 - do not set playing state on "buffering" event
 - ms_played was set to incorrect value when switching between tracks (track_start_time)
 
0.3.0.5
 - for Windows XP IP_ADD_MULTICAST must be after binding
 - do not use WSAPoll() in Windows for XP compatibility
 - fail if mDNS searcher cannot be initialized
  
0.3.0.4
 - add aarch64 version
 - remove libupnp & libthreadutil
 
0.3.0.3
 - (1) REPOS_TIME was set by mistake, preventing proper repositionning ...
 - (1) detect bitrate to not send STMd too early even with stream where duration is wrong (e.g. RadioParadise)
 - (2) slimproto_close must join thread, not detach it otherwise other _close function might destroy buffer mutexes
 - (2) move to "status" cmd query to have only a single query for all metadata
 - (2) remove bitrate usage to send STMd, was not reliable at status or songinfo on RP send the current bitrate 
 - (2) metadata path and track_has removed
 - (3) duration can be taken from "status" as well
 
0.3.0.2
 - mimetype delayed acquisition in output_http is a boolean, relying now on out->mimetype
 - when destroying device that was stuck in cli_send_cmd called by sq_notify, then when the sq_wipe was called, cli_mutex was already destroyed when releasing it
 - read_line for http server timeout up to 200ms
 - forced_mimetypes parameter should use "%s" format string as it can be empty
 - misc squeeze2upnp backport
 
0.3.0.0
 - rewrite of most of the core code: 
		- no use of libupnp webserver anymore
		- own HTTP server for fine control
		- streaming moved to modified squeezelite part of code
		- remove temporary file buffering	
 - new handling of mDNS discovery for on-the-fly player addition / removal (no more discovery time and removal counts)
 - review of all mutex-related issue and various risk of unprotected code
 - review of memory leaks (only the SSL remains, which is the same all over execution)
 - add possibility to dump players ('dump' & 'dumpall' commmands)
 - SSL context is not deleted at every disconnect to avoid memory leaks
 - wrong volume local change when mute was set
 - (2) cli_send_comment uses timeout in connect (for freebsd)
 - (3) mDNS search for group was messed-up
 - (5) if STMd is sent early (happens with when ratio player buffer/compression is large, e.g. mp3) services like Deezer close the stream connection because it's idle for too long
 - (5) add ogg,aac and aiff
 - (5) when missing codec, thru flag must be set otherwise output.c tries to use the encodebuf
 - (5) duration and ms_played status acquisition was protected by wrong mutex-related
 - (5) in main.c, LOCK_O/UNLOCK_O was empty
   
0.2.12.1
 - build system problem ==> LOAD_FLUSH inconsistency
 
0.2.12.0
 - add a "restart" button

0.2.11.1
 - Discovery and exit (-i) could abort without searching
  
0.2.11.0 
 - HasDigitalOut =1 so that volume can be fix
 
0.2.10.1
 - update renderer thread cleaned for better exit
 - properly stop scan thread
 - various fixes 
 
0.2.10.0
 - LOAD over LOAD flushed the queue
 - remove earlySTMD option
 - read_file needed to declare ready_buffering
 - Send stop when waiting for media session (a LOAD has been sent)
 - (P) do not save config file if it does not already exist
 - Volume can be queued
 - Volume is double
 - Reorganize code for AirConnect compatibility
 
0.2.9.1.1
 - another silly mistake in 32 bits timer comparison :-(

0.2.9.0
 - Make the 1.24 FW workaround a runtime parameter (corrected in FW 1.25+)
 - Better 32 bits rollover correction

0.2.8.0.2
 - clean settings.pm to make it more readable
 
0.2.8.0.1
 - correct timegaps using gettime_ms() to manage 32 bits rollover
 
0.2.8.0
 - Fix for FW 1.24.88047 to stop recever when user stops
 
0.2.7.4
 - Initial server detection should be in slimproto thread, otherwise sq_run_device can block 
 - When receiving media session, the setmediavolume should cause a wait
 - Workaround for FW 1.24.x problem (stop does not stop media session)
 - ActiveRemote bound to 0.0.0.0

