winbot tests fail since nov 9

[ghost]

Originally reported by: agroszer (Bitbucket: agroszer, GitHub: agroszer)

---

since rev df4db882199e5d45ced6aff1da53d4c70fd6e6ef

e.g.
http://winbot.zope.org/builders/setuptools_master%20py_270_win32

---

• Bitbucket: https://bitbucket.org/pypa/setuptools/issue/101

[ghost]

Original comment by jaraco (Bitbucket: jaraco, GitHub: jaraco):

---

@philip_thiem Can you look into this? Digging into the failures, I see mentions of SvnDummy. I suspect, although Adam has blamed the 'bump' commit that the failure was actually triggered in a0586d7e713eb (merge of fix for #99) or its ancestors.

Adam, the tests pass on Travis - can you provide any insight into how your tests are run and why setuptools might not be importable at that stage?

Actually, I might have an idea. It may be these SvnDummy tests are invoking a setup script not in the root of the repo, and in that context, 'setuptools' won't be present from the repo, and the reason the tests pass on Travis is because setuptools is being invoked from the system location (not from the location under test).

[ghost]

Original comment by agroszer (Bitbucket: agroszer, GitHub: agroszer):

---

Invocation is simple:

cmd /c c:\Python27_32_clean\setupcompilerandexecute.bat c:\Python27_32_clean\python.exe setup.py test

where

c:\Python27_32_clean\python.exe setup.py test

is the relevant part.

c:\Python27_32_clean is a bare naked python install, probably ONLY with pywin32

[ghost]

Original comment by philip_thiem (Bitbucket: philip_thiem, GitHub: Unknown):

---

I have to admit my python 2.7 isn't clean but I do make sure setup.py passes on python 25-27 and 32-33 on my local win32 laptop. (All 32-bit).

So c:\Python27_32_clean is something that I can download from zope.org?

I have been noticing a error on win32 being throwing with atexit() errors at the end of the tests.

Also test.py does throw a deprecation warning (for direct .svn parsing) in one of the tests. This might be the reason. I could try capturing that for setup.py test to make the test output clean.

[ghost]

Original comment by agroszer (Bitbucket: agroszer, GitHub: agroszer):

---

no need to download anything

• install a python from python.org,
• optionally install pywin32,
• nothing else

that's the environment where winbot runs the tests

[ghost]

Original comment by philip_thiem (Bitbucket: philip_thiem, GitHub: Unknown):

---

Ok I I'll try a clean python

[ghost]

Original comment by philip_thiem (Bitbucket: philip_thiem, GitHub: Unknown):

---

Yes, there was a module issue like Jason thought. I think I have it fixed, will get the pull request in tonight.

[ghost]

Original comment by philip_thiem (Bitbucket: philip_thiem, GitHub: Unknown):

---

https://bitbucket.org/pypa/setuptools/pull-request/24/issue-101-and-tweak-legacy-fallback/diff

[ghost]

Original comment by jaraco (Bitbucket: jaraco, GitHub: jaraco):

---

For .svn legacy fallback, look for the files in the .svn not the directory.
(Fixed unexpected deprecation warning from prombredanne)
Also removed the warning from fallback, only a deprecation warning is issued.
Environment.py whitespacing
Created a specialized command executor for tests in Environment.py
Legacy Test in test_egg_info now supresses the deprecation warning.
PythonPath is now explicitly controlled to allow setup.py test on clean
python installations. Fixes Issue #101
Moved some dummy svn tests from test_sdist to test_egg_info since
they are egg_info tests.
Downgraded a with statement in a test since we haven't offically dropped
2.4 support, however, maybe it is time.
Added a test case to ensure no extranuous output on sdist with a simple
dummy package without rev ctrl.