Re: [cjk] revisiting ttf2tfm and dvipdfmx

[Hin-Tak Leung]

Werner LEMBERG wrote:
<snipped>
> Actually, I no longer maintain ttf2tfm and ttf2pk (mainly due to lack
> of time), sorry.  What you can find in the TeXLive SVN repository is
> the most recent version, and I suggest that you send patches to Peter
> directly (address@hidden) or to the TeXLive team.

Was rather surprised when I ran a gdb session with fedora's debug info
and saw the source path mentioning freetype 1.5 . I just assumed most things
are freetype 2 based by now. Anyway, it seems to be a regression, or I was
very lucky not to see the segfault 10 years ago due to some padding from quirky 
compiler versions, etc.

> Regarding the cjk git repository, just clone it and commit your
> changes, then produce a series of patches so that I can apply them.  I
> can also give you write access on Savannah if you prefer.

I only have one issue with freetype 2 at the moment - ftdump (or ftview) on 
NISC18030.ttf, one of the fonts shipped with Mac OS X 10.7 (and possibly other
versions also) don't work. I haven't looked much further, but ghostscript
seems to be happy loading the font file, at least the headers, and I'd be 
surprised that Apple, of all people, ships a broken font, so ftdump refuses to 
do essentially gs's ttfloadtable is strange.

If you have access to a recent Mac OS X box, you can have a look yourself.

> > - "\begin{CJK*}{SJIS}{wnmc}\end{CJK*}" used to look for
> >    c49wnmc.fd. Now it looks for c40wnmc.fd.  Is that intentional?
>
> For me it apparently works (using SVN revision of TeXLive 30186 from
> 2013-04-30):

Fedora ships 20130608 r30832 , so mine is just a little more recent than yours,
but that shouldn't be an issue.

>    (.../texmf-dist/tex/latex/cjk/texinput/SJIS/SJIS.bdg
>    File: SJIS.bdg 2012/05/07 4.8.3
>    )
>    (.../texmf-dist/tex/latex/cjk/texinput/SJIS/SJIS.enc
>    File: SJIS.enc 2012/05/07 4.8.3
>    )
>    (.../texmf-dist/tex/latex/cjk/texinput/SJIS/SJIS.chr
>    File: SJIS.chr 2012/05/07 4.8.3
>    )
>    LaTeX Font Info:    Try loading font information for C49+wnmc ...
>    LaTeX Font Info:    No file C49wnmc.fd. ...

I'll give that a try again.

> > - I am surprised to find that many tfm's change when I run ttf2tfm
> > repeatedly.  Is it supposed to do that? (I now store these generated
> > files in a local git repository, so it is rather noticeable).
>
> No; the contents should be identical always if the incantation is
> exactly the same.

Hmm, the difference is quite extensive - both the number of subfonts and
the number of actual fonts. I was thinking of reading them back.

Though, in the documentation, it says something about not-used glyph slots
could contain anything (presumably 'random anything').

" Unused positions (either caused by empty code points in the mapping
table or missing glyphs in the TrueType font) will be filled (rather 
arbitrarily) with characters present in the input  encoding
but  not  specified  in the output encoding (without the -q option ttf2tfm 
prints the final output encoding to standard output)."

> > - how to do the equivalent of ttf2tfm for opentype fonts? There is a
> > otftotfm, but it doesn't seem to have options to do cjk-compatible
> > outputs.
>
> No idea, sorry.  I think that the future is xetex and/or luatex which
> both can handle OTF directly, so extended support for ttf2tfm is
> probably wasted time.  If you *really* need the old route, I today
> recommend to use fontforge for generating good old subfont PFBs.

I played with xetex a bit a while ago, but found it doesn't do line breaks, 
glyph spacings as nice as CJK. Maybe I am old-fashioned :-).

> > - updmap doesn't work like it used to, and seemed to be totally
> > re-written - got new map files to work with pdftex by appending
> > directly to pdftex.map. Is there some secret recipe for updating
> > pdftex.map with new entries?
>
> Have you tried `texdoc updmap' to read the man page?  I forgot the
> details...

I'll keep looking. The doc suggests "updmap --enable Map=MAPFILE" to do each one 
by one.

> > - I think there is a latex command (or should be) to make messages
> > like "Missing character: There is no y in font xxx" more visible?
> > They appears in the log file but not on the console.
>
> This is an e-tex extension, and you have to explicitly activate it
> with
>
>    \tracinglostchars 2
>
> (see `texdoc etex' for details).

Thanks for the tips.

> > - I have a font which when processed through LaTeX for "ABCD" (for
> > illustration, not really english), resulted in "D C B A" where A is
> > approximately at the right place.  What kind of problem may that be?
>
> Good question.  Maybe negative advance widths?  I'm not sure whether
> TeX allows this at all.  Maybe you can run `tftopl' to check the TFM
> file, or maybe you call `t1disasm' or `ttx' to disassemble the font.

Thanks for the tips. It is a typeface to mimic stone-cuttings in seals (i.e.
official looking thing, substitute for signatures), so probably have
always been used in a glyph-by-glyph basis, and may have been always strange
that way.

> > - two fonts works okay during the latex-> dvi stage, but refused to
> > be processed in the dvi-> stage (both pdfTeX and dvipdfmx) for
> > having unknown post-version .  Which part of TexLive should I look
> > into?
>
> You mean the `post' table in the TTF file, right?  This looks rather
> strange, since the available formats of this table haven't changed or
> extended in the last ten years, AFAIK.  Maybe a font bug?  Otherwise,
> you have to directly check the source code of dvipdfmx since it parses
> the TrueType fonts by itself, IIRC.

Yes - see the other thread. I just mod'ed the font files to read v3 (no post
glyph name table), then they work. They are of the win 3.x era, I believe, newly
acquired, not part of the font sets I played with 10 years ago.

> > - About font embedding flag within a truetype font. I see dvipdfmx
> > honour that, pdflatex completely ignores it, which xdvipdfmx (xetex)
> > can be overridden on a command line. what's your view on it?
>
> It's a gimmick, essentially.  Everyone can easily circumvent the
> restriction.  However, it is a good idea to set them correctly IMHO
> just to feel better :-)

:-). It probably make sense 10 years ago to restrict to printing only (just use 
private for hard-copy), but these days everything is digital, not allowing 
embedding significantly reduces the usefulness of a type face.

OTOH, I'd prefer not to modify font files - the number of Arphic GPL cjk font
derivatives out there make me crinch... just because one is allowed to modify
a font, and can do it, does not mean one should. :-(.

> > - I can't get dvipdfmx to do landscape mode (most of the latex
> > packages (lscape and pdflscape)/dvi specials don't work) - this
> > seems to be an old silent problem which now has a visible message.
> > I'd also like to do the equivalent of dvips -E - i.e. tight bounding
> > box also, for generating pdf's to be embedded by another pdf's.
>
> Have you tried dvipdfmx's command line option `-l' to set landscape
> mode?

Yes, already tried that. It does something wierd and not quite expected.

I'll probably drop you a sizeable patch to update cjk/doc/pdfhowto/* at some
stage. (if I find the time gradually...).

Hin-Tak