UnicodeDecodeError on html.py

[nedbat]

Originally reported by Simon Coulton (Bitbucket: simoncoulton, GitHub: simoncoulton)

---

Using Python 3.2.3, the app occasionally fails to build the HTML required, and errors on line 277: status = pickle.load(fstatus). This could just be that the original status.dat file had some Python 2 code in it, in any case the following resolves the issue:

try:
if sys.version_info < (3, 0):
status = pickle.load(fstatus)
else:
u = pickle._Unpickler(fstatus)
u.encoding = 'latin1'
status = u.load()

---

• Bitbucket: https://bitbucket.org/ned/coveragepy/issue/193

[nedbat]

Could you send or attach a data file that has this problem? Also, can you explain more about how Python 2 and Python 3 are mixing in your environment?

[nedbat]

Original comment by Peter Portante (Bitbucket: portante, GitHub: portante)

---

Hi Ned, I seem to be encountering the same problem from pycscope.

I am using Fedora 16, where by default /usr/bin/python is a link to python2. I run the runtests script which just invokes nosetests like:

nosetests --with-coverage --cover-package pycscope -cover-erase --cover-html $@

When I then switch the default python by relinking /usr/bin/python to python3 I get the decode error.

If I remove the "cover" directory then python 3 coverage works.

For now I have specified a separate directory for my python2 and python2 coverage runs.

[nedbat]

Fixed in <<changeset 16020b834643 (bb)>>, but the old status is discarded and the coverage report recreated from scratch.

[nedbat]

Original comment by Simon Coulton (Bitbucket: simoncoulton, GitHub: simoncoulton)

---

Hi Ned,
Sorry for not getting back sooner, didn't see the emails come in. Using virtualenv with 3.2, the nose.cfg used was:
[nosetests]
cover-erase=1
with-coverage=1
cover-package=watson
cover-html=1
cover-html-dir=docs/tests/html
with-xunit=1
xunit-file=docs/tests/nosetests.xml
nocapture=1

Thanks for the fix :)