Change Log for Philip's Music Writer (Linux/Unix version)
---------------------------------------------------------


Version 4.12
------------

 1. Transposing the key/chord name in a string such as "\tCm9" up by one
    semitone caused the string to be mangled and an incorrect error message to
    be generated, whereas other strings such as "\tCm11" worked fine. (This was
    a buffer length error.)
    
 2. Transposing a key/chord name in a heading line did not always work 
    correctly. For example, transposing Cm (C minor) up by one semitone
    resulted in D$m (D-flat minor) instead of C#m (C-sharp minor), even if the 
    movement's key signature preceded the heading line. This transposition now 
    works as long as the key signature is defined before the heading that 
    contains the transposition.
    
 3. Code tidies to avoid compiler warnings related to string signedness. 
 
 4. The midistart directive was in the code (from the old Acorn version) but
    was not documented, and didn't work as expected because no time deltas were
    written to the file. All the events in the midistart data are now preceded 
    by a zero delta so that they happen at the start of playing, and the 
    directive is documented.   


Version 4.11
------------

 1. The sorting of MIDI events was not quite as intended, leading to occasional 
    cases where a repeated note was turned on for the second time before 
    (rather than after) being turned off (at the same point in time). Some MIDI 
    players seem to "do the right thing" with this, but apparently others do 
    not. The sorting has been mended.
    
 2. A new character has been added to the PMW-Music font. Its number is 183 
    (decimal) or 267 (octal). The character is an "x" that is suitable for 
    printing at the head of a guitar grid to indicate a string that is not 
    played. Like the open and closed circles, its typographic width is exactly 
    right to move the printing point on to the next guitar string position.
    
 3. The heading directives playtempo, playtranspose, and playvolume, and the 
    stave directives playtranspose and playvolume have been given synonyms with 
    "play" replaced by "midi" because the old names (which originally referred 
    to MIDI and non-MIDI playing) are not longer sensible. They are, however, 
    retained for backwards compatibility. 
      

Version 4.10
------------

 1. A major new feature: text strings are now interpreted as UTF-8 character 
    strings using the Unicode encoding for non-Music and non-Symbol fonts 
    (strictly, fonts that use the Adobe standard encoding by default). For 
    backwards compatibility, a byte with a value greater than 127 that cannot 
    form part of a valid UTF-8 sequence is interpreted as a Unicode character 
    with its byte value. This change is almost backwards compatible. See the 
    documentation for exact details. 
    
 2. PMW could crash for certain command lines because a vector used during 
    argument decoding wasn't big enough. (The number of possible options has 
    increased recently.)
    
 3. PMW crashed if the quoted string was missing from a "heading" directive.
  

Version 4.07
------------

 1. Fixed some errors in the PMW-Music.afm file. These did not affect the 
    running of PMW, but could cause trouble when using the PMW-Music font with 
    other applications.
 
    (i)  The "B" for the bounding box was missing in all but the first 
         character.
 
    (ii) Some semicolons were not preceded by a space. This is not incorrect, 
         as far as I can tell, but I had a report that the afm2tfm program
         didn't like it.
         
 2. The Makefile contained just "RunTests" for running the tests; it should be
    "./RunTests". 
    
 3. Changes to the PMW-Music.pfa Postscript font:
 
    (i)  There was a typo in the data for the "diminished chord sign" character 
         (character 181) - "hlineto" instead of "hmoveto". This didn't seem to
         bother GhostScript or PostScript printers, but the "fontforge" utility
         noticed it and complained. (Note: fontforge has other problems with
         the font, which I have not resolved.)
 
    (ii) I have increased the thickness of the minim notehead by a small amount.
    
 4. There is a new command line option -nowidechars (-nw) which stops PMW from
    using the 100-point wide stave characters when printing staves. Instead,
    staves are constructed entirely from the 10-point wide characters. This
    option is provided because it seems that some PostScript interpreters
    cannot deal correctly with the very wide characters. Using the standard
    5-line stave, a 310-point stave is normally printed as the string FFFC.
    With the new option, it is printed as CCCCCCCCCCCCCCCCCCCCCCCCCCCCCCC.
    
