@htgoebel Can you go ahead and squash-merge this?

@htgoebel I would recommend merging this ASAP to resolve test failures.

Python 3.3 has reach EOL.

What is the urgent need to drop this? (I did not have time yet to check the test failures.)

Also when dropping all code specific to Py 3.3 should be removed.

Dependencies are dropping 3.3. Dropping all code is going to need a longer PR, and it's important to fix the failing tests.

But why they are now failing? They did not fail yesterday. Why can't they be fixed?

(Yes, I know, 3.3. is EOL. But we need a PyInstaller bug-fix release and can't drop Py3.3. support for this. Thus even removing Py 3.3 from the tests can not be done quickly.)

See pytest-dev/py#175.

So the simpler solution would be to pin pytest – as long as we no not require a new version.

Yes but that kills #3049. If it was worth killing, then I would say it's fine, but I don't see the point of working on an EOL Python. Even security issues are not fixed.

I'd just drop 3.3, hardly anyone uses it.

Here's the pip installs for pyinstaller from PyPI for the last month (via pypinfo --percent --pip pyinstaller pyversion), showing a very small fraction for 3.3:

When dropping Py 3.3, this should be done completely, this pull-request did only remove it from the tests.