Python 3 bug in zipfile

[embray]

There's a doctest that invokes this bug in Python 3.6's zipfile module:

sage: from sage.plot.plot3d.shapes import Cylinder
sage: C = Cylinder(3, 1, closed=False)
sage: C.jmol_repr(C.testing_render_params())
Exception ignored in: <bound method ZipFile.__del__ of <zipfile.ZipFile [closed]>>
Traceback (most recent call last):
  File "/home/embray/src/sagemath/sage-python3/local/lib/python3.6/zipfile.py", line 1663, in __del__
    self.close()
  File "/home/embray/src/sagemath/sage-python3/local/lib/python3.6/zipfile.py", line 1681, in close
    self._write_end_record()
  File "/home/embray/src/sagemath/sage-python3/local/lib/python3.6/zipfile.py", line 1783, in _write_end_record
    centDirSize, centDirOffset, len(self._comment))
struct.error: argument out of range
['pmesh obj_1 "obj_1.pmesh"\ncolor pmesh  [102,102,255]']

The Graphics3d.testing_render_params() opens a ZipFile in write mode to /dev/null, so the bug seems to have to do with writing a zipfile to /dev/null (probably to do with a tell() call somewhere).

For the purposes of testing there's no reason this needs to go to '/dev/null'. It can just as well use a temp file, so we can implement that workaround. But we should also open an upstream bug report for the zipfile problem, if it hasn't already been reported.

Upstream bug report: ​https://bugs.python.org/issue34904

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​60fb5bf

trac 26301 replace dev/null by temporary file

[embray]

Sure, looks good to me. I still need to make sure to make an upstream bug report (but this is pretty minor, and after fixing it in Sage I doubt we'll revisit the issue).

[embray]

Interestingly, I can't reproduce the bug on Cygwin, only on Linux. I was trying to see if it was still broken in master, and at first I thought maybe it was fixed. But no, it just doesn't happen on Cygwin that I can tell. Maybe some difference in how /dev/null works; not sure.

[embray]

Ok, upstream bug report submitted. LGTM.