 5. A simple facility for specifying default options has been created. When
    PMW starts up, it looks in the user's home directory for a file called
    .pmwrc. If this file exists, its contents are read and used to modify the
    PMW command line. White space (spaces, tabs, or newlines) in the file are
    used to separate items. Each item is added to the command line, before the
    given arguments. Thus, for example, if you always want to make use of the
    -nowidechars option, all you need to do is to create a .pmwrc file that
    contains
    
      -nowidechars
      
    and the effect is as if you type "-nowidechars" immediately after "pmw"
    every time you run it. If you insert an option that requires data, the data
    item must also be given in the .pmwrc file. For example, the file might
    contain
    
      -midi /usr/tmp/pmw.midi
      
    if you always want to create MIDI output. Omitting the file name causes an 
    error.
    
    Note that PMW does not allow options to be repeated, so if an option is
    present in the .pmwrc file, it cannot also be given on the command line.
    There is no way to override individual options that are set in the .pmwrc
    file. However, if the first option on the command line is "-norc", the file
    is not used at all.
    
 6. A new option, -includefont, is provided to cause PMW to include the music
    font within the PostScript output that it generates. (If the PMW-Alpha font
    is used, that is also included.) If you use this option, there is no need
    to install the font(s) for GhostScript (or any other display program) to
    find. However, it does mean that each PMW output file is bigger by 37K for
    PMW-Music and 31K for PMW-Alpha.
    
 7. New options -duplex and -tumble are provided to set the duplex and tumble 
    printing options in the PostScript output.
    
 8. A semicolon that did not immediately follow a beamable note was ignored. 
    (Way back in pre-history, space was the beam separator, and semicolon was 
    needed as a general separator, but this changed a very long time ago.) This 
    meant that errors such as "b-b-[endslur];" were not diagnosed, though if a 
    comma were used instead, there would be an error. I have now removed the 
    use of semicolon as a general separator - hopefully it is rarely used - and 
    stuck in explicit diagnostics for misplaced semicolons and commas. 
    
 9. Fixed an extremely obscure bug that caused the loss of some ties in a chord 
    when the "p" facility was used. For example, for this sequence:
    
       (#a`-d-)_ p_ P+._ | (A`+.D+.) |   
       
    one tie was lost between the final two chords. It is possible that other 
    specific sequences could also have provoked the bug.
   
10. It is now possible to use "p" and "x" at the start of a bar to repeat the 
    note or chord at the end of the previous bar. If there is no tie, the 
    accidentals are repeated; if there is a tie, they are not. If there is a 
    subsequent use of "p" or "x" after one or more tied notes, the accidentals 
    *are* repeated, according to the usual notation convention.
    
11. The Makefile has been updated so that, after installing pmw, it runs a 
    script that looks for the "gs" command and, if it finds it, writes some 
    text about how to get GhostScript to use the music fonts.
    
12. The gracespacing directive has been extended to take a second, optional 
    value. This applies when there is more than one grace note before a main 
    note. The second value specifies the space between the grace notes. The 
    default is the same as the space between the grace note and the main note, 
    and that in turn defaults to 6 points. A "+" or "-" can be used to specify
    a change to the existing value. For example:
    
      gracespacing +2 -1
      
    increases the space between the last gracenote and the main note by 2
    points, but reduces the space between multiple gracenotes by one point. If
    only one number is given to gracespacing, it is used for both values. For
    example:
    
      gracespacing +3
      
    increases both spaces by 3 points.  
    
13. Added an "uninstall" target to the makefile. 

14. Transpositions of an octave or more caused PMW to fail with an error
    message such as "Internal failure - transposition (48, 0, 60, 48)". Such 
    transpositions should now work.
    
15. There is a new command line option -printgutter. This specifies a distance
    by which righthand (recto) page images are moved to the right, and lefthand
    (verso) page images are moved to the left, thus creating a "gutter" for 
    binding when the images are printed doublesided. The -printgutter setting 
    applies only when pages are being printed 1-up. It is ignored for any of 
    the 2-up printing styles.
    

Version 4.06
------------

1. I had screwed up spectacularly in the representation of variable length 
   numbers in MIDI files, setting the top bits of the bytes to precisely the 
   wrong values throughout. This meant, for instance, that any long rests in a 
   piece threw the whole MIDI thing out, leading to chaos.
    

