This change does 2 things:
- Skip 2 tests: they are not actually _testing_ anything, just running functions
Let's skip these for now until we can dedicate some time to investigate what
should be tested here. This will also help the tests breath a little bit.
- Add a small sleep to the async test. This test needs a little breathing room so
the procs don't get stepped on. I think this is what is causing the master to
shut down sometimes during test runs:
```
08:41:47 test_wheel_async (integration.netapi.test_client.NetapiClientTest)
08:41:47 [CPU:0.0%|MEM:13.2%] ... 12:41:47,280 [salt.utils.process :670 ][ERROR ] An un-handled exception from the multiprocessing process 'SignalHandlingMultiprocessingProcess-24' was caught:
08:41:47 Traceback (most recent call last):
08:41:47 File "/tmp/kitchen/testing/salt/utils/process.py", line 663, in _run
08:41:47 return self._original_run()
08:41:47 File "/usr/lib/python3.5/multiprocessing/process.py", line 93, in run
08:41:47 self._target(*self._args, **self._kwargs)
08:41:47 File "/tmp/kitchen/testing/salt/client/mixins.py", line 457, in _proc_function
08:41:47 salt.utils.daemonize()
08:41:47 File "/tmp/kitchen/testing/salt/utils/__init__.py", line 514, in daemonize
08:41:47 os.dup2(dev_null.fileno(), sys.stdout.fileno())
08:41:47 io.UnsupportedOperation: fileno
08:41:47 Process SignalHandlingMultiprocessingProcess-24:
08:41:47 Traceback (most recent call last):
08:41:47 File "/usr/lib/python3.5/multiprocessing/process.py", line 249, in _bootstrap
08:41:47 self.run()
08:41:47 File "/tmp/kitchen/testing/salt/utils/process.py", line 663, in _run
08:41:47 return self._original_run()
08:41:47 File "/usr/lib/python3.5/multiprocessing/process.py", line 93, in run
08:41:47 self._target(*self._args, **self._kwargs)
08:41:47 File "/tmp/kitchen/testing/salt/client/mixins.py", line 457, in _proc_function
08:41:47 salt.utils.daemonize()
08:41:47 File "/tmp/kitchen/testing/salt/utils/__init__.py", line 514, in daemonize
08:41:47 os.dup2(dev_null.fileno(), sys.stdout.fileno())
08:41:47 io.UnsupportedOperation: fileno
```
The fix in #49002 was incomplete and when run on Debian, it hits a
`ValueError` when the RELEASE variable is a string with a decimal.
This change fixes the error.
The `test_grains_append_val_already_present` test can be flaky, especially on
CentOS 7 when it gets bogged down. Let's give the test a little breathing room
by adding a sleep so the grains.append calls don't stack up too quickly.
Marked as flaky originally, but they're still flaky with undetermined
race-conditions. These are skipped in 2018.3 and forward, so let's skip
them here as well.
This test fails sometimes, so let's mark it as flaky.
I have tried to reproduce the failure and fix it, but to no avail.
Let's mark it as flaky for now.
When one match fails (and some of these can be flaky) the whole test fails.
This makes it somewhat difficult to track down why the test fails.
I am hoping we won't need the sleeps in this change when the tests are separate,
but let's see how that goes. It worked fine in my testing.
I also updated the test to use pytest notation, rather than unittest.
Git 2.18.0 changed the output format against which some of the asserts
in this test are performed. Instead of displaying the full path to the
worktree, it now references the branch name instead. This test updates
the asserts to look for the basename instead, which should allow it to
pass on Git releases both before and after 2.18.0.
Related to https://github.com/saltstack/salt-jenkins/issues/378
This is likely causing all of the flakiness we're seeing on test runs
when the master stops responding.
These tests should be in their own, isolated run, rather than with
the rest of the test suite.
