ttf2pt1-devel Mailing List for TTF to PS Type 1 Font Converter

I have an updated ttf2pt1 at https://github.com/william8000/ttf2pt1
It fixes an issue dropping upper case 'A' in some OTF fonts, and it fixes a number of compile warnings, including for C99.
William

We are trying to port Fedora to a newer version of the C language, with
the goal to eventually change compiler defaults:

  <https://fedoraproject.org/wiki/Changes/PortingToModernC>
  <https://fedoraproject.org/wiki/Toolchain/PortingToModernC>

The ttf2pt1 collection of programs (our version is based off 3.4.4) will
fail to build with future compilers because it relies on implicit ints
and implicit function declarations.  The attached patch fixes that.

Thanks,
Florian

> Is there some algorithm in FreeType or some there else which
> simplifies outlines with bezier curvers preservation?

AFAIK, ttf2pt1 can't do that.  It is a non-trivial algorithm to
convert from overlapping to non-overlapping contours.  It would be
great if someone could contribute such a module to FreeType.


    Werner

Hello Werner,

On Mon, Apr 29, 2002, Werner LEMBERG wrote:
> 
> >   But still produce the odd *.afm file, my example is bsmi00lp.ttf.
> >   in my bsmi00lp06.afm,
> >
> > C 27 ; WX 1000 ; N ZdZ2708 ; B 41 322 956 25886 ;
> >                                           ^^^^^
> >
> >   So the bsmi00lp06.tfm(from afm2tfm) cannot use by TeX,
> >
> > % Args: -GE -pft -Ohub -W4 -L cubig5.map+06 bsmi00lp.ttf bsmi00lp06
> 
> Using the CVS of ttf2pt1 from Apr. 22nd, and using the CVS from
> FreeType 2 of today, and using exactly your command line, I get the
> right result:
> 
>   ...
>   C 26 ; WX 1000 ; N _d_2433 ; B 30 -83 964 844 ;
>   C 27 ; WX 1000 ; N _d_2708 ; B 41 -47 956 818 ;
>   C 28 ; WX 1000 ; N _d_2711 ; B 25 -89 969 822 ;
>   ...

  It might be off topic, sorry. But I DID find a strange thing.
  * maybe just my ignorance.

  yes, if just one line command, then everything is OK. But if
  you try the sh script the weird thing happen(see attachment).

> I'm working on a Linux box, and I've compiled everything with gcc
> 2.95.3 (using `-O0 -g' in case this matters).

  Yes, I try gcc-2.95.3 and gcc-2.96(Red Hat version), but seems
  not the gcc's fault.

  Thanks.


Rgds,
Edward G.J. Lee

>   But still produce the odd *.afm file, my example is bsmi00lp.ttf.
>   in my bsmi00lp06.afm,
>
> C 27 ; WX 1000 ; N ZdZ2708 ; B 41 322 956 25886 ;
>                                           ^^^^^
>
>   So the bsmi00lp06.tfm(from afm2tfm) cannot use by TeX,
>
> % Args: -GE -pft -Ohub -W4 -L cubig5.map+06 bsmi00lp.ttf bsmi00lp06

Using the CVS of ttf2pt1 from Apr. 22nd, and using the CVS from
FreeType 2 of today, and using exactly your command line, I get the
right result:

  ...
  C 26 ; WX 1000 ; N _d_2433 ; B 30 -83 964 844 ;
  C 27 ; WX 1000 ; N _d_2708 ; B 41 -47 956 818 ;
  C 28 ; WX 1000 ; N _d_2711 ; B 25 -89 969 822 ;
  ...

I'm working on a Linux box, and I've compiled everything with gcc
2.95.3 (using `-O0 -g' in case this matters).


    Werner

Thanks Werner,

On Sun, Apr 28, 2002, Werner LEMBERG wrote:
> >   freetype2-current(the same with CVS version).
> > 
> >   The weird thing is, everything is Ok, no error messages,
> >   the *.t1a/*.pfb only have ascii characters no Chinese fonts
> >   inside.
> 
> I found the problem.  The new charmap routines contain validation code
> for TrueType cmaps.  Unfortunately, all (4,3) cmaps in the Arphic
> fonts are buggy, not passing the tests.  Consequently, no Unicode
> cmap is provided by FT2.
> 
> Please compile FT2 with FT_CONFIG_OPTION_USE_CMAPS undefined to use
> the old code, then it should work again.

