This version extends the kanji coverage to roughly halfway through Grade
Four. There have also been a lot of infrastructure changes, some of them
with important visible effects. This version has taken a long time to bring
to release, and the rate of kanji per unit time is a fair bit below what I'd
like it to be. That's partly because of issues outside the project that
have limited my available time to work on defining kanji. There's a lot of
important material here other than kanji, however, and my hope is that the
infrastructure improvements in 0.7 will help improve the rate of kanji
addition further in the future and keep the project sustainable long enough
that it can reach full coverage. Here's a summary of the new stuff.
Font weight (bold, light, and so on) has been added, along with what I think
is a reasonable if idiosyncratic way of handling "style" definitions like
"Anbiruteki" that are actually weight-related. This feature basically
multiplies by six the number of fonts the package can generate. It was
gratifying that almost all the code I wrote a long time ago to support font
weight in the future and never really tested, actually worked when I added
the missing parts and flipped the switch.
"Anchors," which are primarily used for attaching combining accents in
Latin, have been reworked almost completely. They were broken in the
previous version and should at least be less broken now.
A general mechanism for "parasite" packages was added to the build system,
and several of my other projects were reorganized to use this mechanism.
They now come included with full distributions of Tsukurimashou. Genjimon
and IDSgrep were already affiliated with Tsukurimashou, but I've made their
interfaces to Tsukurimashou's build system more consistent and have added
OCR and Beikaitoru.
Dependency scanning for PFB files has been split into smaller chunks that
can run simultaneously; this significantly speeds up some common scenarios
when running the build system on a multicore computer.
Build-system dependency handling should have been improved in the case of a
brand-new Metapost source file being added. This is difficult to test,
however, and may require further work in the future.
IDSgrep has many new features. I released a version 0.3 of it in late 2012,
but Tsukurimashou 0.7 now ships with a pre-release of version IDSgrep 0.4,
containing some interesting features beyond those of IDSgrep 0.3.
Some lowercase Latin letters in TsuIta have been redrawn. This was a to-do
list item; the last released version of TsuIta was experimental and this
version is less so.
There is more support for "Hangul Extension" Unicode blocks, covering rare,
historical, and dialect vowels and consonants in Korean.
The "xpua" feature was added, to remove private-use code points from glyphs
meant to be accessed via OpenType features. This may improve the ability to
cut and paste from PDF documents made with Tsukurimashou fonts.
The bundled version of METATYPE1 has diverged further from GUST's
distributed version; many features not used by Tsukurimashou have been
removed, and the hope is that this stripped-down version will be easier to
Low-level shape-drawing operations in Tsukurimashou have been modified to
insert extra nodes in some curves in order to produce more accurate shapes
and line widths, especially when a stroke turns a tight corner. This has
necessitated some changes to the definitions some shapes (especially in the
hiragana, which contain some of the oldest code still in use) because of
unexpected dependence on the old, broken behaviour. It is possible that
some glyphs may have visibly changed shape, but the goal is for that not to
have happened to any significant degree.
FontForge's fontlint script has been replaced. This was a last-minute
addition and detailed documentation isn't written yet, but there's a strong
chance that Tsukurimashou's replacement for fontlint will end up being
adopted by FontForge mainline anyway. Shipping it with Tsukurimashou is
still worthwhile for the benefit of people who aren't using cutting-edge
Another thing that came up at the last minute: CHISE's servers appear to
be shutting down, and I have posted a snapshot of the part of their Git
repository that is relevant to IDSgrep, in my own Github account.