0.2.7.3
 - merge early_STMd modifications 
 - Update file headers

0.2.7.2
 - correct some strings
 - merge EARLY_STMD and pacing_size modifications 

0.2.7.1
 - clean SendARP
 
0.2.7.0
 - set UPnP max HTTP/SOAP content *after* stack initialization
 - improve track transition speed by requesting next track from LMS earlier
 - big fat cleanup of all old useless squeezelite items (RESAMPLE, DSD, PROCESS, decoder, mode, codecs)

0.2.6.7.1 (not published)
 - backport some UPnP compatibility for HTTP transfer (no impact as casting using -2 for transfer mode)
 - on Windows, retry sending on CLI command socket as it fails with ENOTCONN sometimes at first send after opening ...
 - development versions now have an extra digit to enable auto-update
 
0.2.6.6
 - when payload for device (name or metadata) was containing %s, incorrect string was sent
 
0.2.6.5
 - when using LMS as image proxy, artwork_url must be prefixed with http://<server>:<port>
 - when killed by signal, CastStop was given the wrong argument

0.2.6.4
 - Queue everything, including media status requests
 - SSL and select() usage was broken, use SSL_pending instead ==> makes the playe much more snappy
 - (dev-2) clean some code in cast_util.c and castcore.c
 - (dev-2) do not queue status request
 - (dev-2) wait for exact waitId 
 - (dev-3) GetMediaStatus was using waitId and not checking MediaSession

0.2.6.3
 - do not remove missing player when remove_count is set to 0
 - limit sample rate when slimproto sends a rate above max_supported
 - in the UI, rename 'default parameters' to 'common parameters'

0.2.6.2
 - when IP address of <common> was set, all devices were forced we the same address
 - correct XML updating of <name>
 - change seek access to maitain same line of code with upnp version
 
0.2.6.1
 - lock state happened when receiver launch as failing 
 
0.2.6.0
 - 'pause' command was ignored !!!
 - handle CCA bug with mute (level must be changed for unmute to work)
 - allow disconnect on poweroff (but not if LMS to "pause" on poweroff)
 - filter quick pause/play
 - redo connect/ping/disconnect logic to be less aggressive
 - (dev-3) lauching receiver must wait for SSL connection, otherwise 1st play is missed

0.2.5.3
 - add Linux sparc support (experimental)

0.2.5.2 
 - re-introduce armv5te with soft floating only
 - limit the binary offering to what's supported by the platform
 - option to force use of LMS network interface (GUI and -b on command line)

0.2.5.1
 - move to cross-compilation chain to solve rpi-1 compatibility
 - detect float problem at startup for soft float version (armv5el)
 - add Linux/sparc version
 
0.2.4.2
 - Compiling on Wheezy and eglibc 2.13 causes a fail strtod on arm5el (softfloat) when used with GLIB 2.19 (more recent debian) ... no fix yet, need to use statically linked library or armv6hf
 - clean armv6 compilation flags
 
0.2.4.1
 - Volume command must be queued otherwise CCV ignores (not CCA) the previous request (could be a LAUNCH) 

0.2.4.0 
 - add <roon_mode> for Roon compatibility (default on and no cli use)
 - Roon server can't accept a time spurious time report ==> make sure a mediasession is active
 - try to avoid deadlock when responseId is newer than WaitId
 - buffer directory properly logged
 
0.2.3.0
 - send corrected time position only if seektime > LMS thread
 - remove USE_UPNP option, all is now mDNS
 - remove a few uPnP useless code oldies
 
0.2.2.6
 - get_time/set_time and get_metadata were not re-opening CLI socket (testing it before)

0.2.2.5
 - too many CLI socket were consumed when all bridges were opened - open on demand and close after 10sec
 - volume change in 2 steps: un-muted then set level, otherwise CCA do not unmute properly

