Fix #9352: Add ability to guess tie attachment point for different noteheads

[asattely]

Resolves: #9352

Noteheads are now taken into account when determining where to attach ties. For noteheads with cutout information, this is the average of the cutouts. Also, now that noteheads have a single attachment point, ties adjust for each other when they share that point:

For slash noteheads, which are special, the tie attachment point takes the stem anchor into account:

For all other noteheads, the exact center is used.

[asattely]

Added some extra space for the ties to adjust to each other

[Jojo-Schmitz]

3 vtests show differences, all with ties, unsurprisingly. 3 look OK, 3 don't:
Before:

After:

The tie in the 2nd measure is too high, IMHO. On the other hand the 1st tie, grace to note, looks better now, but IMHO too low.
Before:

After:

In top staff the 1st tie in the 1st measure is too high IMHO, in bottom staff the last ties in the 1sxt meaure seem to be to high/low, not really sure there.
But the worst, before:

After:

The bottom tie starts way too late. It started to late before already, but now even later.

(Sorry for transparent PNG, but that's what the vtests produce)

[Jojo-Schmitz]

Actually there's one more, before:

After:

See the tie for the 1st grace to note (was the case before already) and the 2 gracenotes at the end, they are too far to the right.
Seems the smallness of notes need to get taken into considaration?

[asattely]

Yup, I noticed those as well. Thanks for looking through and getting the screenshots, I'll definitely be taking another whack at this once I finish something I'm working on tomorrow.

[asattely]

So, the extra height on some of the ties were due to the fact that the vtests still use Emmentaler rather than Leland, for the record. The vtests will almost certainly still fail because the ties will look different, but now the ties from gracenotes will be a lot nicer with a little special handling. The wholenotes one was a weird oversight that should be fixed now. Let's see what the vtests say on this one.

edit: a few fixes still needed, should be easy

[Jojo-Schmitz]

Looks better, but this here does not, before:

after:

before:

after:

The 1st tie on each looks better (but to narrow IMHO), the other 3 don't, they go too far past the note they tie to

[asattely]

Yep, very well aware. The problem is that headWidth (and, bafflingly, the bounding box as well) always returns the width of a regular-sized notehead instead of the grace-sized one, so I just need to figure out how to get the grace-sized head width and we're golden.

[Jojo-Schmitz]

See (in 3.x) Format > Style > Sizes > Grace note size. Defaults to 70%, so you'ds need to pick on that style setting

[asattely]

The answer was mag() (or magS()) which holds information about the note's relative size

[Jojo-Schmitz]

Still the issue with these too narrow ties (same as in the above images), they are not going from/to center of notehead

[asattely]

Those are inside-style ties because they are going to or from a multinote chord. This is an issue in other inside-style tie situations as well, and in the future there will be a minimum tie width issue where that will be looked at in greater detail.

[asattely]

Still messing around with different fonts and noteheads, there are so many combinations to go through..!

[asattely]

Rewrote the Y adjust code to be simpler and also more robust in the face of different fonts. Still needs a rebase.

[asattely]

Rebase finished. I was also requested to add a small left- or right-offset to outside tie endpoints, which I ended up making 1/8 sp.

edit: i think i know what is killing the utests, give me a sec.

[Jojo-Schmitz]

4 utests crash with a failed assertion:
ASSERT: "_anchor == Anchor::CHORD" in file .../src/engraving/libmscore/spanner.cpp, line 879
Not sure how (or whether) this relates to this PR?

Edit: I see, you're on it

[asattely]

Tests should be fixed now

[Jojo-Schmitz]

Nope, exact same crash

[asattely]

The weird thing is that these tests pass when I run them locally. I think I identified the problem, we'll see if the tests are good this time.

edit: There we go. The problem was that Tie::startChord() doesn't actually return the correct chord, so you have to use Tie::startNote()->chord() in order to access it. I'm not sure why this is the case, but changing it to the latter fixes it, so it's all good now.

[AntonioBL]

The vtests are now better, except for "grace-5-1".
Before

After

First, second and third ties are better than the reference, but there is something strange in the vertical position of the tie between the two grace notes (i.e. the fourth tie in the file).

Ciao

[Nick-Mazuk]

Does this work as expected on chords where every note has a different notehead?

[asattely]

Fixed a little issue with forced-direction outside ties colliding with hooks, and also added another special case for slashed noteheads, which are different from slash noteheads for the record.