Re: Error message

[Aaron Hill]

On 2019-06-25 4:46 am, David Kastrup wrote:
> Craig Dabelstein <address@hidden> writes:
>
> > And the guilty line is ...
> >
> > \consists "Melody_engraver"
> > %\override Stem.neutral-direction = #'()
> >
> > Commenting out the \override Stem.neutral-direction line fixed the 
> > problem.
>
> More likely than not half of a red herring: garbage protection problems
> are quite elusive to track down.  This propably changes the actions the
> Melody_engraver takes.

Unless I am completely lost myself, this is not the trampoline problem 
from before, though I suspect the appearance of "Melody_engraver" has 
confused things.

Here is the proper repro MWE:

%%%%
\version "2.19.82"
{
  \override Stem.neutral-direction = #'()
  \tuplet 3/2 { r4 b' c'' }
}
%%%%

This is bad input because neutral-direction must be strictly UP or DOWN 
(or rather it can be a procedure that ultimately evaluates to 1 or -1).  
From what I can see, the point of the property is to determine what 
happens to a note on the middle line of a staff when there is nothing 
else (like a beam) that can help sway the decision one way or the other. 
 Setting it to #'() would seem to be meaningless.

Consider the output from the above MWE:

====
GNU LilyPond 2.19.82
Processing `drul-array-bad-direction.ly'
Parsing...
Interpreting music...
Preprocessing graphical objects...
programming error: Stem dir must be up or down.
continuing, cross fingers
Finding the ideal number of pages...
Fitting music on 1 page...
Drawing systems...lilypond: 
/home/gub/NewGub/gub/target/linux-64/src/lilypond-git.sv.gnu.org--lilypond.git-stable-2.20/flower/include/drul-array.hh:35: 
T& Drul_array<T>::at(Direction) [with T = double]: Assertion `d == 1 || 
d == -1' failed.
The terminal process terminated with exit code: 134
====

LilyPond is already emitting a programming error: "Stem dir must be up 
or down".  Whenever I see "cross fingers", I never make any assumption 
of what will work or not.  So a failing assertion afterwards is not 
surprising.

Craig: Would you confirm whether your project output contains the 
programming error line?  I would say there is a bug if you are *only* 
seeing the failing assertion.

Perhaps an additional check could be added to stem.cc:649 to ensure that 
neutral-direction has a sane value, which could catch this particular 
error earlier in the process.


-- Aaron Hill