Tue Aug 17 09:07:51 SRT 1999
----------------------------

- Fixes to the source tree so it compiles for gcc 2.95. Thanks to everyone
  who sent in patches :)
- NEW Synaescope! This is a port of Synaesthesia to the AlsaPlayer scope
  plugin architecture. Thanks Richard, it rocks!
- Stereoscope returns, lets see if its more stable now.
- Other cleanups to the code


Sat Aug 14 18:37:07 SRT 1999
----------------------------

- Fixed levelmeter and spacescope set_data functions. It should fix the
  problem on SGI and Sparc boxes
- Added correct casting in several places to keep gcc 2.95+ happy  


Wed Aug 11 09:54:43 SRT 1999
----------------------------

- Pause button now toggles between pause and play
- UI changes, for the better I hope. I hacked in an Eject button but removed
  it again (the code is still there). Somehow I still think an "Eject" button
  is not right. File selection will improve soon so that should make it
  obsolete anyway.
- Increased pitch range to 233%
- Scope plugins are now loaded dynamically! Except for Monoscope and FFtscope
  which keep the fft and convolve routines linked in until that prob is
  resolved.
- As a result of the dynamic scope plugins the stripped main binary is only
  about 150K now!  


Tue Jul 20 14:44:41 SRT 1999
----------------------------

- Minor changes to the MikMod plugin

Mon Jul 19 15:01:31 SRT 1999
----------------------------

- Major changes in thread synchronization, using pthread_join()
- Fixes for MikMod crashes
- New experimental layout of main GUI
- Updated audiofile library plugin to v0.2.1


Tue Jul 13 17:28:26 SRT 1999
----------------------------

- Bumped the default SCOPE nice level to 15 (again). This makes a HUGE
  difference for slower machines. Performance should be much better now.
  Scopes should not suck up CPU cycles that are badly needed by the decoder 
  and soundcard feader threads!!
- Modifications to the audiofile library plugin, should work much better now,
  but we're not there yet
- Added yet antoher incarnation of the FFTscope, this one is called the
  logFFTscope with a logarithmic X-axis (Petr Janecek)
- Disabled the default reverb level in the MikMod plugin. This should
  dramatically decrease the CPU usage and actually improves the sound output.
- Added genuine audiofile library detection for SGI boxes (Michael Pruett) 
- Fixed some crashing MikMod plugin bugs. Started looking at potential race
  conditions in the code. We should be save for now.
- Removed the 50ms delay time before feeding the next song to the soundcard.
  This might have a negative effect on slow boxes (stuttering output at the
  start of the first song). Please report problems.


Thu Jul  8 12:35:35 SRT 1999
----------------------------

- Minor changes to the MikMod plugin. Requires libmikmod 3.1.7 now.
- Modified time indicator to display "No time data" when there's none
  available (the plugin hasn't implemented it yet, etc..)


Tue Jul  6 03:07:06 SRT 1999
----------------------------

- New MikMod plugin from Paul Fisher. Thanks! AlsaPlayer can now do MODS,
  weehaa!!
- Removed aplay plugin for now. It is cuasing a bit of trouble for other
  plugin coders and the code is really hackish anyway.
- Also disabled the sterescope. It has been the source of crashing bugs on
  Sparc and selected x86/linux boxes. Have to figure out how/where.
- Added audiofile plugin from Michael Preutt. Thanks!
  It still needs work though. This will perhaps replace the aplay
  entirely since audiofile supports many more formats.
- Preparing code for the CVS tree
- Added gtk_set_locale() call


Thu Jul  1 11:21:30 CEST 1999
-----------------------------

- Bug fixes to the gtk+ code, delete_events were mishandled and that caused
  many unexplained segfaults when closing scope windows.
- Improved monoscope and fft code
- Removed 'active' icon in Scopes Window, it didn't work very well and I got
  too many 'bug' reports on it not working ;-]
- New SGI audio library output plugin by Michael Pruett. He also discovered
  the gtk+ bugs. Thanks!

MAJOR CHANGE:

I switched to the new mpg123 0.59r code. This needs lots of testing. It seems
to work ok on my x86 box. The nice thing is that I finally understand most of
the mpg123 code that's getting called from the glue code. This also brings
support for shoutcast/icecast a little closer (it's not here yet!!). There
are some internal changes needed to support 'streaming' media. Support for
3DNOW ccelleration is now also a possibility. Now, who wants to make libmpg123
reentrant? :)


  
Sun Jun 27 20:15:26 CEST 1999
-----------------------------

- fft lib thread-safe fixes from Richard Boulton
- Modified Monoscope by Ralph Loader! Cool stuff
- New FFTscope II!, modified FFTscope with few bars and fall-off
- Minor changes in the engine


Fri Jun 25 17:59:09 CEST 1999
-----------------------------

- Fixed some configure.in bugs
- NEW! UltraSparc native output addon!!!

