Re: Wrong type in part-combiner.scm

[Dan Eble]

On Oct 5, 2020, at 12:02, David Kastrup <dak@gnu.org> wrote:
> 
> Jonas Hahnfeld <hahnjo@hahnjo.de> writes:
> 
>> Am Montag, den 05.10.2020, 10:55 -0400 schrieb Dan Eble:
>>>>> Absolutely no idea which is correct. Sorry. :/
>>>> 
>>>> Dan? I think this issue started occuring after a change of yours,
>>>> so do you have an idea about the meaning of raw-voice becoming #f
>>>> here?
>>> 
>>> I'll take a look.
>> 
>> I think this has been taken care of by
>> https://gitlab.com/lilypond/lilypond/-/merge_requests/427 ?
> 
> That deals with the symptom.  It doesn't address the meaning of
> raw-voice becoming #f.  The changed function is now more robust against
> different input, but I have no idea if the different input constitutes a
> bug.
> 
> So it would be good if Dan checked whether the current behavior is the
> desired one.

[Moving discussion to lilypond-devel.]

After some investigation, I'd like to ask the other contributors some questions 
because I don't see very strong reasons to go one way or the other.

A. Should context-specific music with no music inside it create the context or 
not?  Right now, it doesn't.  I'm slightly in favor of changing it because it 
would allow reverting David's recent fix and one of my earlier ones and it 
seems that it might raise fewer questions; however, changing this would have 
some consequences (item B and perhaps others).

B. Which conditions should trigger the "no music found in score" warning?  Are 
the cases in input/regression/no-music.ly working as desired?  Would it be a 
bad change if \new Voice {} were considered music?  If that were a problem, 
maybe we could broaden the test to warn about music with no duration (grace or 
main).

Your thoughts?

Regards,
— 
Dan