Re: [ft-devel] longjmp issue 1

[Behdad Esfahbod]

On 14-12-07 10:20 PM, Werner LEMBERG wrote:
>> > I highly suggest you remove ftvalid (and gxvalid).
> They are not used in a default build.
> 
>> > It's unused, which makes it untested, and as such broken.
> `Not in the default build ' means `broken'?  You are exaggerating.

Oh no, I mean unused by widely used+tested clients, like the Linux desktop,
Android, Chrome, Firefox, etc.


>> > OTS does the same thing, but is actually used in Firefox and Chrome,
>> > and as a result tested in practical situations.  But even that was
>> > very broken in certain ways around GSUB/GPOS and we have just got to
>> > a point that it can be called ok.
> Thanks for the pointer to OTS.  Can you elaborate on the issues you've
> experienced with validation?

Many small things, as discovered over time by various "invalid" fonts that do
certain things.

Ok, for example, invalid glyph IDs are a perfectly valid use in GSUB as
intermediate values.  Currently ftvalid downright rejects those.


>> > A more extensive font lint tool is a great idea, and one that is
>> > sorely missing right now.  But FreeType isn't the place to do it.
> Well, I could move those two modules to a different place, creating
> another library, for example, but what are the benefits?  Just look at
> the work to be done to split HarfBuzz into two libraries so that the
> chicken-and-egg problem with FreeType could be solved.  It's a lot of
> work, and the benefits are small since compiling FreeType, then
> HarfBuzz, then FreeType again is not rocket science and works without
> troubles.

For a lint tool I think using a Python framework like FontTools makes much
more sense.

Cheers,
-- 
behdad
http://behdad.org/