[tlbuild] ttf2pk versus ttf2pk2

[Hin-Tak Leung]

------------------------------
On Mon, Jul 22, 2013 10:59 BST Peter Breitenlohner wrote:

>On Tue, 4 Jun 2013, Peter Breitenlohner wrote:
>
>> Hi Norbert,
>>
>> ttf2pk2 is my attempt to rewrite ttf2pk, but using libfreetype instead of
>> libttf -- all that based on some earlier work from
>> http://chem.skku.ac.kr/~wkpark/project/ktug/ttf2pk-freetype2_20030314.tgz
>>
>> At the moment ttf2pk2 does the same as ttf2pk except for the substitution
>> from the 'vert' feature for rotated CJK fonts.  Once that has been done and
>> tested I'll change ttf2pk2 => ttf2pk and libttf can be discarded.
>
>Hi Karl, Norbert,
>
>as you probably have already seen, I have added the 'vert' substitutions for
>rotated CJK fonts to ttf2pk2 and it is now the default in the TL SVN
>repository (since r30893 from 2013-06-24).  The old version is still
>available if explicitly requested via '--disable-ttf2pk --enable-ttf2pk2'.
>
>Hi all,
>
>I'd apreciate any tests you can perform and feedback (about both any
>problems and success) in order to eliminate bugs, in particular those
>introduced by the transition FreeType 1 => 2.

Here is what I think:

- a few of the existing problems with ttf2pk must be addressed - in particular, the indeterminism issue.
(I am refering to runs separated by a few hours or a few days giving
non-identical results, not compared with 10-years ago) - it is pointless to talk about "does the same",
when even that "same" changes every few hours.

- then you write down the known progressions, uninteresting differences, and regressions; and fixes
all the regressions

- then you test against the common fonts on common platforms, i.e. all the platform-bundled fonts
on windows and/or Mac OS X. e.g. there are quite a few on-line tutorials on using mssimsun and
msmincho/msgothic (I am not going to debate the legality of it. "testing" does not need to mean
re-distributing the outcome - the tfm/enc files - nor encouraging such use).

- then you testing against rarer fonts from rarer vendors. e.g. there are again, a few on-line tutorials
on using Netscape's cyberbit - I believe somebody have even tar'gz'ed the resulting tfm/enc files
for convenience of use for this. And rarer fonts.

When you have done all of that, *then* you can talk about withdrawing the old version.
You can treat 2. and 3.as the same - i.e. testing against common fonts on common platforms
and detailing all the known progressions, uninteresting differences, and *no* regressions.

It is just not appropriate to talk about withdrawing the old, at this stage.