Add regtest for issue 5181

Add regtest for issue 5181


Also contains:

Define ly:music-error

[Carl, 2017-09-25 04:18:07]

LGTM

[thomasmorley651, 2017-09-25 21:28:50]

Very nice, i.e LGTM

Not sure whether this should be reflected by a reg-test, but currently there is
no case in it triggering the warning like:

warns = { 
  \override NoteHead.color = #red 
  c'1 d'
  \revert NoteHead.color
}

{ \warns -1 e' }

And there are cases which will probably never work, like:

fails = {
  d'-\tweak color #red 
     %% with or without \etc
     \etc
}

{ \fails -1 e' }


Not sure if we should mention it in the docs at all, or in "Known issues and
warnings" or elsewhere ...?

[dak, 2017-09-25 21:53:06]

On 2017/09/25 21:28:50, thomasmorley651 wrote:
> Very nice, i.e LGTM
> 
> Not sure whether this should be reflected by a reg-test, but currently there
is
> no case in it triggering the warning like:
> 
> warns = { 
>   \override NoteHead.color = #red 
>   c'1 d'
>   \revert NoteHead.color
> }
> { \warns -1 e' }
> 

Touché.  I was not sure about whether or not the behavior and warning
for << {c'} -1 >> should be retained.  But your case looks pretty bad.
I mean, sure, one could skip backwards over unusable music expressions
as long as their length amounted to zero...  But it seems like
inviting trouble.

> And there are cases which will probably never work, like:
> 
> fails = {
>   d'-\tweak color #red 
>      %% with or without \etc
>      \etc
> }
> 
> { \fails -1 e' }

Ugh.  That's another of those "what do you even think \etc is" cases.
Basically you need to come up with a reasonable definition of the
expression

{ ... \etc }

first.  I mean, the syntax does not yet seem taken so that does not
seem impossible but it may still end up being tricky.

> Not sure if we should mention it in the docs at all, or in "Known
> issues and warnings" or elsewhere ...?

So far I think it's more a case of reviewing "Known issues and
warnings" for stuff that is no longer an issue.  I don't think that we
have rigidly defined the limits of where you could place post-events,
but we may have warned people off a few things.

[thomasmorley651, 2017-09-26 00:04:34]

On 2017/09/25 21:53:06, dak wrote:
> On 2017/09/25 21:28:50, thomasmorley651 wrote:
> > Very nice, i.e LGTM
> > 
> > Not sure whether this should be reflected by a reg-test, but currently there
> is
> > no case in it triggering the warning like:
> > 
> > warns = { 
> >   \override NoteHead.color = #red 
> >   c'1 d'
> >   \revert NoteHead.color
> > }
> > { \warns -1 e' }
> > 
> 
> Touché.  I was not sure about whether or not the behavior and warning
> for << {c'} -1 >> should be retained.  But your case looks pretty bad.
> I mean, sure, one could skip backwards over unusable music expressions
> as long as their length amounted to zero...  But it seems like
> inviting trouble.
> 
> > And there are cases which will probably never work, like:
> > 
> > fails = {
> >   d'-\tweak color #red 
> >      %% with or without \etc
> >      \etc
> > }
> > 
> > { \fails -1 e' }
> 
> Ugh.  That's another of those "what do you even think \etc is" cases.

This one is easy.
\etc is a blackbox containing all my wishes to christmas.
;)

> Basically you need to come up with a reasonable definition of the
> expression
> 
> { ... \etc }
> 
> first.  I mean, the syntax does not yet seem taken so that does not
> seem impossible but it may still end up being tricky.
> 
> > Not sure if we should mention it in the docs at all, or in "Known
> > issues and warnings" or elsewhere ...?
> 
> So far I think it's more a case of reviewing "Known issues and
> warnings" for stuff that is no longer an issue.  I don't think that we
> have rigidly defined the limits of where you could place post-events,
> but we may have warned people off a few things.

Don't get me wrong, I don't suggest to make those examples work.
But we should think about increasing "naive" user expectations after 5181 will
be released, imho.

5181 is great, but has limitations, we should be prepared to explain them, i.e.
point to relevant documentation or the like.