Re: Moving Mark_engraver causes programming error

[Thomas Morley]

Am So., 18. Juli 2021 um 17:24 Uhr schrieb Jean Abou Samra <jean@abou-samra.fr>:
>
>
>
> Le 18/07/2021 à 11:12, Thomas Morley a écrit :
> > Hi all,
> >
> > consider the follwing code, moving Mark_engraver:
> >
> > \score {
> >    <<
> >        \new Staff b1
> >        \new Dynamics
> >          \with { \consists "Mark_engraver" }
> >          { s2 \mark \default s2 }
> >    >>
> >    \layout {
> >      \context {
> >        \Score
> >        \remove "Mark_engraver"
> >      }
> >    }
> > }
> >
> > with 2.22.0 it works, with 2.23.3 I get:
> > programming error: cyclic dependency: calculation-in-progress
> > encountered for #'adjacent-pure-heights (VerticalAxisGroup)
> > continuing, cross fingers
> >
> > Similar for https://lsr.di.unimi.it/LSR/Item?id=1010
> >
> > Is this a bug or on purpose?
> > If not a bug, how to avoid the programming error?
> >
> > Thanks,
> >    Harm
>
>
> 15e9d3010cfcede094e5907b87d55737df839572 is the first bad commit
> commit 15e9d3010cfcede094e5907b87d55737df839572
> Author: Jean Abou Samra <jean@abou-samra.fr>
> Date:   Wed May 26 16:56:06 2021 +0200
>
>      Unconditionally warn about cyclic dependencies
>
>      Check for cyclic dependencies and warn about them even in non-debug
>      builds.  This is useful even for users: if LilyPond has a bug that lies
>      in a cyclic dependency, it is better to be informed about it so you
>      have a change to report it to the mailing lists rather than have
>      LilyPond behave surprisingly without any indication why.
>
>      It is also useful for those who extend LilyPond in Scheme, who
>      (from experience) can hit cyclic dependencies relatively easily.
>
>      Finally, it is useful for developers, because we might get more
>      bug reports.
>
>      This is also consistent with the code for pure heights, which
>      always warns about dependency cycles.
>
>      As a result, the check in get_grob_direction can be removed
>      since get_property now does the job.
>
>   input/regression/debug-property-callbacks.ly | 22 ++++++++++++++++++++++
>   lily/directional-element-interface.cc        |  7 -------
>   lily/grob-property.cc                        |  8 --------
>   3 files changed, 22 insertions(+), 15 deletions(-)
>   create mode 100644 input/regression/debug-property-callbacks.ly
>
> So it's a bug uncovered in development versions but
> already present in 2.22.

Probably a long standing bug.

> I'm rather glad that this
> commit had the intended effect! I just opened
> https://gitlab.com/lilypond/lilypond/-/issues/6156.

Indeed.
A bug is found and without bug-report no fix ...

> Given that so far this input seemed to work without
> surprising effects in the printed score, I suppose
> you could simply ignore the warning.

Likely.

Many thanks,
  Harm