Re: Tremolo and partcombiner: core dump (assertion failure)

[Rutger Hofman]

On 02/10/2018 03:13 PM, James Lowe wrote:
> Rutger,
>
> On Sat, 10 Feb 2018 14:30:13 +0100, Rutger Hofman <address@hidden> wrote:
>
> > Good afternoon list,
> >
> > the following combination of \partcombine and \tremolo gives a core dump
> > with the following error message:
> >
> > Preprocessing graphical objects...lilypond:
> > /home/gub/NewGub/gub/target/linux-64/src/lilypond-git.sv.gnu.org--lilypond.git-stable-2.20/flower/include/interval.hh:227:
> > T Interval_t<T>::center() const [with T = double]: Assertion `!is_empty
> > ()' failed.
> >
> > If the % \partcombineApart below is uncommented, there is no core dump.
> >
> > Lilypond:
> >
> > \version "2.19.80"
> >
> > upp = \relative c'' {
> >       \repeat tremolo 32 { d64( b) } |
> >       d1 |
> >       b'1:64 |
> > }
> >
> > low = \relative c'' {
> >       % \partcombineApart
> >       b1 |
> >       a1 |
> >       g1:64 |
> > }
> >
> > \score {
> >       <<
> >           \new Staff \with {
> >               instrumentName = "upp"
> >           } <<
> >               \upp
> >           >>
> >           \new Staff \with {
> >               instrumentName = "low"
> >           } <<
> >               \low
> >           >>
> >           \new Staff \with {
> >               instrumentName = "partcombined"
> >           } <<
> >               \partcombine
> >                   \upp
> >                   \low
> >           >>
> >       >>
> > }
> >
> > Rutger Hofman
> > Amsterdam
>
>
> Is this the same thing you reported a while back
>
> e.g
>
> Rutger Hofman said:
>
> > > This is still an issue of current master ... has anybody investigated the 
> > > problem ?
>
> > I don't know whether we have an issue in the tracker for this, but it
> > will always be an issue since there is no valid output for this input
> > (I don't think one can have both a flag and a stem tremolo). Aborting
> > with a core dump (and a cryptic message not obviously related to the
> > input) is of course about the worst possible diagnostic, but we would
> > not be able to produce valid graphics either.
>
> > David Kastrup
>
> I didn't notice if there was ever a bug filed for this.
>
> James

I don't think so. This core dump is related to the partcombiner, which 
has a lot of other issues anyway!

This bug certainly has to do with the partcombiner, but I seem to 
remember that the core dump goes away if the upper-voice last note is 
raised so there is room for the tremolo beams. But it would be 
interesting for our bug killers to inspect how the spanner stuff reacts 
to that.

Relating to the other partcombiner issues I mention above: after a lot 
of experience, my current intuition is that the partcombiner runs into 
problems with closing spanners (hairpin, dynamic, text, slur, ...) and 
changing partcombine 'mode' (unisono, apart, chords, ...), which is 
automatic. But there, the behaviour is spanners that get terminated 
/while later/, leading to e.g. hairpins that erroneously cover many, 
many systems. I would be deeply gratified if this set of bugs is also 
attacked.

Rutger