Re: segfault in grob.cc

[Han-Wen Nienhuys]

address@hidden writes:
> 
>    > What is a broken smob, anyway?  Is it possible that my score has some
>    > diabolical feature that makes a bunch of those?
> 
>    Lily constructs  the score as one very long line (or system). After
>    line breaking, this long line is broken into the lines which are
>    printed, and spanners (slurs, ties, etc.) are broken along. This
>    linebreaking requires that lots of pointers have to be remapped, which
>    is what handle_broken_smobs () does.
> 
> A really high estimate for me would be 800 measures with a linebreak
> every five, and 12 staves.  If I had a meter change at every
> linebreak, that would still only be 1920 (x2 for each part?) broken "smobs".

Actually, no: at every feasible line break, bar lines and clefs are doubly
duplicated to deal with linebreaks, so a better estimate would be

  800 measures * 3 items / linebreak (clef, left edge, barline,
  various grouping items) * 2 * 12 staves = 57600

but, to do the substitution, you have to examine *all* pointers (which
means that you have to recurse on lists that are a lot longer than
10000).  Only after this substitution thing, hara kiri removes unused
staves.

> In reality, a lot of those staves are blown away by the HaraKiri mechanism.
> Would it help you guys if I hacked in some kind of counter?

lilypond -V gives element counts 

> John

-- 

Han-Wen Nienhuys   |   address@hidden    | http://www.cs.uu.nl/~hanwen/