Problem with coverage and sqlalchemy declarative synonym?

[Doug Latornell]

Over on the TurboGears list a TG2 user pointed out a problem that
arises when nosetests --with-coverage is run on a project with a
sqlalchemy identity model: http://groups.google.com/group/turbogears/t/7fd3639a5a4d4b8c

I dug into it and have reproduced the problem outside of TurboGears 2
and without nose: http://paste.turbogears.org/paste/7051

I've also shown that the problem isn't there for a plain Python
property (in contrast to a sqlalchemy
synonym): http://paste.turbogears.org/paste/7052

So, it looks like a problem maybe with the metaclass that sqlalchemy
uses to build properties, or with coverage not understanding what that
metaclass produces, or something. Or is there something else I'm
missing?

Doug

[jason kirtland]

Seems to be a problem in coverage. This narrows down a problem that was
first seen only when running inside Bitten.

$ cat c.py
def foo(somearg):
class Bar(object):
def __getattr__(self, attribute):
return getattr(somearg, attribute)

assert 'somearg' not in Bar.__dict__

foo('hi')

$ python c.py
$ coverage -e
$ coverage -x c.py
Traceback (most recent call last):
File "/Users/jek/bin/coverage", line 8, in <module>
load_entry_point('coverage==2.80', 'console_scripts', 'coverage')()
File "build/bdist.macosx-10.5-i386/egg/coverage.py", line 978, in main
File "build/bdist.macosx-10.5-i386/egg/coverage.py", line 398, in
command_line
File "c.py", line 8, in <module>
foo('hi')
File "c.py", line 6, in foo
assert 'somearg' not in Bar.__dict__
AssertionError

[Michael Bayer]

I've seen this identical issue with Bitten, and its not a SQLAlchemy
one.

Here's the ticket: http://www.sqlalchemy.org/trac/ticket/1138

Replace your test case with this script:

def foo(somearg):
class Bar(object):
def __getattr__(self, attribute):
return getattr(somearg, attribute)

assert 'somearg' not in Bar.__dict__

foo('hi')

I just ran it with the "coverage.py" that we have in SQLAlchemy and
the bug reproduces.

[Doug Latornell]

Thanks for the quick replies, Jason and Michael, and for the very
succinct test case. I will try to raise this with Ned Bachelor over
at http://nedbatchelder.com/code/modules/coverage.html

Doug

[Doug Latornell]

With some help from Ned Batchelder I was able to confirm that this is
a Python bug: http://bugs.python.org/issue1569356 that has been fixed
since the 2.5.2 release. Ned confirmed that the fix is included in
Python 2.6a3.

Doug

On Sep 16, 5:22 pm, Doug Latornell <djlatorn...@gmail.com> wrote:
> Thanks for the quick replies, Jason and Michael, and for the very
> succinct test case.  I will try to raise this with Ned Bachelor over

> athttp://nedbatchelder.com/code/modules/coverage.html

[Michael Bayer]

On Sep 18, 2008, at 4:08 PM, Doug Latornell wrote:

>
> With some help from Ned Batchelder I was able to confirm that this is
> a Python bug: http://bugs.python.org/issue1569356 that has been fixed
> since the 2.5.2 release. Ned confirmed that the fix is included in
> Python 2.6a3.

well thats....peachy. I suppose we might have to find a way to work
around it in SQLA (changing names in that area I think).