Re: [ft-devel] Type 1C font that is not rendered correctly

[Albert Astals Cid]

A Diumenge 27 Juliol 2008, vàreu escriure:
> > > This font is buggy.  I assume that you've taken it out of a PDF
> > > file, and Acroread is displaying it fine, right?
> >
> > Right,
>
> Sigh.  I really wonder whether I shall break the specification just to
> support this particular font...

Well, i'd say it all boils down to how much the breakage implies in terms of 
work to fix it and if it can break correct fonts and how difficult it'll be 
to maintain in the future.

If all that factors are not much negative i think having a happy user is worth 
it :-)

>
> Can you send me the PDF (or a link to it) with this font?

Here it goes.

https://bugs.freedesktop.org/attachment.cgi?id=17908

Albert

>
> > can you point which the problem is?  Just to make sure that we are
> > not extracting it badly from the PDF file.
>
> For example, the glyph of digit `4' looks like this after
> desassembling
>
>  glyph index 12:
>   1000 387 hmoveto
>   163 vlineto
>   69 hlineto
>   47 vlineto
>   -69 hlineto
>   449 vlineto
>   -35 hlineto
>   -311 -441 rlineto
>   -55 vlineto
>   298 hlineto
>   -163 vlineto
>   -207 567 vmoveto                <---
>   -357 vlineto
>   -251 hlineto
>   endchar
>
> (You can get this with saying `FT2_DEBUG=any:7 ftview -f 12 20 font0',
> using the current CVS.)
>
> The marked line is the bug: According to the specification, `vmoveto'
> takes only single parameter, not two.  To cite 5177.Type2.pdf:
>
>   A number, decoded from a charstring, is pushed onto the Type 2
>   argument stack.  An operator expects its arguments in order from
>   this argument stack with all arguments generally taken from the
>   bottom of the stack (first argument bottom-most); [...]
>
> The specification further says:
>
>   [...], in general, arguments are not accumulated on the Type 2
>   argument stack for later removal by a sequence of operators,
>   arguments generally may be supplied only for the next operator.
>   Notable exceptions occur with subroutine calls and with arithmetic
>   and conditional operators.  [...]
>
>   [...]
>
>   Many operators take their arguments from the bottom-most entries in
>   the Type 2 argument stack; this behavior is indicated by the stack
>   bottom symbol ‘|-’ appearing to the left of the first argument.
>   Operators that clear the argument stack are indicated by the stack
>   bottom symbol ‘|-’ in the result position of the operator
>   definition.
>
> Finally, the `vmoveto' operator is defined like this:
>
>   vmoveto:
>     |- dy1 vmoveto   ===>   |-
>
> And this is exactly what FreeType does: It takes the bottom-most
> argument (-207) and clears the stack afterwards.
>
>
>     Werner