On python2.7, the fonttools py23 module registers a 'lastResort' StreamHandler
similar to the one found in python3's logging module, that always writes
to the current value `sys.stderr`.
This also applies to any python library that imports from fontTools.misc.py23
under python2.7.
The logging module has a 'shutdown' atexit handler that flushes all the
logging handlers' streams just before the python interpreter exits.
Sometimes (e.g. when calling `python setup.py test` as in MutatorMath's test
suite), the interpreter termination ends with a traceback, which is
triggered by the atexit handler failing to flush the lastResort handler's
stream, sys.stderr
AttributeError: None has no attribute 'stderr'
This is because during module teardown, the globals (in this case 'sys')
are set to None, and the order in which modules are deleted is not
guaranteed.
See 58531934a8
We are using unicode_literals in this module since 329261b.
Because of that, on py27 sometimes a TTFont glyphOrder may end up with
a mix of `str` and `unicode` glyph names.
While I'd love to change every single 'text' string to unicode, maybe
in this case it makes sense to use the native `str` type (`bytes` on
py2, unicode string on py3) for the UV2AGL and AGL2UV dictionaries, as
glyph names can only contain ascii characters anyway.
See https://github.com/fonttools/fonttools/pull/774#discussion_r98327429
When a subroutine contains no explicit hint stem operators (has_hint=False),
but it contains a hintmask operator which, in the context of the calling
charstring, would be understood as implying a vstemhm, then we need to set
has_hint=True on the subroutine, and update the last_hint index.
Otherwise, the drop_hints function leaves behind the arguments of the implicit
vstemhm operator.
This case is exemplified in the test font Tests/subset/data/Lobster.subset.ttx,
for charstrings "B" and "B.salt", and subroutine index="2".
--- /Users/cosimolupo/Documents/Github/fonttools/Tests/subset/data/expect_no_hinting_CFF.ttx
+++ /var/folders/jb/rjz76yw92w7144mwqg119jnm0000gn/T/tmpO_XOWh/tmp3.ttx
@@ -47,7 +47,7 @@
107 return
</CharString>
<CharString index="2">
- 230 636 rmoveto
+ 119 230 636 rmoveto
-136 -636 rlineto
144 hlineto
return
@@ -94,7 +94,7 @@
endchar
</CharString>
<CharString name="B">
- 187 -105 callsubr
+ 187 6 93 362 139 -119 101 -101 -105 callsubr
82 383 rlineto
2 18 20 1 8 hhcurveto
73 22 -57 -70 hvcurveto
@@ -109,7 +109,7 @@
endchar
</CharString>
<CharString name="B.salt">
- 185 -105 callsubr
+ 185 6 93 350 149 -119 105 -105 -105 callsubr
6 30 rlineto
-41 39 41 -17 39 hhcurveto
125 110 175 136 72 -32 62 -82 15 hvcurveto
_DehintingT2Decompiler now inherits from T2WidthExtractor, so we can save
the charstring width and insert it back after we dropp the hints.
This avoids the need to do an extra .draw() just for the sake of
extracting the width.
We shall reuse it as base class for the _DehintingT2Decompiler in subset module,
as we need to save the width in order to re-insert it after dropping the hints.
This is the same patch as PR #606
Quoting myself:
```
It seems that the horizontal advance of hinted charstrings gets lost when the CFF hinting is stripped with '--no-hinting' option...
T2CharString objects only get assigned a 'witdh' attribute after their 'draw' method is called.
The subsetter's drop_hints function attempts to insert the width back at the beginning of the de-hinted charstring's program, but can do that only if the charstring does have a 'width' attribute:
c63fea0f8f/Lib/fontTools/subset/__init__.py (L1928)
Hence, we must 'draw' the charstring (with a NullPen) before stripping the hints.
```
Now this method fixes the issue (advances are kept) when doing _both_ --no-hinting and --desubroutinize (the test_no_hinting_desubroutinize should now pass, while it was failing before).
However, when one only does does --no-hinting, this method raises an error:
AttributeError: 'NoneType' object has no attribute 'nominalWidthX'
Lib/fontTools/misc/psCharStrings.py:282: AttributeError
(this is reflected in the failing `test_no_hinting_CFF`)
Indentation for TTX XMLF is 2 spaces. I chose 4 spaces for indentation of instructions to make it more noticeable. I hope it’s not a problem, because assembly code lines are usually very short.
Also fixes issues with merging of PairPos.
Trying on Noto Sans Thai still fails, now because of issues in class differences in
PairPosFormat2. :( Investigating.
This is part of fixing https://github.com/fonttools/fonttools/issues/719
though, the changes are currently in interpolate_layout, and need to be ported /
merged with varLib.__init__ variation-font-builder.
We should probably use cliTools.makeOutputFileName here for consistency with ttx,
but I know some prefer the approach taken by pyftsubset so I don't want force this.
Howver, I find it a bit annoying that when one is too lazy (like me) to specify
the --output-file=, the subsetter outputs a file ending with an invalid ".subset" extension, which Finder or Windows Explorer don't recongize as a font, and so one has to rename it anyway.
This makes the '.subset' string is inserted between the file name and its original extension (if any).
The code in main() assumes that the first of the args which are not consumed by Options.parse_opts(args) is the positional argument (ie. does not start with '--') for the input font file, and that is what the usage() says too.
However I find it myself writing --options=... first, and then at the end the positional arguments, as is the convention for many other Unix tools.
This patch makes this possible, while also keeping the current behavior.
Rounding of coordinates is now disabled by default.
The pen now accepts an optional 'roundTolerance' float, with values between
0 and 1 (default is 0).
Values >= 0.5 mean round all coordinates to integers.
Values between 0 and 0.5 mean round only when the absolute difference
between the original float and the rounded integer is within the tolerance
Fixes#769
* [feaLib.ast] fix checking multiple markClass definitions don't redefine same glyphs
As pointed out by @mhosken, we are looking in the wrong list for glyphs
that have already been defined in a previous markClass definition.
With this patch, the current markClass.fea test case fails becuase it defines 'acute'
twice for the same @TOP_MARKS class:
fontTools.feaLib.error.FeatureLibError: Lib/fontTools/feaLib/testdata/markClass.fea:6:5: Glyph acute already defined at Lib/fontTools/feaLib/testdata/markClass.fea:3:1
Also see conversation at:
3b79d51755 (r94622074)
* feaLib/testdat/markClass.fea: remove duplicate 'acute' in @TOP_MARKS
Now the test pass, after 45c77b7
* [feaLib.builder_test] test case for redefined glyph in multiple markClass definitions
This should be a bit more efficient as we don't need to create a temporary list
object before passing it to "".join() method of str; the latter accepts any
Iterable
