LM fonts (version 0.98.3): Bug report 001
I have prepared the first bug report with 'a visual explanation'
for the 0.98.3 version of the LM fonts.
I verified the quality of glyph outlines (or the quality of conversion
to outline curves) and some other global features like hinting.
I had generated and then have checked the glyph proofsheets.
Because the files with complete proofsheets are large or huge
they are not usually available on this web page.
I have selected the cases significant or/and important to illustrate
some principal effects.
The aim of this my first attempt was not a completeness.
And also I 'publish' only a short sample part
of a glyph bug list (TXT) here.
Mistakes in glyph outlines
A general comment
What is a bug in a glyph outline contour curves? (a local meaning)
The outline font glyph contains elements (points, curves, straight lines,
corners, arcs, bumps, holes, steps, etc.) absent in the original script
design or in the corresponding METAFONT source
elements present in the METAFONT definition are missing
in the result of an outline conversion.
The fonts need a complete revision:
missing extrema points - wrong (vertical) hinting !!!
The complete glyph proofsheet files are huge. Therefore only
lmbo10 (PDF) (>2 MB) and
the METATYPE1 form (PDF) (>4 MB)
proofsheets are presented here to demonstrate this crucial error.
Single glyph bugs
Several (probably new) additions/extensions have not been converted
to outlines accurately.
Verification of occurrences of such glyphs in all the fonts is strongly
recommended. I suppose we are interested in the precise fonts (like
as in the precise TeX relative products in general) and any elements
with 'alpha-version' artifacts should be improved in the 'official'
The following example illustrates a proper conversion of the METAFONT
source to contours (the 'wings' of arrows have the sharp corners,
the stem has no arcs):
lmbx12: correct arrowleft v.s.
lmbx6: wrong arrowleft
(bumps and arcs are wrong because they are absent in MF!).
A selected list of glyph names I recommend to check
(with the complete corresponding proofsheet files in PDF):
The result of a mutual comparison -
an inconsistence may be a signal that something is not correct;
the outline curve topology cannot be different for various sizes
of the same font. The serifs in the lmro* font group
Good serif (lmro9) and
bad serif (lmro10).
or with more details:
lmro10: u before and
after modification and simplification,
the modified glyph has better agreement with the MF source,
is 'cleaner' and has the smaller number of points (invalid points was
The following overview lists several
individual glyph contour artifacts:
serifs in lmro*; bumps in nine, six an oldstyleone, etc.;
lmbxi10: Eng is under a question.
Good Oslash (lmro9) and
bad Oslash (lmro10, lmro12).
A special section for Vietnamese (VN)
shows a common ('rounding') problem
of correlations between simplification of
glyph outlines and rounding coordinates to integer.
These operations must be done in a 'proper' order depending
on an individual features of glyph outline contour curves
(attention: short segments in non-integer representation!).
Inaccurate conversion of hookabove
in lmbxi10 and a couple of fonts demonstrates that problem.
(Please zoom the details in Acrobat).
An acceptable conversion (in the 'non-slanted' lmr10)
and how to do the conversion better
for lmbxi10 shows the next file
- please compare the both lmbxi10 hookabove representations.
Other glyphs in VN being not 'clean' (I mean):
Karel Piska, email@example.com
Created: 20 Apr 2005
Last modified: 21 Apr 2005