Re: Suspected Bug in Lilypond: pushToTag vs EventCord

[David Kastrup]

Bjørn Bäuchle <lilybugs@frankfurtium.de> writes:

> Happy New Year!
>
> I have come across some odd behaviour with \pushToTag.
>
> See this MWE:
>
> x = \relative c' { \tag #'dyn c \tag #'dyn <c e>4 }
> \relative c' { \pushToTag #'dyn \p \x }
>
> I receive the following message:
>
> test.ly:...: warning: \pushToTag cannot push post-event
> \relative c' { \pushToTag #'dyn
>                                 \p \S }
> test.ly:...: to this music
> S = \relative c' { \tag #'dyn c
>                                 \tag #'dyn <c e>4 }
>
> Investigating, I found the error message in ly/music-functions.ly:1880.
> This code has been introduced in
> fda7efb2254751862897d66c6b27b099fb2975b6 ; the commit message says
> "rhythmic events or articulations can be pushed to chords, and
> articulations only can be pushed to rhythmic events." Now I don't quite
> understand how articulations can be pushed to chords and then only to
> rhythmic events, but in any case passing a chord into add-left which is
> defined in lines 1874-1909 triggers the first condition "(music-is-of-
> type? m 'simultaneous-music)" (line 1877).
>
> Changing the order of the conditions - first checking event-chord, then
> sequential-music and simultaneous-music, i.e., moving lines 1884-1897
> above line 1876, works and allows to push a dynamic mark to an
> EventChord.

Being the one responsible for the code, the reason for this was that I
was completely unaware that an EventChord is not just of music type
event-chord but also of simultaneous-music .

That is kind of unexpected and I am surprised that it doesn't get us
into more trouble.  The question is whether that is useful to any
degree.

A quick scan sees simultaneous-music being checked for twice in
ly/music-functions.ly (probably both instances from the commit you
mention) and then in scm/auto-beam.scm and scm/modal-transforms.scm .

The instance in auto-beam.scm is run on the output of event-chord-reduce
and thus is unlikely to see any event-chord.  The instance in
modal-transforms.scm explicitly catches event-chord before
simultaneous-music .  I doubt either of those instances navigated that
problem consciously.

My vote would be to stop an event-chord from being marked as
simultaneous-music.  Neither are treated as events but are processed by
individual iterators.  And it does not look like this "feature" has been
employed anywhere consciously.

-- 
David Kastrup