Re: Nested contexts - fix

[Rune Zedeler]

Jan Nieuwenhuizen skrev:

> In software there's no such thing as a temporary solution.

Quoting the changelog:

1.3.142.rz1
===========

* Added "hacky" drum notation/midi. See input/tricks/drums.ly and
  ly/drumpitch.ly. Moved midi-instruments to scm.

* Added temporary nested staffgroups/choirstaffs support
  use Context InnerStaffGroup / InnerChoirStaff.

> Forgive my ignorance but is testing for A->B->A cycles not easily
> possible?

Yes.
But then A-B-C-A loops will still not be detected. Testing for 
A-B-C-A-loops is also easy, but then A-B-C-D-A-loops will not be 
detected. Etc.

> We must not have an easy infinite loop attack?

If this is official policy then it was a really bad move to include a 
turing complete language (scheme) in lilypond :-)

%%% BEGIN coredump.ly %%%
#(define (inflist) (cons 42 (inflist)))
#(inflist)
{
  c d e f
}
%%% END coredump.ly %%%

> If we detect a loop, should there not be a warning or something?
> Should this patch not include the removal of Inner*, including
> convert rules?

Yep.
As I said I have git problems and just included the patch here because I 
was unable to commit.
I am in the middle of doing pitches, so doing another non-minor 
restructuring, such as removing a context, is not something I want to do 
at the same time.
I posted the patch here in the hope that somebody would do some testing 
to see whether it works and do not lead to infinite loops or other 
errors. No need to "polish it" before we have verified that it works.
Of course convert-ly-rules should be added, and the superfluous contexts 
should be removed.
And I admit that also testing for level-2-loops is a good idea, so that 
staffgroups and choirstaves can accept each other.

In all cases I think that this should wait till 2.15.
I can imagine that introducing cycles in the context hierarchy could 
introduce some hard-to-find random-segfault-bugs - and it would 
therefore be a bad idea to apply it just before a new stable release.

-Rune