Version 4.05
------------

1. It is now possible to specify alterative MIDIvoices and MIDIperc files by
   means of the command line options -MV and -MP, respectively.
   
2. An incorrect MIDI file was written if there was more than one stave, and
   there were notes tied over bar lines. An attempt to play such a file using 
   Timidity caused it to stop at the point of error.
   
3. If just one bar is selected for MIDI generation, and it happens to be the 
   end of a repeat section, PMW does not now carry on to play the repeat. 
   
4. Added the -norepeats (synonym -nr) command line option, to suppress repeats
   in MIDI output.
    

Version 4.04
------------

1. Internal re-arrangement of flags for each note into two 32-bit words instead 
   of one, in order to get more flags available.
   
2. Added support for writing MIDI files. 

3. Added support for transposing chord names in text strings by means of the \t 
   escape sequence.


Version 4.03
------------

1. The PostScript music font PMW-Music has been converted from a Type 3 to a 
   Type 1 font. The file is now called PMW-Music.pfa, as per the normal 
   convention. This font should display better on the screen, and should also 
   be included in PDF files, thereby making the music therein display better.
   
2. Added three extra characters to the music font.

3. The PSheader file used a variable called PMSencoding, a hangover from the 
   previous incarnation. Renamed as PMWencoding.
   
4. If *include was used when the input was the standard input, that is, not a
   named file, PMW crashed. It no longer crashes; however, if the included name
   is not absolute, it cannot make it relative to the main input (because there 
   is no name). It just goes ahead and tries to open the file - that is, is it 
   taken as relative to the current directory. 
   
5. --help is now the same as -help, because some people are in the habit of
   using --help. 
   
6. Arguments in macro calls are now scanned (recursively) for macro calls. For 
   example, &a(&b) first expands &b, then uses the result as the argument for 
   &a. The use of & as an escape now applies only to non-alphanumeric 
   characters.
   
7. When cue notes and full-sized notes are vertically aligned, augmentation
   dots are supposed to line up. They weren't. In solving this, I've introduced 
   two states: in the default state, the dots are spaced proportionately 
   according to the size of the cue notes. There is a new option, invoked by 
   [cue/dotalign] which causes them to be placed so that they align with any 
   full-sized notes above or below on the same stave. This is useful when using 
   the cue facility for alternative notes.  
 

Version 4.02
------------

1. It was still saying "Scribe" in the PostScript output instead of "Writer".

2. Added -manualfeed to set the manual feed flag in the PostScript.

3. Put back the MIDI and playing directives for compatibility with old PMS
   files, though they have no effect. (They were made to work later - see 
   above.)


Version 4.01
------------

1. Building problem. It was including one too many "/pmw" components in the 
path used for finding the PSheader and fontmetrics.

2. Crash if sluroverwarnings was set and the bar following a warning bar did 
not exist.

3. The -debug option wasn't behaving as documented (debug output to stderr). It 
was taking the next item on the command line as a file to write to. This is 
dangerous! It now behaves as documented.

4. PMW-Music change: when being processed by GhostScript, the widths of the 
short stave characters are now (incorrectly) set as long as the long stave 
characters. This gives a much better effect in the examples in the manual when 
converted to PDF. It may also give a better screen display.

5. The PostScript command "a4" was put in the output (this dates from very 
early PostScript printers). It seems to cause trouble with GhostScript, causing 
problems in the zoom window, and it doesn't seem to be necessary any more, so I 
have cut it out.


Version 4.00
------------

Version 4.00 is the first version ported from the RISC OS program to Unix-like 
systems. A lot of changes were made as part of the porting process. However,
the following changes were also made to the application-level logic of the
program itself.

1. Removed (commented out) code for playing. (This got added back in later -
see above.)

2. Fixed a bug with very long arpeggio squiggles - buffer not big enough - 
which just doesn't show up on RISC OS.

3. A macro defined with arguments, but no text, didn't removed any arguments 
that were present when it was called. For example, if you had

  *Define X(1) @....
  
followed by &X(3) in the later text, "(3)" got left instead of being removed.

4. There was a place I forgot to record where the code was indirecting through 
a NULL pointer (for an empty string); on the Acorn this didn't produce an 
error.

****
