Re: [ft] Issues with fonts embedded in PDF

[Jeremias Maerki]

After identifying which fonts really caused the bug, I finally found the
problem: The "cvt ", "fpgm" and "prep" table all seem to be optional for
TrueType fonts and therefore not all fonts have them. Apache FOP handled
those optional tables incorrectly (preparing the table entry but not
filling it). I ran FOP over my complete set of fonts I have accumulated
over time and all are finally displayed correctly in
GhostScript/GhostView (on Windows). I wasn't able to reproduce the
errors in evince or xpdf (on Ubuntu) which kind of surprised me. But
having GS as an indicator should be good enough.

Change in FOP:
http://svn.apache.org/viewvc?rev=616691&view=rev

Thanks everyone for your help in tracking down the problems!

On 29.01.2008 18:55:24 Derek B. Noonburg wrote:
> On 29 Jan, George Williams wrote:
> > On Mon, 2008-01-28 at 11:18, Derek B. Noonburg wrote:
> >> I had another report of this problem a month or two ago.  It seems to be
> >> a bug in FontForge.  In that case, someone was converting metafont fonts
> >> to TrueType, using mftrace, which uses FontForge.
> > 
> >> If you've already tried the latest version of FontForge, I would suggest
> >> reporting the bug to them.  Please feel free to forward this message,
> >> and they're welcome to contact me if they have questions on the info
> >> above.
> > Um... how does fontforge enter the picture in this problem? FontForge
> > doesn't generate pdf files (well, not beyond font samples). If you have
> > a fontforge bug, please explain it in greater detail.
> 
> The original email (not sure if you saw it, sorry) said that the fonts
> were converted from OTF to TrueType using FontForge.
> 
> In the earlier bug report, someone was using mftrace to convert
> MetaFonts to TrueType.  According to the mftrace web page, it uses
> FontForge to generate TTF output.
> 
> TrueType fonts can be embedded directly in PDF files, so my assumption
> was that the TTF error was being introduced by FontForge.
> 
> But it looks like both cases were also using fop as the last stage (to
> actually generate the PDF file), so the bug may actually have been in
> fop.
> 
> Jeremias - can you look at this?  I know you found one bug in fop
> yesterday (regarding the numTables problem).  I'm wondering if the other
> bug (invalid table directory entry - all zeros other than the tag) is
> also in fop.
> 
> - Derek




Jeremias Maerki