#5540 clash between staves

[Werner LEMBERG]

[lilypond git 49f41bf1c64ae1e2490571e73a85cbc4b2a9fb27]

The example below exhibits a clash between staves. It's easy to circumvent, but it definitely shouldn't happen, since it gets triggered by non-musical parameters. Maybe a rounding issue?

% Sizes 17 and 18 also cause a clash.
#(set-global-staff-size 16)

\paper {
  % Commenting out this line removes the clash;
  % values like 5cm and larger also remove the clash.
  indent = 4\cm
}

% The notes below trigger the clash;
% I wasn't successful in simplifying it more.
<<
  \relative c' {
    d2 d4 d |
    d8.[ d16] << { d4 } \\ { <a c> } >>
  }

  \relative c' {
     e2 e4 e |
     e4 <d b'>
  }
>>

[Malte Meyn]

You don’t need another staff-size:

\version "2.21.0"

\paper {
  indent = 10\cm
}

<<
  \new Staff << d' \\ <c' a> >>
  \new Staff <a' b'>
>>

Alternatively, you can keep the default indent and

#(set-global-staff-size 9)

[Werner LEMBERG]

Thanks for the simplification! Here's the corresponding image.

[Thomas Morley]

Interestingly the problem happens only for quaters not for other durations.
And if I extend the quater's X-extent to that of a half note with
\override NoteHead.X-extent = #'(-0.0 . 1.377346)
(applied to Malte's code) the clash vanish.
Ofcourse this is not a workaround (the value would need adjustment for changed staff-sizes, the stem-attachment is off and likely other problems), but may point to the direction where to look for the cause of the clash.

---

And yes, first bad commit is:
commit 87eb2f9fe1be3a532675fe4b7322bbba5a60ba5c (HEAD)
Author: Carl Sorensen carl.d.sorensen@gmail.com
Date: Sat Mar 5 21:30:46 2016 -0700

Use a negative value of overdone_noteheads to shrink the head slightly

Changing recent master to the former value:

$ git diff
diff --git a/mf/feta-params.mf b/mf/feta-params.mf
index 36549e85ed..bb3dc2867f 100644
--- a/mf/feta-params.mf
+++ b/mf/feta-params.mf
@@ -250,7 +250,7 @@ slash_thick# := 2/3 * 0.48 staff_space#;
% reduce the notehead height. Empirically, we have determined that
% reducing by 10% of stafflinethickness solves the problem.

-overdone_heads = -0.1;
+overdone_heads = 0.0;
noteheight# := staff_space# + (1 + overdone_heads) * stafflinethickness#;

define_pixels (slash_thick);

returns not clashing output.
Ofcourse the (here reverted) patch was meant to fix a problem, which now will arise again...