clear AND reload analysis config on tests’ tearDown

[JGreenlee]

There are several tests that call etc.set_analysis_config in their setUp() to set config options in certain testing scenarios. This creates a config file, and in tearDown of each test the config file is deleted via os.remove.

However, just deleting the file is not enough!

When emission.analysis.config is intialized, we read the config and cache it in a variable config_data

e-mission-server/emission/analysis/config.py

Line 19 in 4034533

config_data = get_config_data()

When get_config() is called while the pipeline is running, it returns the value of config_data, which can be out of sync with the contents of the config file if the config file was added/removed/modified and reload_config() was not called. So instead of just calling os.remove() these tests also need to call reload_config() in their tearDown().

Added this by way of a new function in emission.tests.common

Also, I made the config file pathnames into constants for tidiness and to eliminate the risk of typo-related bugs

[shankari]

@JGreenlee I have merged this since it is clearly the right thing to do, but I don't understand how the analysis config would affect the pipeline timestamps, which is what you reported in

#993 (comment)

We do not understand why the pipeline start ts comes up as 1440688707.867 during runAllTests.sh.

Is the pipeline not run properly because of the configs? But in that case, the pipeline range should not be set at all (it should be like a new user), right? Can you explain this further?

[JGreenlee]

Is the pipeline not run properly because of the configs? But in that case, the pipeline range should not be set at all (it should be like a new user), right? Can you explain this further?

The pipeline runs properly but behaves differently depending on "intake.cleaning.filter_accuracy.enable" in the analysis config.
When true, the first trip on Aug 27 starts at 1440688707.867.
When false, that trip starts a bit later at 1440688739.672.

When the test was run in isolation, "intake.cleaning.filter_accuracy.enable" was false.
When the test was run during runAllTests.sh, it was true.