Re: Crash with tupletFullLength = ##t

[Simon Albrecht]

Hello Mark,

this would have been in the domain of the bug list.

On 23.12.2015 19:58, Mark Knoop wrote:
> When tupletFullLength is set to ##t, the TupletBracket is actually
> *shortened* by a following RehearsalMark. This doesn't happen with
> tupletFullLength = ##f.
>
> This is merely ugly when a RehearsalMark is not so big, but if it
> pushes the righthand end of the TupletBracket too far, it causes a
> crash.
>
> This would seem to be related to issue 2362
> <http://sourceforge.net/p/testlilyissues/issues/2362/> however that
> doesn't mention actually crashing.

It would appear to be the same issue, yet the two examples behave 
differently: The one from issue 2362 compiles up to the most recent 
versions, if also with the console warning, while yours compiled up to 
2.19.19 (if also with bad output) and crashes from 2.19.22. We’ve had a 
number of such cases, where some stricter checking introduced around 
this time made scores actually not compile, which didn’t give sensible 
output before.
At any rate, I’d tend to add this to 2362 nevertheless.

Yours, Simon

> \version "2.19.33"
> music = { c'2 \tuplet 3/2 { c'4 c' c' } }
> {
>    \music
>    \mark \markup "long mark"
>    \music
>    \set tupletFullLength = ##t
>    \music
>    \music
>    \mark \markup "mark"
>    \music
>    \music
>    %\mark \markup "long mark"
>    %{ when uncommented, the previous line causes a crash:
>
> Drawing systems...lilypond: ../flower/include/interval.hh:227: T
> Interval_t<T>::center() const [with T = double]: Assertion `!is_empty
> ()' failed. Aborted (core dumped)
>
>    %}
>    \music
> }