\version "2.12.1"

% A mode that contains a quarter tone generates the warning: % "warning: No ordering for key signature alterations"

% and if a \midi { } statement is included this error is also generated: % "lily-library.scm:149:5: Wrong type (expecting exact integer): -1/2"

% Here's a minimal file to demonstrate the problem:

mode = #`((1 . ,SEMI-FLAT)) \score { \relative c'' { \key c \mode c4 } \layout { } \midi { } }

% This example is musical nonsense, but the same construct is % used, for example, in the bayati mode in ly/arabic.ly, and % generates similar errors.

The first warning arises from the lack of quartertones in the standard alist for keyAlterationOrder; setting it to #mode gets rid of the message:

mode = #`((1 . ,SEMI-FLAT)) \score { \relative c'' { \set Staff.keyAlterationOrder = #mode \key c \mode c4 } \layout { } %\midi { } }

keyAlterationOrder specifies which accidentals to display in the key signature and the order in which they appear. Using the mode definitions in ly\arabic.ly as they stand for this generates key signatures containing accidentals for all pitches, and in the wrong order. So should the present definition of, for example, bayati be changed from

bayati = #`( (0 . 0) (1 . ,SEMI-FLAT) (2 . ,FLAT) (3 . 0) (4 . 0) (5 . ,FLAT) (6 . ,FLAT) )

to

bayati = #`( (6 . ,FLAT) (2 . ,FLAT) (5 . ,FLAT) (1 . ,SEMI-FLAT) ) ?

No, ideally the file would have a separate keyAlterationOrder alist for each mode if necessary (assuming arabic key signatures are standardized enough to have particular orderings of accidentals).

A patch addressing both "no ordering" for Arabic modes, and the crash with midi output, is at http://codereview.appspot.com/4237041/

The midi portion of the patch is due to Graham Breed, http://lists.gnu.org/archive/html/lilypond-devel/2011-02/msg00616.html

LGTM, nothing obviously wrong.

Fixed with commit 9513397, due to appear 2.13.53.

The midi error is resolved. Using pre-defined modes like \key d \bayati now gives no warnings. User-defined modes might still produce a warning, so that warning now includes "keyAlterationOrder" to help find in the manual the advice of comment 3.