Re: [Devel] Bug in ttinterp.c ???

[Rogier van Dalen]

> > At g0018, the program calls function 6. At f0398, the MSIRP tries to
> > move a point horizontally at a distance from point 0 in the twilight
> > zone. The y coordinate, however, seems to be set to the twilight point y
> > coordinate, which is 0. It seems to me this has got something to do with
> > the /* XXX: UNDOCUMENTED! behaviour */ part in Ins_MSIRP, which I don't
> > understand at all. This seems to happen at all sizes.

> As reported by other posters on the list, this might be a bug in the
> TrueType bytecode interpreter that comes with FreeType. Several patches
> have been submitted but we need to test them extensively to see if these
> do not degrade the rendering of several popular high-quality fonts. I
> hope to do this kind of work this or next week..

Do you have any idea what the reason was for adding the "undocumented" bit?
Surely it must have been a specific font that used this feature?

Is there a patched version in the CVS, or could you send me one for testing?

Two more questions about (or rather, feature requests for) the interpreter:
1. Is it possible to make the MPS (Measure Point Size) instruction return a
sensible value? (Windows 95 & 3.1 just return 12, but Windows 2000 (and
probably other versions) returns the real point size). This enables fonts to
implement optical size features very easily in fonts, so it would be good to
have FreeType using this as well.

2. Is it possible to make the GETINFO instruction return whether
anti-aliasing is enabled or not? In the past, fonts have presumed no
anti-aliasing is used, but this is not the case anymore, I'd guess. I think
Windows rasterisers supporting aa do all return this information. Since
FreeType doesn't use the GASP table, it would be sensible to give the
instructions information about whether or not aa is used.

Thanks,

Rogier van Dalen