Bug in Lookup::bezier_sandwich

[Jürgen Reuter]

   Hi all,
   there is a bug in Lookup::bezier_sandwich that severely affects ancient
   notation.  This method was originally added to lookup.cc for
   flexa/porrectus support.

   In version 2.14, the bezier_sandwich curve still looks correctly, see
   here:
   http://lilypond.org/doc/v2.14/Documentation/a9/lily-551aed0c.png
   or (with more context) here:
   http://lilypond.org/doc/v2.14/Documentation/notation/ancient-notation

   In version 2.15 and later, the bezier_sandwich curve has zero height at
   its right end, which is bad; see here:
   http://lilypond.org/doc/v2.15/Documentation/bf/lily-ac979051.png
   or (with more context) here:
   http://lilypond.org/doc/v2.15/Documentation/notation/ancient-notation

   I tried to track down the problem and found the following suspicious
   commit:

   commit 35725a573e47be7c02c51964641ea534fb88be6b
   Author: Mike Solomon <address@hidden>
   Date:   Mon Feb 6 15:03:20 2012 +0100
       Gets rid of bezier-sandwich stencil
   diff --git a/lily/lookup.cc b/lily/lookup.cc
   index 3f393e0..7b63b83 100644
   --- a/lily/lookup.cc
   +++ b/lily/lookup.cc
   @@ -449,22 +449,32 @@ Lookup::slur (Bezier curve, Real curvethick, Real
   linethick,
    Stencil
    Lookup::bezier_sandwich (Bezier top_curve, Bezier bottom_curve, Real
   thickness)
    {
   -  /*
   -    Need the weird order b.o. the way PS want its arguments
   -  */
   -  SCM list = SCM_EOL;
   -  list = scm_cons (ly_offset2scm (bottom_curve.control_[3]), list);
   -  list = scm_cons (ly_offset2scm (bottom_curve.control_[0]), list);
   -  list = scm_cons (ly_offset2scm (bottom_curve.control_[1]), list);
   -  list = scm_cons (ly_offset2scm (bottom_curve.control_[2]), list);
   -  list = scm_cons (ly_offset2scm (top_curve.control_[0]), list);
   -  list = scm_cons (ly_offset2scm (top_curve.control_[3]), list);
   -  list = scm_cons (ly_offset2scm (top_curve.control_[2]), list);
   -  list = scm_cons (ly_offset2scm (top_curve.control_[1]), list);
   -
   -  SCM horizontal_bend = scm_list_n (ly_symbol2scm ("bezier-sandwich"),
   -                                    ly_quote_scm (list),
   +  SCM commands  = scm_list_n (ly_symbol2scm ("moveto"),
   +                              scm_from_double
   (top_curve.control_[0][X_AXIS]),
   +                              scm_from_double
   (top_curve.control_[0][Y_AXIS]),
   +                              ly_symbol2scm ("curveto"),
   +                              scm_from_double
   (top_curve.control_[1][X_AXIS]),
   +                              scm_from_double
   (top_curve.control_[1][Y_AXIS]),
   +                              scm_from_double
   (top_curve.control_[2][X_AXIS]),
   +                              scm_from_double
   (top_curve.control_[2][Y_AXIS]),
   +                              scm_from_double
   (top_curve.control_[3][X_AXIS]),
   +                              scm_from_double
   (top_curve.control_[3][Y_AXIS]),
   +                              ly_symbol2scm ("curveto"),
   +                              scm_from_double
   (bottom_curve.control_[2][X_AXIS]),
   +                              scm_from_double
   (bottom_curve.control_[2][Y_AXIS]),
   +                              scm_from_double
   (bottom_curve.control_[1][X_AXIS]),
   +                              scm_from_double
   (bottom_curve.control_[1][Y_AXIS]),
   +                              scm_from_double
   (bottom_curve.control_[0][X_AXIS]),
   +                              scm_from_double
   (bottom_curve.control_[0][Y_AXIS]),
   +                              ly_symbol2scm ("closepath"),
   +                              SCM_UNDEFINED);
   +
   +  SCM horizontal_bend = scm_list_n (ly_symbol2scm ("path"),
   scm_from_double (thickness),
   +                                    ly_quote_scm (commands),
   +                                    ly_quote_scm (ly_symbol2scm
   ("round")),
   +                                    ly_quote_scm (ly_symbol2scm
   ("round")),
   +                                    SCM_BOOL_T, SCM_UNDEFINED);

      Interval x_extent = top_curve.extent (X_AXIS);

   I do not fully understand the rationale / implications of this change,
   so I do really know what to do here without affecting other places in
   the code.  By the way, ancient notation does not make use of the
   "thickness" argument; probably it has been introduced for some other
   use elsewhere.

   Could someone of the active developers look into this?  That would be
   great!

   Thanks a lot,
   Juergen