Unlike other database adaptors, memcache adaptors can be shared among
transactions. So pool them in order to reduce the number of connections
established to memcached servers. And as the connections handle
thread-safety internally (by inheriting from threading.local), threads
can share the same connection instance.
Also, do not create a new connector each time we get disconnected, just
reconnect.
Also, memcached internally retries connection, so no need to retry it
ourselves.
Also, do not log when failing to update & delete entries in cache
server: this is a racy cache, each read much be checked anyway, so
failed writes are just business as usual.
Also, document a bit more why we catch "any" exception happening during
finish, and specify base exception class following python coding best
practices.
Also, use named constant for loggin levels for remaining log directives.
Also, display traceback when catching a connection issue during
__getitem__, as the original exception gets converted into a KeyError.

Business templates are installed by system user, which is a special
user not returned by getWrappedOwner. Because of this, the "fixing
problems or activating a disabled alarm is not allowed" error was
raised when checking if the owner of the alarm has manage portal
permission on the alarm.

This switches the implementation to explicit creation of the user
when user id is the system user, so that we have a user with the
permission to solve the alarm.

It did not allow iteration, probably since python 2.7.
We also need to allow objects implementing __reversed__ with a custom
type, I only found list.__reversed__

Commit 89a17f04 broke live tests using
ERP5TypeTestCase.publish when they are run from command line. Note that such
live test could not be run from ERP5.

More refactoring is needed because ERP5TypeLiveTestCase should only take care
of adapting ERP5TypeTestCaseMixin to an ERP5 environment, and runUnitTest
should never use it. Commit fd9219af looks a
wrong approach.

A change that also applies to traditional tests is that the password does not
need to be known when logging with basic authentication. If the given password
is empty (e.g. 'ERP5TypeTestCase:'), password validation passes regardless the
actual password: to keep things secure, this exception is only valid for the
test thread and during the call to publish(). This is a simple way to deal with
random passwords in live tests.

This results with calling correct methods on the server.

This hopefully fix the random issue of having test results not linked to
a project. Indeed, distributor/createTestResult was called in two different
times:
- by the testnode, with the project title
- by the runTestSuiteCommand, without the project title. This was
  not an issue most of the time, since the test result is already
  created. But if it was not there (like due to a cancelled test),
  a test result without project was created

erp5/tests/testERP5TestNode.py: Fix test_scalability_04_constructProfile. repository option is required for scalability test profile and it was added on commit 58d92a84.

erp5/tests/testERP5TestNode.py: Fix test_zzzz_scalability_19_xxxx. prepareFrontendMasterInstance was removed on commit 4048ab39.

erp5/util/testnode, scalability_test: Stop using a dummy frontend master and use host.vifib.net frontend with a valid SSL certificate instead. Always use https.

Conflicts:
	scalability_test/__init__.py

erp5/util/testnode: hosting subscription url was not updated once it was set. As a result, since testnode watched an old destroyed instance, the next test never started. Thus don't store hosting subscription url. It changes from time to time.

erp5.util.testbrowser: Refactor of openNoVisit() and {Image,Submit}Control for recent zope.testbrowser (followup of 5a0b3763).

Instead of sleeping, try using an API method until it stops raising
ConnectionError.

/reviewed-on !650

- hardcoded parameter removed
- minor changes in slapos master communicator

- allow slapos master communicator to supply software destruction
- runner reinstalls software on every commit (temporarily)
- refactor in software profile construction
- some try controls added

/reviewed-on nexedi/erp5!644

- before create a new instance checks if the current testsuite has an old one
- re-uses file-dictionary functions of frontend software checks
- minor fixes

/reviewed-on nexedi/erp5!615

By using `os.path.normpath` we also support paths with a trailing slash.

- requests to bootstrap and site availability urls using a script run with userhosts
- hardcoded parameters removed
- try-exception controls in communication with master
- general refactoring and cleanup