  It works well thanks. Meanwhile the same problem is solved
  on my mlterm. Oh, it's too hard to me to find out this problem.

  But still produce the odd *.afm file, my example is bsmi00lp.ttf.
  in my bsmi00lp06.afm,

C 27 ; WX 1000 ; N ZdZ2708 ; B 41 322 956 25886 ;
                                          ^^^^^

  So the bsmi00lp06.tfm(from afm2tfm) cannot use by TeX,

! Font C00/aming/m/n/14.4/06=bsmi00lp06 at 14.4pt not loadable: \
Bad metric (TFM) file.

  My ttf2pt1 arguments,

%%CreationDate: Sun Apr 28 21:41:51 2002
% Converted by ttf2pt1 3.4.1-CURRENT/ft
% Args: -GE -pft -Ohub -W4 -L /usr/local/share/ttf2pt1/maps/cubig5.map+06 bsmi00lp.ttf bsmi00lp06
%%EndComments

> We will try to find a solution to the problem.

  Hope there are no side effect if I undef the FT_CONFIG_OPTION_USE_CMAPS.

> 
>     Werner
> 
> 
> PS: You must use the latest CVS version; right now I've fixed some
>     typos which prevented compilation if FT_CONFIG_OPTION_USE_CMAPS
>     isn't set.

Rgds,
Edward G.J. Lee

>   freetype2-current(the same with CVS version).
> 
>   The weird thing is, everything is Ok, no error messages,
>   the *.t1a/*.pfb only have ascii characters no Chinese fonts
>   inside.

I found the problem.  The new charmap routines contain validation code
for TrueType cmaps.  Unfortunately, all (4,3) cmaps in the Arphic
fonts are buggy, not passing the tests.  Consequently, no Unicode
cmap is provided by FT2.

Please compile FT2 with FT_CONFIG_OPTION_USE_CMAPS undefined to use
the old code, then it should work again.

We will try to find a solution to the problem.


    Werner


PS: You must use the latest CVS version; right now I've fixed some
    typos which prevented compilation if FT_CONFIG_OPTION_USE_CMAPS
    isn't set.

David Jarvie wrote:
> 
> I had to edit the name of one of the include files in ft.c in order to be able
> to build ttf2pt1-3.4.0 from source. The edit was to rename ftsnames.h to
> ftnames.h . I am running on a Linux system with freetype2 2.0.1 (SuSE RPM)
> installed.

It was one of the differences introduced in freetype somewhere 
between 2.0.1 and 2.0.4. ttf2pt1-3.4.0 expects ttf2pt1-2.0.4.

-SB

I had to edit the name of one of the include files in ft.c in order to be able 
to build ttf2pt1-3.4.0 from source. The edit was to rename ftsnames.h to 
ftnames.h . I am running on a Linux system with freetype2 2.0.1 (SuSE RPM) 
installed.

--
David Jarvie  <dj...@li...>

Werner LEMBERG wrote:
> 
> > The daily CVS archive is availabel from the Download page.
> > There is also a link to the on-line repository browser among
> > the list of pages (on each page):
> >
> > Background | What Is TTF2PT1 | History | Download | CVS | Credits | Todo
> >                                                     ^^^
> 
> Hehe.  This link is broken (at least if you start with
> http://ttf2pt1.sf.net), so I've assumed that no public CVS is
> available.

I've fixed that link, this time hopefully to a more stable
page.

-SB

Roland Mainz wrote:
> 
> Hi !
> 
> ----
> 
> I have a problem with ttf2pt1 - trying to convert parts of japanese
> truetype fonts (get them from
> http://puck.informatik.med.uni-giessen.de/people/gisburn/ttf2pt1_test_sergey_babkin_ja_ttf.tar.gz)
> causes ttf2pt1 to "hang" (I killed it after one hour) ...
> 
> Any ideas what may go wrong here ?

There was a stupid bug in the code that tries to guess the boldness
of the font by its names: two nested loops used the same
variable. I've fixed it, the new snapshot ttf2pt1-3.4.1-SNAP-020422.tgz
is in the usual place (or you can use CVS). Thank you for the
bug report!

