More ss_outdev rewriting.

[ahxm.git] / TODO

Ann Hell Ex Machina TODO / Bug List
===================================

Release Critical Bugs
---------------------

 * 1017: mp_doccer documentation is severely outdated.
 * 1022: Add file name to compiler errors.
 * 1045: Line numbers in compiler errors are incorrectly reported when
   inserting a block at the end of the line; when the block is inserted,
   the linefeed has already been read and yyline incremented. This probably
   occurs also with other directives.
 * 1051: Unify all header files into annhell.h.
 * 1052: Add quantizations.
 * 1055: Add signal control to cleanly close the devices.
 * 1056: Note printing in midiin.c should correctly print chords.
 * 1060: Add ss_tempo_from_wave to language.
 * 1063: Now that I've tested it with good hardware, the new ss_outdev
   driver is useless as is due to horrible latency. So, it must be rewritten
   to do direct writing to devices.
 * 1065: Create a document about the extended commands.

Open Bugs
---------

 * 1034: tempo_from_wave() will never work; when the wave is loaded, it's
   too late for the tempo to be set, as the ss_song stream has no tempo
   information. The only solution is to just print the optimal tempo, for
   the ahxm programmer to change it manually.
 * 1046: When a compilation error occurs inside a named block, the reported
   line number is the one where the insertion occurs, not the one where the
   block is defined.

Pending Enhancements
--------------------

 * 1002: Build an audio filtering tool to test effects.
 * 1003: There should exist 'some kind' of effect for note volumes
   to add a random variation, to add 'humanity'. This cannot
   be a standard digital effect as it should be different
   for each note.
 * 1005: Vibrato should be variable (from 'no-vibrato' to 'some-vibrato'
   in the same note).
 * 1006: A 'choose' option, to take randomly one block or another
   (blocks seem appropriate, but further study is necessary). A random
   seed script command is also necessary.
 * 1007: Support for 'tonalities'; fixed transpositions for each note of
   the octave so that, for example, a 'f' is always a 'f#'.
 * 1008: Support for 'keyboard splits': ranges of notes that a track will
   ignore. This should allow to have multiple tracks playing the same
   notes but with different channel maps and, for example, having a
   guitar glissando pan from left to right.
 * 1010: Loop stretching ('stretch wave without pitch change to last
   exactly N time units').
 * 1020: When a source file includes multiple songs, there should exist a
   way to mark song / track changes. This should be used to include text
   messages as ID3 tags or even to generate .cue / .toc files.
 * 1042: Add multiple MIDI devices; a configuration file will match device
   numbers to real addresses in /dev. Then each track in the song file could
   have a command to set which MIDI device will the events be sent to.
 * 1043: The same as the multiple MIDI devices, there should exist multiple
   dsp devices. This way, a unique computer with more than one sound card
   can output multichannel songs at a time.

