Re: [Lilypond-auto] Issue 2657 in lilypond: font kerning on Windows is broken

[lilypond]

Comment #5 on issue 2657 by address@hidden: font kerning on Windows  
is broken
http://code.google.com/p/lilypond/issues/detail?id=2657

I've done some more investigation into this.  I'm assuming it's related to  
Issue 2656.  Anyway, looking at the .ps output from Windows and Linux, we  
see, for example, this on a PS file from Windows:

1.8969 -4.8157 moveto /CenturySchL-Roma 11.59570312 output-scale div  
selectfont
1.7072 0.0000 0.0000 /D
0.8991 0.0000 0.0000 /I
1.4682 0.0000 0.0000 /L
1.5934 0.0000 0.0000 /A
1.5934 0.0000 0.0000 /V

The same output from Ubuntu is this:

1.8969 -4.8157 moveto /CenturySchL-Roma 11.59570312 output-scale div  
selectfont
1.7072 0.0000 0.0000 /D
0.8991 0.0000 0.0000 /I
1.4682 0.0000 0.0000 /L
1.5934 0.0000 0.0000 /A
1.3771 0.0000 0.0000 /V

In the Windows example, the widths of A and V are the same.  On Ubuntu, V  
is narrower.  The widths seem to be set in pango-font.cc, with this call:

pango_glyph_string_extents_range (pgs, i, i + 1, pa->font, &ink_sub_rect,  
&logical_sub_rect);

I added some debug code to this to print out PANGO_RBEARING  
(logical_sub_rect).  On Ubuntu, I get this:

warning: 41984.000000
warning: 48128.000000
warning: 44032.000000
warning: 26624.000000
warning: 51200.000000

On Windows:

warning: 48128.000000
warning: 48128.000000
warning: 44032.000000
warning: 26624.000000
warning: 51200.000000

So it looks like pango_glyph_string_extents_range returns a kerned value of  
V for Ubuntu, but not for Windows.  I was out of my depth before I looked  
into this, but I'm well out of it now.  I assume the pango call is a call  
to the pango library, and this is returning different results on the 2  
platforms.  Does this seem correct, and should it be reported as a bug to  
Pango?