-SB

Roland Mainz wrote:
> 
> Edward Lee wrote:
> 
> > > An attempt to use these fonts always ends like this:
> > > -- snip --
> > > % xfd -fn '-hebrew-times-medium-r-normal--0-0-0-0-p-0-iso8859-8'
> > > X Error of failed request:  BadAlloc (insufficient resources for
> > > operation)
> > >   Major opcode of failed request:  45 (X_OpenFont)
> > >   Serial number of failed request:  41
> > >   Current serial number in output stream:  42
> > > -- snip --
> > > (I have attached the matching font "timehr__.pfa" to this message).
> > [...]
> >
> >   Perhaps it isn't related with ttf2pt1.
> >
> >   I can render it without error messages.
> >
> >   XFree86-4.2.0 + FreeType-2.0.9 on Linux box.
> >
> >   But this font has some problem, when open with pfaedit i got,
> >
> > Copyright ? 2000-2002 by George Williams.
> >  Executable based on sources from 19:55 5-4-2002.
> > Didn't understand |/.notdef put dup 10 /.notdef put dup 11 /c11 put dup 12 /c12 put dup 13
> >
> >   Maybe you can try t1utils,
> >
> > t1diasm timehr__.pfa timehr__.ps
> >
> >   And edit timehr__.ps, change UnderlinePosition/UnderlineThickness
> >   to inetger then,
> >
> > t1asm -b timehr__.ps timehr__.pfb
> >
> >   And try to render timehr__.pfb again.
> 
> Looks that the font was somehow broken... running "type1fix.pl" (from
> the TeXTrace package) over the fonts fixed the issue... :)

If you can find out what exactly it has fixed, that would be great.

-SB

Roland Mainz wrote:
> 
> Does someone know how difficult it is to turn "ttf2pt1" into a shared
> library  ?
> The idea is to use "ttf2pt1" from other applications using a
> well-defined shared library API.
> 
> Any suggestions/ideas ?

It will require a major audit and revision of the code. Basically
the problems are that large amounts of the code is written in the
assumption that it's a stand-alone program, with the problems like:

- the global context is truly global, some initializations are
done at compile time
- when something fails, it just does exit(1)
- some parts may be not very careful with freeing the dynamically
allocated memory
- the ttf parser is not doing a very careful error checking so
it may crash on bad fonts (though there has been an improvement
throughout the life of project and the current state is not as bad
as it used to be)

So it's doable but not too easy. Also I don't see why someone would 
want to turn the whole ttf2pt1 into a library: it's easier to just
call it as a separate binary. Turning the parts of it into libraries
makes more sense, in particular the outline processing and font
generation code and I've considered that. One interesting use would
be to reuse the code for an interactive font editor, kind of
like pfaedit. But again, it requires a significant amount of work
and currently I don't have any personal interest in doing that. 
(And I have more interesting things to spend time on than I have
spare time).

If you can turn it into a library, your changes would be definitely
welcome. I can help you with description of the general logic of
things.

-SB

Werner LEMBERG wrote:
> 
> > The daily CVS archive is availabel from the Download page.
> > There is also a link to the on-line repository browser among
> > the list of pages (on each page):
> >
> > Background | What Is TTF2PT1 | History | Download | CVS | Credits | Todo
> >                                                     ^^^
> 
> Hehe.  This link is broken (at least if you start with
> http://ttf2pt1.sf.net), so I've assumed that no public CVS is
> available.

Oops, I guess it got broken during on of the SF download server
reforms. Thanks for the info.

-SB

> The daily CVS archive is availabel from the Download page.
> There is also a link to the on-line repository browser among
> the list of pages (on each page):
> 
> Background | What Is TTF2PT1 | History | Download | CVS | Credits | Todo
>                                                     ^^^

Hehe.  This link is broken (at least if you start with
http://ttf2pt1.sf.net), so I've assumed that no public CVS is
available.


    Werner

Werner LEMBERG wrote:
> 
> >   Sorry, I cannot let ttf2pt1 work normally if I link to
> >   freetype2-current(the same with CVS version).
> 
> There have been heavy changes behind the scene.  It is possible that
> ttf2pt1 uses some internal structures which have changed, I don't
> know.

