As https://github.com/harfbuzz/harfbuzz/pull/1059#issuecomment-397912812 we
like to still work with blobs that harfbuzz itself can't handle directly that are failing sanitization
currently apparently.

We were passing the font path directly to freetype so rendering
was broken when we are getting the font from stdin.

This fixes it by using FT_New_Memory_Face instead.

This fixes:
* build/util/hb-view /dev/stdin text < font.ttf
* build/util/hb-view - text < font.ttf
* cat font.ttf | build/util/hb-view - text

but doesn't work on
* cat font.ttf | build/util/hb-view /dev/stdin text

which I will try to fix separately.

New API:
HB_BUFFER_FLAG_REMOVE_DEFAULT_IGNORABLES
hb-shape / hb-view --remove-default-ignorables

One more text-rendering-tests test passing.  Eleven failing.

When advances are not printed, glyph offsets reflect absolute glyph
positions.

New API:
HB_BUFFER_SERIALIZE_FLAG_NO_ADVANCES
hb-shape --no-advances

Part of fixing https://github.com/behdad/harfbuzz/issues/585

Simplifies compatibility issues by centralizing all boilerplate code
in hb-private.hh.

If buffers are empty, content type should be ignored.

This fixes last of the failing tests: fuzzed.tests.  Green again!

If *I* get this wrong in 2017, I have no idea how others get their head around
cluster math...

Fixes tests/arabic-fallback-shaping.tests

I like to have a mode where CONTAINS_NOTDEF and CONTAINS_DOTTEDCIRCLE are not
returned.  Abused a value of -1 for that.  hb-shape now uses it.  Fixes two
of the six tests failing with --verify in test/shaping/run-tests.sh.

We break and shape fragments and reconstruct shape result from them.
Remains to compare to original buffer.  Going to add some buffer
comparison API and use here, instead of open-coding.

Right now it checks for monotone cluster values.  Other checks to be added.

Not all shapers code is updated to set this properly.
GSUB and Arabic shaper are updated.
GPOS and other shapers are NOT.

Fixes https://github.com/behdad/harfbuzz/issues/224

https://github.com/behdad/harfbuzz/issues/506

Before, that was printed using --debug (and in both hb-shape and hb-view).
Changed it, now hb-shape has a new command-line argument called --show-messages.
When invoked, it also respects other output formatting options.  The messages
are better formatted and printed to te same place that hb-shape output is
directed to.  Previously they were written to stderr.

Fixes https://github.com/behdad/harfbuzz/issues/506

Is hooked up to the font, but not to FreeType, so raster doesn't show yet.

Documentation needs to be done.

Pre-2013 MSVC does not have scalbn() and scalbnf(), which are used in the
utility programs.  Add  fallback implementations for these, which can be
used when necessary.

New API: HB_BUFFER_SERIALIZE_FLAG_GLYPH_EXTENTS

hb-shape now accepts --show-extents.

Patch from Simon Cozens.

TODO: Documentation.

Fixes: https://bugs.freedesktop.org/show_bug.cgi?id=71445

Also makes hb-view to use 8 bits of subpixel precision and shape at
requested size, instead of always shaping at upem and scaling results.

Supports ft and ot right now.  hb-view currently not rendering with ot.
Will fix after some clean up.

Avoids "deprecated" warnings.

Add deprecated alias for old name.

Useful for profiling shapers.

Before, if one called hb_shape() without setting script, language, and
direction on the buffer, hb_shape() was calling
hb_buffer_guess_segment_properties() on the user's behalf to guess
these.

This is very dangerous, since any serious user of HarfBuzz must set
these properly (specially important is direction).  So now, we don't
guess properties by default.  People not setting direction will get
an abort() now.  If the old behavior is desired (fragile, good for
simple testing only), users can call
hb_buffer_guess_segment_properties() on the buffer just before calling
hb_shape().

As reported by Jonathan, OS X has setmode() that is something other
than what setmode() is on Win32.  So, limit invocation to Windows
platforms only.