PyPy's zipimport._zip_directory_cache implementation does not support direct
item assignment, thus breaking our attempts at silently updating the cached zip
archive directory information behind the scene. As a workaround, when using
PyPy, we fall back to simply clearing the stale cached information.

--HG--
extra : amend_source : 991e30244754d8fac042da56ac4cf0ad1a0f50d5

This is an extra safety measure to avoid someone holding a reference to this
cached data and using its content even after we know that the underlying zip
archive has been removed and possibly even replaced.

Change suggested by PJ Eby (pje on BitBucket) in a setuptools pull request #51
comment:
  https://bitbucket.org/pypa/setuptools/pull-request/51/diff#comment-2018183

--HG--
extra : rebase_source : 6de2309bc7446647749cfe78ab00e0230a07f92f

_update_zipimporter_cache() extracted from _uncache() &
_replace_zip_directory_cache_data().

Code cleanup done in preparation for adding a bit more detailed cache item
clearing logic, so that would not require adding a separate function with yet
more code duplication.

--HG--
extra : rebase_source : e2e956e042c7cbfabe2c31ecc58a4f76c91f40aa

Extracted code for collecting a list of zipimporter cache entries related to a
given path into _collect_zipimporter_cache_entries().

--HG--
extra : rebase_source : 54ab881d794f95467e811511433a2cd31595339e

--HG--
extra : rebase_source : c8c77d96880275e34c1580991c2d70486b6d0e00

Change-Id: Ic3cc25a02de71b94a08f0bf64e8d8b01b572a23b

--HG--
extra : rebase_source : a49971d71570380f1ef51a88897d72328de337ed

pypy uses a custom zipimport._zip_directory_cache implementation class that
does not support the complete dict interface, e.g. it does not support the
dict.pop() method.

For more detailed information see the following links:
  https://bitbucket.org/pypa/setuptools/issue/202/more-robust-zipimporter-cache-invalidation#comment-10495960
  https://bitbucket.org/pypy/pypy/src/dd07756a34a41f674c0cacfbc8ae1d4cc9ea2ae4/pypy/module/zipimport/interp_zipimport.py#cl-99

--HG--
extra : rebase_source : 95cff7946455f0a4422d97eecab11164a9ddef10

Change-Id: I6789ef2eda75748597c9ae76f2a5389140a9daab

--HG--
extra : rebase_source : 48bd79fb7ee229a89a3796df14aadbcca8ace016

Set up integration tests that install packages to temporary
directories.

Change-Id: Iec90838fec961228fca24c0decc088de55303350

--HG--
extra : rebase_source : f5219f8411db4b79694a74659e22b0c0b1c771ab

When the tests are run, some packages are installed or created in the
current directory. They should never be checked in, so add them to
.gitignore.

Change-Id: Ic7790ec85c9e2f0fc4bd06b492f75ea6c17ecda4

--HG--
extra : rebase_source : 7ad1502fd5d09d38f521a391e22af46cfce6ea4f

Now that 2to3 is no longer run on the codebase, it's safe for the templates to be syntactically incorrect (prior to substitution).

--HG--
extra : source : f78ffe97964c6becbf42e946bf6504e8dc09f13b

https://bitbucket.org/pypa/setuptools/issue/213/regression-setuptools-37-installation

--HG--
extra : source : 182f68beacf5e436609fb7d1064a18279cbbd24a

--HG--
extra : source : 0d6a065483104cc9781a3ff72c168f059ec11036

The rename of the script template files to be .tmpl put them
into the realm of package data, rather than python files that
would be bundled automatically. Include them specifically in
package data so that they'll actually be installed.

--HG--
rename : setuptools/script template (dev).py => setuptools/script (dev).tmpl
rename : setuptools/script template.py => setuptools/script.tmpl