Thanks for the patch.

-        # Test that the isntalled raises an exception if the project does not
+        # Test that the installed raises an exception if the project does not
It took me many seconds to find the change :)

+    def test_installation_get_infos_with_ClientWrapper(self):
+        # Test the use of get_infos default index
Hm, I don’t like the method name, and don’t really understand from the comment what it is exactly that you’re testing (I’m not very familiar with p7g.pypi).

-        install.install_dists = lambda x, y=None: None
+        install.install_dists = lambda x, y = None: None
PEP 8: Never put spaces in a function (or lambda) signature.

Let me remark that even if we get to 100% line coverage, or even 100% branch coverage, that won’t mean that we have covered everything.  Two personal anecdotes to illustrate my point.  In one project, I had a branch covered but actually untested, because I used a ternary operator (something like “flags = 0 if case_sensitive else re.I”).  Another project was a Pylons web app; I had full coverage, and then I tried using non-ASCII and saw that my app was incomplete.  So I think that full code coverage is mainly useful in a new project, where you can have 100% all the time and use that metric to avoid adding code without sufficient tests.  For a codebase like distutils2 that’s half legacy, half new stuff, it’s harder to achieve that, and maybe less useful than working on other things.

If you grep packaging tests for XXX|TODO|FIXME, you’ll find 22 entries.  Fixing those may not have an impact on coverage numbers, but will definitely improve things.  There are also XXX notes in the code itself, open bugs on this tracker, and some dozen more in my todo lists.  For example, packaging.database is supposed to handle zipped eggs, but does it?  I really appreciate your contributions, and think that you know enough of the code now to take on a larger bug if you want to.  :)