Distinction between hintmask bits and an operator in charstring

While subsetting I got an error raised in op_or(), although this font has no or operators, of course.

$ pyftsubset font.otf --text=島 --no-hinting
Traceback (most recent call last):
  File "/Users/mnakamura/.pyenv/versions/subset-2.7.13/bin/pyftsubset", line 11, in <module>
    load_entry_point('fonttools', 'console_scripts', 'pyftsubset')()
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/misc/loggingTools.py", line 372, in wrapper
    return func(*args, **kwds)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/subset/__init__.py", line 2971, in main
    subsetter.subset(font)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/subset/__init__.py", line 2785, in subset
    self._prune_post_subset(font)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/subset/__init__.py", line 2768, in _prune_post_subset
    retain = table.prune_post_subset(self.options)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/subset/__init__.py", line 2208, in prune_post_subset
    decompiler.execute(c)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/subset/__init__.py", line 2001, in execute
    psCharStrings.T2WidthExtractor.execute(self, charString)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/misc/psCharStrings.py", line 303, in execute
    rv = handler(index)
  File "/Users/mnakamura/.ghq/github.com/fonttools/fonttools/Lib/fontTools/misc/psCharStrings.py", line 375, in op_or
    raise NotImplementedError
NotImplementedError

This glyph has fifteen stems, and contains a sequence hintmask 0110111101110010 in its charstring. Since the mask bits 0110111101110010 (= 6f72 in hex) can be decoded as a string “or” in ASCII-compatible encodings, the sequence is stored as [..., 'hintmask', 'or', ...] in charString.program, and eventually it’s impossible to distinguish the bits from an or operator.

Workaround

Use Python 3.

[anthrotype]

I think I’ve fixed the issue in https://github.com/fonttools/fonttools/commit/c0d672ba70ba6907759b086d59f110496d79c606 Could you please check it works now? Also it would be nice if you could provide a minimal font the reproduces the issue so we can add to our regression tests (one glyph would be enough).

[anthrotype]

@mashabow Thank you! 😄