Thu Jun 24 13:12:06 CEST 1999
-----------------------------

- New FFTscope implementation by Richard Boulton! Many thanks!
- Fixed spacescope hanging in 8-bit mode, thanks Fice :)
- Replaced redundant code in scopes with MACRO


Thu Jun 24 00:23:31 CEST 1999
-----------------------------

- Output plugins are now compiled using plain C. C++ compilation was not
  needed and automake was screwing up with conditional compiles
- Major configure work. The configure process should now build all
  configurable output modules (ALSA, OSS and ESD).
- New -o,--ouput switch to select the output method to use


Mon Jun 21 14:40:08 CEST 1999
-----------------------------

- Important performance bug fixed. An extra memset in the (tight) soundcard
  loop was not needed at all. This should really help especially for low
  latency playback!
- Put the indicator callback update in a thread. I don't trust the gtk+
  timeout system. I keep wondering which commercial company will use gtk+
  for their apps?!
- Fixed ALSA scope synchro bug

  
Sat Jun 19 22:17:13 CEST 1999
-----------------------------

- Yes! Input and output plugins are now loaded dynamically! This means you
  will have to do a 'make install' to copy the files to the right location.
  Future versions will allow for custom plugin paths and more flexibility.
- Changed the font of the main window
- Added another function the output_plugin structure:
	get_latency() - returns the latency introduced by the hardware
- The scopes should be synced with music using esound (0.2.12) now
- Bug fixes, and bug additions no doubt 
- Selecting something in the "Play File" dialog closes the dialog


Fri Jun 18 00:06:29 CEST 1999
-----------------------------

- Weeh! stupid bug in Playlist.cpp. The playlist loader couldn't handle paths
	with spaces in them. Fixed.
- Some instability for mp3 playback under Solaris. Needs more investigation.
  	Fixed up some code, but that's not the main prob.
	
	
Wed Jun 16 15:44:19 CEST 1999
-----------------------------

- Delete dac object at program exit
- Added functionality to load/save buttons in playlist window
- Internal changes/cleanups
- Added 2 new functions to input_plugin struct:
	nr_tracks()	- return number of tracks in stream
	track_seek() - seek to a track
- Many changes in the various input plugins, all except the mpg123 plugin
	should be ready for multi open. the mpg123 plugin needs a lot of work
	since the decoder itself is not reentrant. I hear xaudio is, and perhaps
	the skysound also. I'll be working on plugin docs sometime this summer.
- Upped ringbuffer fragment count from 20 to 44. This uses more memory but
  buffers a lot more data in RAM for playback (see app/CorePlayer.h)
- plugin->close(obj) is now called only once, as it should
- RESYNC_FRAMES is now 3 in mpg123 plugin, should make reverse playback
	a wee bit faster :)
- squashed some more bugs
- ID3 tag info is displayed in the main window (when present)


Mon Jun  7 10:30:30 CEST 1999
-----------------------------

- Optimized display info output. Fields are only redrawn if their text has
  changed. The update frequency has als been bought down from 10HZ to 5HZ
- Added check for snd_pcm_t / void handles so it compiles with older versions
  of alsa-lib
- Disabled the usleep() hack in ALSAOutput.cpp, produces crackles. I'm almost
  100% sure this is an ALSA bug.


Sun Jun  6 11:16:57 MEST 1999
-----------------------------

Some small bug fixes. First public f.n. release 


Tue May 11 09:35:23 CEST 1999
-----------------------------

Woah, what is it with all these gaping holes in the Changelog file?
Frankly I'm still a horrible documenter.

Bug fixes:
- Last (partial) ringbuffer partition is also played. It was being skipped
  earlier since it was marked empty. pcm_worker() does the right thing now
- Fixed a bug where the player would sometimes do a CPU hogging loop
  on MP3 streams with garbage at the start. There was already some code
  in mpg123 to strip this crap out but the read_frame_init() function was
  not called at the start of each new stream. Happens when you don't bother
  checking what use a function is :) 
- Fixed some potential divide by zero errors.
- Added some more 'check pointer for NULL before using it' code
- Fixed a bug in the audio data buffering code. Still wondering when
  that cast from void* to int* made it in there (I hear that's bad practice)
- Cleaned up the CorePlayer class. Not sure if I need to abstract it more
- Cleaned up some stale code in the scopes. Plugification still needs to 
  be done for scopes btw
- Cast sum to (short) in app/mpg123/decode_4to1.c WRITE_SAMPLE(...)
- Cast sum to (short) in app/mpg123/decode_2to1.c WRITE_SAMPLE(...)
- Cast sum to (short) in app/mpg123/decode_ntom.c WRITE_SAMPLE(...)
- More configure.in hacking. The player should compile on Solaris/UltraSparc
- Added an --enable-debug switch for -DDEBUG and -g
- Check for esd only if we compile for esd output

New:

- CD button. When pressed should start playing your AUDIO CD in /dev/cdrom
- Experimental 8-bit WAV support, thanks to Tkil for the algorithm 
- Parameters at startup!

