Re: Semantically equivalent scores look different

[Erik Sandberg]

OK, now I have found an example. It is very constructed and of no practical 
use, but it does show that something is strange in Lilypond:

Compile the following file with Debian's unstable 1.6.4. It gets 2 lines long.
Then comment the 
  \x <{\y} \\ {\z}>
line and uncomment the
  <{\x \y} \\ {s2 \z}>
line. The length of \x is 1/2, so nothing should change (as \stemBoth is used 
in \x). But if you recompile the file now, the notes get identical except it 
takes 3 lines instead of 2.

This required quite much fine-tuning, and since random things might have 
changed in 1.6.5, there might be no longer be a difference in the number of 
lines for this particular example. However, I guess the spacing will be 
slightly different, which will be visible if you compare the tex/dvi/ps 
output with diff, or carefully compare the output pixel-by-pixel.

I use Python 2.1.3, Guile 1.6.0.

\version "1.6.4"
\include "paper16.ly"

x = \notes \relative c'' {
  \stemBoth <e8 c g> \times 2/3 {[g16 e c]} <d8 d, g,> \times 2/3 {[g16 d b]}
}
y = \notes \relative c'' {
  \stemUp \times 4/6 {[c b a e' d c]} \stemBoth [b32 c b c d16 dis] |
  %7
  [e32 fis g fis g fis e d!] [c d e d e d c b] \stemUp [b8 a] \stemBoth r8 d |
}
z = \notes \relative c' {
  d4 s | s2 d4 s4
}
\score {\notes \relative c'' {
  \key g \major
  \time 4/4

  \repeat "volta" 2 {
    <d8 d, g,> <d8 d, g,> \times 4/6 {r16 [b c d e d]} <c8 d,> <c d,> \times 
4/6 {r16 [a b c d c]} |
    <b8 d, g,> <b d, g,> \times 4/6 {r16 [g' fis g g, b]} [<a8 e c> \times 
2/3 {c16 d e]} \times 4/6 {[d c b a g fis]} |
    g8 g, r16 fis' g a <b8 d, g,> <b d, g,> \times 4/6 {r16 [g a b c b]} |
    %4
    <a8 fis d> <a fis d> \times 4/6 {r16 [fis g a b a]} <g8 g,> <g g,> \times 
4/6 {r16 [b a b b, d]} |
    [<a'8 e c> \times 2/3 {a16 b c]} \times 4/6 {[b a g c, b a]} g8 g r 
\times 2/3 {[g''16 d b]} |
    \x <{\y} \\ {\z}>
%   <{\x \y} \\ {s2 \z}>
  }
}}


Erik