Hm, it should not. The only thing that looks semi-internal to me
is looping over the list of splines but even that was documented.
 
> BTW, I would like to access ttf2pt1 using CVS.  Is this possible?
> There is no hint on the project's homepage.
> 
> Then I can try to find out what's failing.

The daily CVS archive is availabel from the Download page.
There is also a link to the on-line repository browser among
the list of pages (on each page):

Background | What Is TTF2PT1 | History | Download | CVS | Credits | Todo
                                                    ^^^

Both are also available from http://sf.net/projects/ttf2pt1                                          

The CVS access to repository is available as usual on
SourceForge (including anonymous access), they have instructions
somewhere.

-SB

Hi !

----

Does someone know how difficult it is to turn "ttf2pt1" into a shared
library  ?
The idea is to use "ttf2pt1" from other applications using a
well-defined shared library API.

Any suggestions/ideas ?

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) Rol...@in...
  \__\/\/__/  gi...@in...
  /O /==\ O\  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
 (;O/ \/ \O;) TEL +49 641 99-41370 FAX +49 641 99-41359

"James H. Cloos Jr." wrote:

> If the curves are sufficiently free, it would be cool to see them
> included in the freefont project <http://www.freesoftware.fsf.org/freefont/>.

I am not sure about the license (yet) ... ;-(

Just wondering: Will the "freefont"-project provide PS Type1 fonts for
the single encodings some day ?

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) Rol...@in...
  \__\/\/__/  gi...@in...
  /O /==\ O\  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
 (;O/ \/ \O;) TEL +49 641 99-41370 FAX +49 641 99-41359

Edward Lee wrote:

> > An attempt to use these fonts always ends like this:
> > -- snip --
> > % xfd -fn '-hebrew-times-medium-r-normal--0-0-0-0-p-0-iso8859-8'
> > X Error of failed request:  BadAlloc (insufficient resources for
> > operation)
> >   Major opcode of failed request:  45 (X_OpenFont)
> >   Serial number of failed request:  41
> >   Current serial number in output stream:  42
> > -- snip --
> > (I have attached the matching font "timehr__.pfa" to this message).
> [...]
> 
>   Perhaps it isn't related with ttf2pt1.
> 
>   I can render it without error messages.
> 
>   XFree86-4.2.0 + FreeType-2.0.9 on Linux box.
> 
>   But this font has some problem, when open with pfaedit i got,
> 
> Copyright ? 2000-2002 by George Williams.
>  Executable based on sources from 19:55 5-4-2002.
> Didn't understand |/.notdef put dup 10 /.notdef put dup 11 /c11 put dup 12 /c12 put dup 13
> 
>   Maybe you can try t1utils,
> 
> t1diasm timehr__.pfa timehr__.ps
> 
>   And edit timehr__.ps, change UnderlinePosition/UnderlineThickness
>   to inetger then,
> 
> t1asm -b timehr__.ps timehr__.pfb
> 
>   And try to render timehr__.pfb again.

Looks that the font was somehow broken... running "type1fix.pl" (from
the TeXTrace package) over the fonts fixed the issue... :)

----

Bye,
Roland

-- 
  __ .  . __
 (o.\ \/ /.o) Rol...@in...
  \__\/\/__/  gi...@in...
  /O /==\ O\  MPEG specialist, C&&JAVA&&Sun&&Unix programmer
 (;O/ \/ \O;) TEL +49 641 99-41370 FAX +49 641 99-41359

On Sat, Apr 20, 2002, Edward Lee wrote:
>   Sorry, I cannot let ttf2pt1 work normally if I link to
>   freetype2-current(the same with CVS version).
> 
>   The weird thing is, everything is Ok, no error messages,
>   the *.t1a/*.pfb only have ascii characters no Chinese fonts
>   inside.

  Seems that cannot get the encoding of the font.
  fail in glenc() of ft.c of ttf2pt1 when use
  freetype2-current version.


Rgds,
Edward G.J. Lee

On Sat, Apr 20, 2002, Werner LEMBERG wrote:
> 
> >   Sorry, I cannot let ttf2pt1 work normally if I link to
> >   freetype2-current(the same with CVS version).
> 
> There have been heavy changes behind the scene.  It is possible that
> ttf2pt1 uses some internal structures which have changed, I don't
> know.
> 
> BTW, I would like to access ttf2pt1 using CVS.  Is this possible?
> There is no hint on the project's homepage.

