Adjust the gfortran regex to require a '.' to be present in the version number.
This avoids matching a hash or text containing a '-' such as the version string
from the compilers produced by crosstool-NG.  For example:
GNU Fortran (crosstool-NG 8a21ab48) 7.2.0

Note that this was discussed extensively in gh-7287, but removing
this -xhost flag was missed in the PR that closed that issue.

[ci skip]

Issue was introduced in gh-8454.

Thanks to @jennirinker for pointing out the issue and fix.  Closes gh-9553.

If we are using the virtualenv module in Python 2.7 (as opposed to venv
in Python 3.3+), then the location of the base distribution is stored in
the real_prefix attribute instead of the base_prefix attribute.

This changeset adds the content of this attribute to the search paths if
it exist (i.e. in practice, if we are running Python 2.7), so that the
runtime libraries do not have to be duplicated in the virtualenv
directory.

The original changeset 1c8ecc7 to make extensions compilable with MinGW
used the base_prefix attribute unconditionally. However, this is not
available in versions before Python 3.3.

This changeset introduces a check for whether the attribute is available
so that the code will work without errors also on Python 2.7. However,
this may lead to surprising results when compiling in a virtualenv.

Hat tip to Diorcet Yann (@diorcety) for pointing out this problem.

Fixes two problems:

* c compilers do not have a find_executables method.
* get_version return a LooseVersion instance, not string.

Closes #9278.

[ci skip]

Echo is a command in the Windows command interpreter

Python 3.5 also added build parallelization at the extension level
instead of the file leve numpy uses.
This causes two problems:

- numpy.distutils is not threadsafe with duplicated source files
When source files are duplicated in multiple extensions the output
objects are overwritten which can truncate in a parallel context.
This is fixed by keeping track of the files being worked on and wait
for completion if another thread is already using the object name.

- The parallelization on two nested levels causes oversubscription.
When building multiple extensions with multiple source files the number
of jobs running is multiplied.
This is fixed by adding a semaphore that limits the number of jobs numpy
starts to the defined amount.

closes gh-7139