Class definitions interpreted as branches

[arttuperala]

Describe the bug
Using coverage run --branch reports an unexpected number of branches in a file with class definitions when using coverage >= 6.2.

To Reproduce

class MyClass:
    def my_method(self):
        return True

    @property
    def my_property(self):
        return True


my_class = MyClass()
my_class.my_method()
my_class.my_property

$ coverage --version
Coverage.py, version 6.4.4 with C extension
Full documentation is at https://coverage.readthedocs.io
$ coverage run --branch test.py && coverage report
Name      Stmts   Miss Branch BrPart  Cover
-------------------------------------------
test.py       9      0      2      0   100%
-------------------------------------------
TOTAL         9      0      2      0   100%

Generating the result XML file shows that the branch statement comes from the first line in the test file.

<class name="test.py" filename="test.py" complexity="0" line-rate="1" branch-rate="1">
    <methods/>
    <lines>
        <line number="1" hits="1" branch="true" condition-coverage="100% (2/2)"/>
        <line number="2" hits="1"/>
        <line number="3" hits="1"/>
        <line number="5" hits="1"/>
        <line number="6" hits="1"/>
        <line number="7" hits="1"/>
        <line number="10" hits="1"/>
        <line number="11" hits="1"/>
        <line number="12" hits="1"/>
    </lines>
</class>

Downgrading to 6.1.2 yields a different (but expected) result.

$ coverage --version
Coverage.py, version 6.1.2 with C extension
Full documentation is at https://coverage.readthedocs.io
$ coverage run --branch test.py && coverage report
Name      Stmts   Miss Branch BrPart  Cover
-------------------------------------------
test.py       9      0      0      0   100%
-------------------------------------------
TOTAL         9      0      0      0   100%

6.2 is the earliest version where I have found the unexpected behaviour.

$ coverage --version
Coverage.py, version 6.2 with C extension
Full documentation is at https://coverage.readthedocs.io
$ coverage run --branch test.py && coverage report
Name      Stmts   Miss Branch BrPart  Cover
-------------------------------------------
test.py       9      0      2      0   100%
-------------------------------------------
TOTAL         9      0      2      0   100%

Tested with Python 3.8.14, 3.9.13, 3.10.6 on macOS 12.6 (x86) with a minimal virtualenv only containing the tested version of coverage.

Expected behavior
No branch statements are reported in a file with a class definition but no branch statements (coverage 6.1.2 behaviour).

[nedbat]

Hmm, I'm not sure why I changed this in 6.2, but I think you are right: classes shouldn't be considered branches. I've put back the old behavior in commit fd36918.

A remaining mystery is why classes were considered branches when def lines were not...

[nedbat]

This is now released as part of coverage 6.5.0.