Regression: 2.19.61 doesn't handle melisma properly with multiple voices

[Marnen Laibow-Koser]

Hi everyone!

I've just run across a bug in 2.19.61. Normally when lyrics are set to a
voice with slurs, syllables don't get assigned to the second and subsequent
notes in a slur. But in 2.19.61, *if the voice in question is in a
polyphonic passage*, they do, which is a regression: 2.18 handles this
correctly.  Here's some example code; note the difference in lyric underlay
between the two measures, which should assign the lyrics the same way (and
do in 2.18).

---begin LilyPond code---
\version "2.19.61" % bug shows up in this version
% \version "2.18.2" % has correct behavior, no bug
\language "english"

global = {
  \key c \major
  \time 4/4
}

melody = \relative c' {
  \global
  c4( d) e f | % works as expected
  <<
    { \voiceOne c( d) e f } % doesn't work as expected in 2.19
    \new Voice { \voiceTwo c2 c4 d }
  >> |
}

verse = \lyricmode {
  \repeat unfold 2 { one two three }
}

\score {
  \new Staff { \melody }
  \addlyrics { \verse }
  \layout { }
}
-----end LilyPond code-----

Let me know if you need any further information to investigate this bug.
Thanks!

Best,
-- 
Marnen Laibow-Koser
address@hidden
http://www.marnen.org