  CVSROOT is,

  :pserver:ano...@cv...:/cvsroot/ttf2pt1

  no password, ``Enter'' instead. Modulename is ttf2pt1.

> Then I can try to find out what's failing.

  I'm reading the code too, but I don't know the exact diffrence
  between 2.1.0 and 2.0.9 of FT2.

  Thanks in advance.


Rgds,
Edward G.J. Lee

>   Sorry, I cannot let ttf2pt1 work normally if I link to
>   freetype2-current(the same with CVS version).

There have been heavy changes behind the scene.  It is possible that
ttf2pt1 uses some internal structures which have changed, I don't
know.

BTW, I would like to access ttf2pt1 using CVS.  Is this possible?
There is no hint on the project's homepage.

Then I can try to find out what's failing.


   Werner

> Please retry with the current development version,
> 
>   ftp://ftp.freetype.org/freetype/unstable/freetype2-current.tar.gz
> 
> I think this bug has been fixed meanwhile.

  Sorry, I cannot let ttf2pt1 work normally if I link to
  freetype2-current(the same with CVS version).

  The weird thing is, everything is Ok, no error messages,
  the *.t1a/*.pfb only have ascii characters no Chinese fonts
  inside.

%%CreationDate: Sat Apr 20 11:22:35 2002
% Converted by ttf2pt1 3.4.1-CURRENT/ft
% Args: -GE -pft -Ohub -W4 -L /usr/local/share/ttf2pt1/maps/cubig5.map+06 bsmi00lp.ttf bsmi00lp06
%%EndComments

  I check the ``CHANGES BETWEEN 2.1.0 and 2.0.9'' but in vain.

  Did I miss something? Thanks.


Rgds,
Edward G.J. Lee

Roland Mainz wrote:
> 
> Hi !
> 
> ----
> 
> I'd like to filter some PFA fonts using ttf2pt1
> (ttf2pt1-3.4.1-SNAP-020329 + using the freetype-2.1.0 library as parser)
> - however the filtered versions do not behave as the original versions.
> 
> Example:
> I am trying to filter the Hebrew fonts from
> http://puck.informatik.med.uni-giessen.de/people/gisburn/elmar8859_8_type1.tar.gz
> :
> -- snip --
> mkdir filtered
> cd filtered
> for i in ../elmar/*.pfa ; do
> echo " ---------- $i"
>  ttf2pt1 -a -b $i - >$(basename ${i%.pfa}.pfb)
> done
> -- snip --
> 
> The original version works with the Adobe-based Type1 font engine in
> Solaris Xsun but causes any X.org-based Xserver to crash. The filtered
> version works with the X.org server but Solaris Xsun does not show the
> hebrew-specific chars in "xfd" or any other application... ;-(
> Some information seems to be lost during the conversion - but which one
> ?

Do these Hebrew characters look complex ? One possibility is that
the substituted hints for them are too big for this particular
interpreter. Try the option -Ou to disable the hint substitution.
If it helps, you may want to try experimenting with lower limits
on hints complexity, say "-mh=64" or lower values.

Of course, you can always run t1disasm, look at the source code
of the fonts and compare them.
 
> Is there any way to prevent this kind of problem (maybe via a "secret"
> ttf2pt1 option) ?

Sorry, all the options are documented :-)

-SB

> I have no idea. Apparently the difference between the afm files is:
> 
> 45c45
> < C 27 ; WX 1000 ; N ZdZ2708 ; B 41 322 956 25886 ;
> ---
> > C 27 ; WX 1000 ; N ZdZ2708 ; B 41 -47 956 818 ;
> 274c274
> < C -1 ; WX 1000 ; N .notdef ; B 0 0 0 0 ;
> ---
> > C -1 ; WX 1000 ; N .notdef ; B 225 68 315 685 ;
> 
> and most probably it does not like the value of 25886. Maybe
> the Freetype people can say how it gets there.

Please retry with the current development version,

  ftp://ftp.freetype.org/freetype/unstable/freetype2-current.tar.gz

I think this bug has been fixed meanwhile.


    Werner