Comment #9 on issue 3041 by [email protected]: Patch: Fix a few segfaulting cases in hara-kiri-group-spanner
http://code.google.com/p/lilypond/issues/detail?id=3041
It has happened to me before that valid input was only ever valid because of magic. It could be that calls to get_column to a non valid column still somehow spit out numbers. It may be the 0 to INT_MAX range that is causing the issue.
The root of the segfault seems to be that this function is called on axis groups that lack left or right bounds that are items, no? I think it'd be better to ensure they have bounds before and issue a nasty programming error if they don't.