Closed
------

 * 1000: Convert current event queue to two event queues. The first
   stage queue should store notes and the second note on / note off
   events (Tue, 17 Aug 2004 12:00:00 +0200).
 * 1004: Vibrato for generators (Fri, 13 Aug 2004 11:05:00 +0200).
 * 1009: Set tempo from a sample loop ('set tempo so that this wave
   will match exactly one/N measures') (Fri, 16 Aug 2004 18:08:00 +0200).
 * 1013: Volume should be a command ('v', for example) and not be
   inherited from previous notes. There should also be settable as
   relative inc/decrements (as octaves) (Mon, 30 Aug 2004 14:37:59 +0200).
 * 1014: doc/language.txt is severely outdated
   (Mon, 30 Aug 2004 15:13:00 +0200)
 * 1015: Upgrade build system (new config.sh, makefile.in, etc.).
   (Thu, 09 Sep 2004 18:56:29 +0200).
 * 1016: Integrate the new compiler (Sun, 12 Sep 2004 16:13:11 +0200).
 * 1001: Generator should use queues of active / free generators.
   (Sun, 17 Oct 2004 12:50:10 +0200).
 * 1012: Portamento should not have a 'time', only a frequency
   inc/decrement per time unit (Sun, 17 Oct 2004 13:18:04 +0200).
 * 1018: The 'default' driver should not be bound to the first usable
   one, but be a sequential try of all available drivers
   (Thu, 11 Nov 2004 12:18:12 +0100).
 * 1019: Add libartsc support (Thu, 11 Nov 2004 12:30:02 +0100).
 * 1023: The SS_FRAMES lex/yacc token should be SS_MSECS, that is,
   the compiler should not use frames but milliseconds
   (Tue, 19 Apr 2005 20:05:46 +0200).
 * 1024: The 'event1' and 'event2' stuff is ugly. It should be called,
   respectively, 'note stream' (as it's, more or less, device independent)
   and 'ss stream', which depends of softsynth, frames and such. In the
   future, there could exist a 'midi stream' if MIDI output is supported
   (Closed by #1025; Wed, 20 Apr 2005 17:30:31 +0200).
 * 1033: SONG_EV_SS_NOTE won't work, because freq can't be calculated
   on compilation time; instead, it should be a generic message
   (Sat, 23 Apr 2005 10:52:35 +0200).
 * 1026: Get rid of event.c and event.h (Sat, 23 Apr 2005 12:28:47 +0200).
 * 1027: Get rid of effect.c and effect.h (Sat, 23 Apr 2005 12:28:51 +0200).
 * 1031: Add lex/yacc code for processing absolute notes (A1, C#4, F&3)
   (Sat, 23 Apr 2005 16:14:21 +0200).
 * 1036: Unify song event types; ss_song should use the same event types
   (Sun, 24 Apr 2005 10:29:22 +0200).
 * 1035: The number of compiled tracks should be available somewhere
   (Sun, 24 Apr 2005 10:35:19 +0200).
 * 1037: ss_ins_frame() is flawed; as it directly mixes into frame[],
   each intrument's ss effects affect the previous ones
   (Mon, 25 Apr 2005 16:42:28 +0200).
 * 1025: 'event1' has been moved to song.c (generic song event stream)
   and 'event2' has been moved to ss_song.c (softsynth song event stream),
   but due to almost all of ss_song.c being private, there is no way to
   order the instruments and the effects to do anything until a way of
   storing them is decided. So, project state is really critical now as
   nothing works (last working version was 'cvs update -D 20050419').
   (Mon, 25 Apr 2005 18:16:48 +0200).
 * 1039: Volumes can't be set as integer numbers (0 and 1 should be valid)
   (Mon, 25 Apr 2005 18:26:06 +0200).
 * 1038: Comments can't be written inside extended commands
   (Mon, 25 Apr 2005 18:31:55 +0200).
 * 1021: Add line number to compiler errors (Mon, 25 Apr 2005 19:11:01 +0200).
 * 1028: input.c should be ss_input.c (Wed, 27 Apr 2005 08:08:49 +0200).
 * 1029: output.c should be ss_output.c (Wed, 27 Apr 2005 16:59:34 +0200).
 * 1044: Line numbers in compiler errors don't work if the error is inside
   a block or after inserting one
   (Closed by #1045; Sun, 08 May 2005 10:22:43 +0200).
 * 1047: Change generators to use just one volume and not a channel map
   (Mon, 09 May 2005 16:24:13 +0200).
 * 1049: Since the closing of #1047, the channel map in ss_ins is
   completely broken (Tue, 10 May 2005 18:27:28 +0200).
 * 1050: New ss_ins is flawed; volumes are never used
   (Tue, 10 May 2005 18:41:30 +0200).
 * 1041: Get rid of CHANNELS and always use dynamic structures or buffers
   (Tue, 10 May 2005 19:40:52 +0200).
 * 1048: ss_wave must be dynamic (Thu, 12 May 2005 19:23:13 +0200).
 * 1032: Refactorize input, ss_gen and ss_ins to always use pointers to
   struct ss_wave instead of copies of them (Thu, 12 May 2005 19:31:44 +0200).
 * 1053: Create a tool that will read from a MIDI keyboard and print
   the played keys on stdout (Closed by #1054;
   Mon, 23 May 2005 17:04:54 +0200).
 * 1054: Modify midiin.c to print currently played notes to stdout
   (Mon, 23 May 2005 19:10:45 +0200).
 * 1057: The Arts driver hangs forever (Sat, 28 May 2005 13:42:36 +0200).
 * 1058: The new ss_outdev model seem to have problems with pipe buffer
   sizes, at least with the OSS driver. Regardless of the number of bytes,
   data is only read in 256 byte chunks (Sun, 29 May 2005 10:35:27 +0200).
 * 1030: Refactorize the output system to always use a forked process
   (Closed by #1062; Sun, 29 May 2005 10:44:39 +0200).
 * 1061: Process the SONG_EV_SS_NOTE_ON_BY_TIME message, probably by
   unifying it with the SONG_EV_NOTE_ON (Mon, 30 May 2005 17:22:21 +0200).
 * 1011: Loop matching ('change wave pitch to last exactly N time units')
   (Renamed as 'pitch_stretch'; Mon, 30 May 2005 17:23:17 +0200).
 * 1062: The forked process for sound writing works, but it has proven
   worse in slow systems. A redesign must be done to also allow direct
   writing to ports (Closed by #1063; Mon, 30 May 2005 18:14:44 +0200).
 * 1059: Add channel maps to language (Mon, 30 May 2005 18:50:38 +0200).
 * 1040: Get rid of all _ prefixed symbols (Mon, 30 May 2005 19:12:19 +0200).
 * 1064: Document or fix the 'feature' of pitch_stretch that, if the chosen
   wave can't handle the final frequency (due to min/max boundaries), it can
   end up not sounding at all (Tue, 31 May 2005 07:41:50 +0200).

Email bugs to angel@triptico.com