Re: rendering differences if using `-j12 CPU_COUNT=12`

[Jean Abou Samra]

Le 01/05/2021 à 11:33, Thomas Morley a écrit :
> Am Sa., 1. Mai 2021 um 11:16 Uhr schrieb Jonas Hahnfeld via
> Discussions on LilyPond development<lilypond-devel@gnu.org>:
> > Am Samstag, dem 01.05.2021 um 10:16 +0200 schrieb Jean Abou Samra:
> > > Since commit 4751d0a66e97a61c763c150bb38f0446cc945f4d,
> > > ly/turkish-makam.ly contains this:
> > >
> > > #(set! standard-alteration-glyph-name-alist
> > >          (append turkishMakamGlyphs
> > >                  standard-alteration-glyph-name-alist))
> > >
> > >
> > > which changes the shared variable
> > > standard-alteration-glyph-name-alist, used by the
> > > \sharp markup command among others.
> > >
> > > [...]
> > >
> > > This means that if a snippet using turkish-makam.ly
> > > is compiled, in the same session, before a file
> > > using \markup \sharp, the glyph changes.
> > >
> > > Basically the mechanism should be changed so that
> > > set! is no longer needed.
> > I think all that's needed is changing
> >      (define-public standard-alteration-glyph-name-alist
> > to
> >      (define-session-public standard-alteration-glyph-name-alist
> > and the session mechanism will take care to restore the initial value
> > after each file (as lang as you don't use destructive operations on
> > that initial value).

Indeed, thanks for reminding me. I'd prefer fixing
the mechanism, making the alist a layout variable
or something so it becomes customizable, but this
could be a good temporary fix.


> > Wouldn't it be better to _not_ reset
> > `standard-alteration-glyph-name-alist' at all?
> > The files hel-arabic.ly, turkish-makam.ly and makam.ly all define some
> > custom-makam-glyph-alist and use this alist in \layout for
> > KeySignature, Accidental, AccidentalCautionary, TrillPitchAccidental,
> > AmbitusAccidental.
> > Is this not sufficient? In other words why not simply delete all
> > occurrencies of:
> > #(set! standard-alteration-glyph-name-alist
> >         (append whateverMakamGlyphs
> >                 standard-alteration-glyph-name-alist))
> > ?
> > Then we would rely only on the mechanism partially demonstrated in
> > input/regression/accidental-ancient.ly
> >
> > I'll test that later this WE

It's slightly more complicated. Chord names use the
alist directly (sigh):

(define-public (alteration->text-accidental-markup alteration)
  (make-smaller-markup
   (make-raise-markup
    (if (short-glyph? alteration)
        0.3
        0.6)
    (make-musicglyph-markup
     ;; FIXME -- use current glyph-name-alist here.
     (assoc-get alteration standard-alteration-glyph-name-alist "")))))

and the markup commands \sharp et allies do the same.

Best,
Jean