Available options:

  -h,--help         print this help message
  -f,--fragsize #   fragment size in bytes [default=4096]
  -g,--fragcount #  fragment count [default=8]
  -c,--card #       use card number [default=0]
  -n,--nogui        start in command line mode
  -q,--quiet        quiet operation. no output
  -v,--version      print version of this program

- NOGUI mode. Yes, you can now use the player in your scripts. Controls are
  still missing though (I'm trying to learn Python. *hint hint*)
- Doesn't do subsecond display by default.


Thu Apr 15 15:02:34 CEST 1999
-----------------------------

Keeping track of what I'm hacking is not something I do often. But here goes:
Many (many!) changes to the internal structures of the player. The plugin
system should be very solid now. It only took me about 1 hour to write a CDDA
plugin, and most of it was cutting and pasting stuff and getting to know the
ioctl CD interface.

New:
- CDDA plugin for true CD Digital Audio playback! You get speed (pitch)
  control for free too!
 

Thu Mar  4 11:57:12 CET 1999
----------------------------

Fixes:
- Cleanup of automake builds. Distrubtions can now be build by simply doing
  a 'make dist' in the source tree, cool! Automatically fixed the .deps/
  inclusion prob!
- Started serious thinking about the marshaller thread. This is the main
  reason there is no function playlist yet. Since the program is threaded
  we can't use simple function calls in some places, messaging is needed.

MAJOR CHANGE:

I switched from the old (and buggy) amp mpeg decoding engine to mpg123. This
makes the whole player about 10x more stable since mpg123 has very robust
error correction (compared to amp, which had none :). The speed is about the
same with full optimization. It also looks like the floating point exceptions
that were occuring were happening inside the amp engine, woohoo! And of course
mpg123 has a much clearer license than amp... *Psyched*


Mon Mar  1 07:39:02 CET 1999
----------------------------

Fixes: 
- Cleanups everywhere, restructuring of code
- I finally sat down and figured out how autoconf works. So from now
  on compiling should be as easy as running configure and make (yeah right)
- packed up a pr3 release, still only really interesting to C coders or
  real enthousiasts


Thu Feb 18 00:19:23 CET 1999
----------------------------

Fixes:
- Started logging changes into a Changelog
- Massive cleanups everywhere, most code gets throught the -Wall test
  The GLADE code is the only part that produces minor warnings. Standard
  flag in the Makefile now
- Finally (finally!) got the buffering code right (crossing fingers). It
  turns out there was way too much (and incorrect) locking done on various
  datastructures. The only mutex that survived was the counter_mutex. The
  producer functions blocks on this mutex as soon as there are enough
  buffers ready for the consumer. The consumers unblocks the producer as
  soon as its starts consuming the next block. 
- Fixed a nasty bug where consumer could start reading a buffer that was
  being written to by the producer! The solution was very simple. Since
  we don't really care about a missed buffer read that much we just return
  an empty PCM block if there are currently no buffers available. The reason
  why we don't care is because this case is very rare in normal day use.
  This is a simple and elegant solution. This should also fix the race
  condition (a real life one!) which occured sometimes (the producer would
  start producing at full CPU speed never blocking untill it ran out of
  st(r)eam.
- Made the playlist do something. gtk_object_set_data() rocks!
- Tidying up for PR3 release
- Threaded file/directory loading in playlist (directory recursion is done
  in a separate thread)
- Minor GUI updates
- Scope threads now die when closing their window


??? Jan 24 00:00:00 CET 1999
----------------------------

- Somewhat usable now
- Fixed loads of race conditions/threading issues
- Implemented all scopes for 8-, 16- and 32-bit X displays
- Added more threading (each scope runs in its own thread)
- Functional FX and SCOPES button
- Software volume and balance controls (leaves the hardware mixer alone)
- Minor GUI redesign (glade is ok)
- Double buffered counters
- Further design on Playlist layout (it's *NOT* working yet)
- Minor code cleanups, still very hackish/ugly


Earlier additions
-----------------

- The base code was started on in August 1998, with on and off coding. 
  Various redesigns and new ideas were implemnted/scrapped/reimplemented
- A semi working player core appeared at the end of 1998. 
- Coding on my BeOS player BeMP has virtually stopped, there are no more
  challenge to it. I haven't even bothered porting it to R4 (maybe to R4.5)
  since BeOS boots are few and far apart 
