Re: Possible regression with changing time-signatures and order of input

[Thomas Morley]

Am Mi., 29. Juli 2020 um 23:39 Uhr schrieb Thomas Morley
<thomasmorley65@gmail.com>:
>
> Am Mi., 29. Juli 2020 um 23:37 Uhr schrieb Thomas Morley
> <thomasmorley65@gmail.com>:
> >
> > Hi,
> >
> > this just popped up in the german forum:
> >
> > The two scores below return different output since 2.20.0. (see
> > attached) and warnings/errors are emitted for the first score:
> > warning: barcheck failed at: 1/4
> >   R1
> >
> > programming error: Multi_measure_rest::get_rods (): I am not spanned!
> >
> > With 2.18.2 it's the same.
> >
> > \version "2.20.0"
> >
> > takte = { \time 3/4 s2.*2 \time 4/4 s1*2 }
> >
> > noten = {
> >   \repeat volta 2 { R2. }
> >   \alternative { { r2. } { r1  }
> >   }
> >   R1
> > }
> >
> > \score { \new Staff << \takte \noten >> }
> > \score { \new Staff << \noten \takte >> }
> >
> > Cheers,
> >   Harm
>
> And the image:

Here some more research on it, with extended test-cases:

control = {
  \time 3/4
  \repeat volta 2 { s2. }
  \alternative { { s2. } { \time 4/4 s1 } }
  s1
}

notes = {
  \repeat volta 2 { R2.  }
  \alternative { { R2. } { R1 }
  }
  R1
}

%% Order matters
\score { \new Staff \with { instrumentName = "1" } << \control \notes >> }
\score { \new Staff \with { instrumentName = "2" } << \notes \control >> }


%% From commit-message of f372c695a52cf83ea00ff3be40acc542b4d46cbd
\new Staff \with { instrumentName = "3" }
\relative {
  \partial 4
  \repeat volta 4 { e'4 | c2 d2 | e2 f2 | }
  \alternative { { g4 g g } { a a a a } }
  a b2.
}

%% Accidental missing
\new Staff \with { instrumentName = "4" }
\relative {
  \partial 4
  \repeat volta 2 { d'4 | g1 | }
  \alternative { { fis2. } { fis1 } }
  g1
}

(1)
With released 2.18.2
Scores 1+2 look good, 3+4 not, score-4 misses the accidental in 2nd repeat.
No error/warning

(2)
With a build out of
commit f372c695a52cf83ea00ff3be40acc542b4d46cbd
Author: David Kastrup <dak@gnu.org>
Date:   Sun Nov 3 15:01:10 2013 +0100

    Issue 355: \repeat bar lines get messed up

    % the second repeat should start at the beginning of a bar, not
    % continue from the incomplete bar in the first ending.

    \version "2.17.30"
    \new Staff \relative {
      \partial 4
      \repeat volta 4 { e' | c2 d2 | e2 f2 | }
      \alternative { { g4 g g } { a a a a } }
      a b2.
    }

(first commit changing the previous behaviour)

Scores 1+2 look bad and emit barcheck-warnings and errors about not
spanned MultiMeasureRest
Score 3 looks good
Score 4 looks good as well, apart from the missing Accidental

(3)
With a build out of

commit b79db5deaf2159b569b1f9dab9c0fd59c3ff0533 (HEAD)
Author: David Kastrup <dak@gnu.org>
Date:   Mon Jan 13 12:45:33 2014 +0100

    Issue 3792: \time change under last \alternative does not persist

(first commit changing the previous behaviour)

Score 1 looks bad and emits barcheck-warnings and errors about not
spanned MultiMeasureRest
Score 2, 3, 4 look good (no warning or error)
Score 4 still misses the Accidental

This is still the behavior for current master.


I don't understand why score-1 is still bad and of course the
Accidental shouldn't be suppressed.

David I cc you, because you're the author of the mentioned patches,
any idea to improve?

Thanks,
  Harm