0.2.2.4
 - When changing IP::PORT, a race condition could lead to use a NULL SSL context and crash
 - Volume filtering inverted: ignore commands from LMS that are result of a local change feedback (less than 1s)
 - seek offset trace had wrong format (%zu)
  
0.2.2.2 
 - when using "songinfo" query, LMS wants a client_id, although documented differently
 - max read_wait fixed at 20s
 - add pacing option to prevent buffering to be greedy (use <pacing_size)
 - cli_wait set at 1s
 - (dev-3) Handle PPC byte ordering
 - (dev-3) mDNS interface change
 - (dev-4) force MAC address creation from UDN for virtual groups
 - (dev-4) armv5 is renamed armv5el
 - (dev-4) setname was not working when device was powered off
 - (dev-4) re-connect if CLI socket is closed
 - (dev-5) cli_send_cmd returns immediately in case of socket error (no wait)
 - (dev-5) server address is memorized on LMS change
 - (dev-6) CLI commands can return empty string !
 - (dev-6) path can be NULL in metadata if index is wrong
 - (dev-7) cli_cmd parser checks for exact sent string as LMS seems to send garbage at the beginning from time to time
 
0.2.2.1
 - Player name set from LMS are now persistent, no need to set it from the plugin settings
 - adding <friendly_name> parameter to keep trace or true player name
 - Player auto refresh could be in conflict with external modifications
 
0.2.2.0
 - move <server> parameter from global to <common> (parameter copy for compatibility needs to be removed later)
 - fix server switch & rediscovery
 - armv6 static build was not listed
	
0.2.1.0
 - if MAC cannot be retrieved, create a fake one from UDN
 - force first 2 bytes of MAC to be 0xcc unless <mac_address> is set in config file
 - remove <base_mac> parameter

0.2.0.3-dev-1
 - in stream thread, endtok static parameter was breaking multi-thread !
 - (dev-2) hash32 check for NULL string

0.2.0.2
 - Change libupnp to handle AndroidTV request (range 0-1 to try to guess file size)
 - Bring back ppc version
 - Add x86-64-static version
 - Build a proper softfp for armv5
 - (dev-3) mdns search was missing some answers ==> causing device to be missed
 - (dev-4) detect ip:port change
 

0.2.0.1
 - Duration should not be adjusted when playing a live webradio
 - armv6hf version runs on raspberry pi B
 - Remove ppc build
 - Gathering metadata when scanning was creating usage of a NULL pointer
 - mDNS search can be improved by forcing host IP address using <upnp_socket>
 - When LMS was scanning, an overflow was happening in metadata gathering

0.2.0.0
 - Use mDNS for device detection
 - Support grouped devices (appear as an extra player)
 - Volume changes are now at device level (not session level)
 - Volume commands can ignored when not playing
 - Volume changes on the CC are forwarded to LMS, unless device is a group
 - Add volume scaling factor (using media session volume for individual players and ratio for group)
 - Free device search was missing ;
 - Filter volume speed changes
 - Reorganize log management
  
0.1.3.1
 - Documentation update on ReplayGain
 
0.1.3.0
 - Backport squeezetiny changes from changes made for squeeze2upnp 0.2.7.0 
 - armv5 build static and dynamic
 - powerpc build
 - ReplayGain available for raw audio only (pcm, aif)
 - add aif as an accepted codec
 - in CastLoad, MediaSession must be reset 
 - ON/OFF fully cleans any pending response as, in some cases, CC does not respond to a STOP
 - Corrected memory leak when no matching codec was found
 
0.1.2.0
 - Always gather metadata from LMS, <send_metadata> only stops sending of coverart, album, artist and title 
 - <send_metadata> was ignored except for 1st track

0.1.1.2
 - When streaming a webradio using wav format, pseudo file size was not set properly
 - MAC address parameter was not read properly in config file (was overflowing on <send_icy> parameter)
 - Add build date/time to version
 

0.1.1.1
 - Buffer overflow was preventing artwork to be sent (potentially overflow on long metadata even w/o artwork)