by calling pip.main() directly there since it should be available in
those version of pip that are able to run on it and getting rid of
a conditional expression.

(cherry picked from commit cd1d9e4a)

using those included in virtualenv 16.0.0 as the baseline.

Since pip 10.0.0 has dropped support of Python 2.6 and we plan to
keep it until PyCharm 2019.1, now we need to bundle older compatible
versions of all related packages as we previously did for Python <= 2.5.

(cherry picked from commit bff4ba25)

(cherry picked from commit c11f8dd7)

On Python < 3.3 we rely on virtualenv script to update setuptools and
pip in created virtualenvs instead of using those we bundle. On the
affected version of MacOS virtualenv failed to download newer versions
of the packages and, thus, fell back to pip 9.0.1 (not able to handle
newer TLS on pypi.org) that it's shipped with itself.

(cherry picked from commit 48ade29f)

(cherry picked from commit 86f5b376)

that are identical to their versions shipped with Python 3.6.5.

Also, I included setuptools 36.8.0 which is the last supporting
Python 2.6 according to their changelog.

(cherry picked from commit d7d8616a)

(cherry picked from commit 5d101bfa)

since the only place where it was actually used is launching pip, and
the latter terminates the process with sys.exit() itself now thanks to
runpy.run_module().

(cherry picked from commit d77d0ac5)

Previously in case of error pip run with runpy.run_module() terminated
the execution of the script abruptly with SystemExit and this cleanup
wasn't performed.

Normally it didn't happen though, since temp directory was created on
Java side in PyPackageManagerImpl.install() that was responsible for
its removal.

(cherry picked from commit d280c276)

packaging_tool.py is terminated with the right exit code from pip
anyway with this approach because pip.__main__ calls sys.exit() itself.

(cherry picked from commit 7e9a2b87)

(cherry picked from commit ea562342)

(cherry picked from commit 38ece86a)

(cherry picked from commit d9f32f65)

(cherry picked from commit 1b469a75)

Some of which still use no longer public API of pip

(cherry picked from commit ec95d05c)

It's obviously only a workaround, we should find a better way
not to use internal modules of the package altogether.

(cherry picked from commit a7ec5833)

(cherry picked from commit 5252a51)

this is simple fix for correct processing of removed files (IDEA-181227, IDEA-160008)

(cherry picked from commit a9c04718), IDEA-CR-27877

When compacting fragmented PersistentHashMap in some cases records got
content not in the write order. In particular, append calls can produce
fragments A,B,C and after compact record read loads C,B,A

(cherry picked from commit beaf591d), IDEA-CR-28595

Backport from 181 (886c08c1, 29b9f231, 8bb0035a, 4f7fd895, 44b3bc0) but not possible to use Maven libs in this version.
Related reviews - IDEA-CR-29575, IDEA-CR-29739, IDEA-CR-29651

Additional change - use the lib both in git and svn.

Fixes IDEA-140196

Before the change all modifiers were computed and cached.
Now each modifier is handled separately.

(cherry picked from commit d8421f3e)

(cherry picked from commit 54e8d422)

(cherry picked from commit 759c9082)

(cherry picked from commit df9fe451)

(cherry picked from commit b59e4cad)

(cherry picked from commit fa618c74)(Reviewed by Roman Shevchenko and Vladimir Krivosheev)

(cherry picked from commit 276d27c5)
(cherry picked from commit 3c481325)

(cherry picked from commit 8e19d90c)

https://youtrack.jetbrains.com/issue/IDEA-180954

(cherry picked from commit 4af803cb)

(cherry picked from commit dc0a2903), IDEA-CR-25878

(cherry picked from commit e214b837)
IDEA-CR-25505

(cherry picked from commit 5954ef17)
IDEA-CR-25050

(cherry picked from commit 1b2e03a7)
IDEA-CR-24775

(cherry picked from commit 63399396)

(cherry picked from commit 9754b97e)

(cherry picked from commit d5757454)

(cherry picked from commit d7fc83aa)

Unfortunately, neither the reporters nor me have a reproducible example
for this one.

(cherry picked from commit c264ec31)

Since the switch to the builtins from Typeshed we no longer have
the definitions for True, False, None for PY2. We have to ignore these
references via null highlighting severity as we do in the base method
for PyReferenceImpl.

(cherry picked from commit cb22683f)

(cherry picked from commit 85ace079)

We already to it for regular (positional/keyword) parameters even
though (**kwargs) is not strictly compatible with (arg1, arg2, ...).
Now we do it for (*, kwarg1, kwarg2, ...). We don't know much about
**kwargs so it's better not to show any warnings here.

(cherry picked from commit 235b76c5)