TypeError from _intSecAtan on previously valid arguments (valid before 4.40.0)

[stenson]

Hi fontTools — long time fan, first time issue reporter — apologies if this has already been reported, but I've been using fontPen’s FlattenPen in coordination with ttf variable fonts for a number of years now with no issue, but when I recently updated a fontTools install (as a requirement of a fontmake install), I started getting this error whenever I tried to use the FlattenPen on glyphs with quadratic curves.

After some investigating, I narrowed the issue down to FlattenPen calling calcQuadraticArcLength and then, in turn, that function calling _intSecAtan — fontTools 4.39.0 works fine for all of my use cases, but 4.40.0 introduces the issue.

Here's an example traceback, with a minimal example (a direct call to calcQuadraticArcLength):

Traceback (most recent call last):
  File "/Users/robstenson/Type/flattentest/test_minimal.py", line 7, in <module>
    print(calcQuadraticArcLength((210, 333), (289, 333), (326.5, 290.5)))
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "Lib/fontTools/misc/bezierTools.py", line 183, in fontTools.misc.bezierTools.calcQuadraticArcLength
  File "Lib/fontTools/misc/bezierTools.py", line 233, in fontTools.misc.bezierTools.calcQuadraticArcLengthC
  File "Lib/fontTools/misc/bezierTools.py", line 148, in fontTools.misc.bezierTools._intSecAtan
TypeError: must be real number, not complex

Not sure if it's of any value, but I put together a small repo to demonstrate the issue here: https://github.com/stenson/flattentest. My main issue is demonstrated in test.py, with this code snippet:

from fontTools.ttLib import TTFont
from fontTools.pens.recordingPen import RecordingPen
from fontPens.flattenPen import FlattenPen

glyph = TTFont("MutatorSans.ttf").getGlyphSet()["B"]

rp = RecordingPen()
fp = FlattenPen(rp, approximateSegmentLength=5)
glyph.draw(fp)

As far as I know, I'm not doing anything out of the ordinary here, but I get the TypeError with all recent versions of fontTools.

[anthrotype]

Seems like a regression, thanks for the report, I will take a look at it today

[justvanrossum]

I wonder if it relates to #3121. The failing code is optionally optimized by cython, but has otherwise not been changed in a while.

[anthrotype]

argh.. yes seems to be related to cython, again..(6012643)

[anthrotype]

actually in this case cython might be right.. the value is wrongly typed, it should have been typed as double, not as complex:

diff --git a/Lib/fontTools/misc/bezierTools.py b/Lib/fontTools/misc/bezierTools.py
index 7772a4bf8..21ab0a5d0 100644
--- a/Lib/fontTools/misc/bezierTools.py
+++ b/Lib/fontTools/misc/bezierTools.py
@@ -141,7 +141,7 @@ def _dot(v1, v2):
 @cython.cfunc
 @cython.inline
 @cython.returns(cython.double)
-@cython.locals(x=cython.complex)
+@cython.locals(x=cython.double)
 def _intSecAtan(x):
     # In : sympy.integrate(sp.sec(sp.atan(x)))
     # Out: x*sqrt(x**2 + 1)/2 + asinh(x)/2

this fixes it.