Prevent save config to home in tests

[GateBuilder]

This PR removes redundant and unnecessary execution of save_to_home function in tests.
The functionality of the corresponding context managers are not effected and they do not need this command. Although I tested their functionality without this attribute, please confirm that this is indeed the case.

@jenshnielsen

[jenshnielsen]

@GateBuilder I suspect but I am not sure that these context managers does this because there is a bunch of places in the code where a new config is generated by calling Config() these really should read the config from the default qcodes.config instead. We may need to change those. The way to check I think is to see if the location and work_station not set warning is triggered when running the tests after all the tests passes

[jenshnielsen]

For refererence the config instance creation below all look suspicious

qcodes\tests\test_config.py
177:        qcodes.config = Config()
224:        self.conf = Config()
322:        cfg = Config()
341:    cfg = Config()
363:        cfg = Config()

qcodes\tests\dataset\test_guids.py
24:    ocfg: DotDict = Config().current_config
30:        cfg = Config()
41:        cfg = Config()
66:    orig_cfg = Config().current_config
73:        cfg = Config().current_config
86:    orig_cfg = Config().current_config
93:        cfg = Config().current_config
129:        cfg = Config()

qcodes\tests\dataset\test_database_creation_and_upgrading.py
51:    cfg = Config()

qcodes\tests\dataset\test_database_extract_runs.py
514:    cfg = Config()

qcodes\dataset\sqlite\queries.py
1596:    cfg = Config()

qcodes\dataset\guids.py
28:    cfg = Config()
84:    cfg = Config()
109:    cfg = Config()

[jenshnielsen]

These calls to save_to_home could also be problematic if they are called by tests

qcodes\dataset\guids.py
102:    cfg.save_to_home()
128:    cfg.save_to_home()

We may need to make sure that we set the home dir to a isolated temp dir before any of these are called in a test

[GateBuilder]

@jenshnielsen I tested the one of the context managers without save_to_home where I first change my location and work_station numbers. They returned back the original state in the config file at my home directory without a problem.

[jenshnielsen]

@GateBuilder great lets hope that is the case for all of them :)

[GateBuilder]

@jenshnielsen Ok I understand what you mean, indeed why don't we use the default config?. Let me change this PR to a draft; rather tricky stuff here I do not wanna rush.

[GateBuilder]

I have made some changes in separate tests and files. The followings are the remaining problems?

• I cannot get rid off two Config() calls. Both resides in test_config.py. First one is in line 177. Removing this one while does not effect the test itself, causes failure of the test test_subscribing.py which uses the context manager created in the former. Second one is in line 224. Removing this call or replacing it with in memory config results in multiple failures. I could not figure out a way.

• Good news is with these changes now xdist works for 2 cores. For higher core numbers for example 4, I got 2 failures relating to logging. Hopefully can be solved.

Overall, this is a fine improvement. Please let me know if you have any suggestions concerning getting rid off remaining Config() calls.

@jenshnielsen @astafan8 @lakhotiaharshit

[jenshnielsen]

I think we should land this as is if passing and then revisit the remaining ones

[codecov]

Codecov Report

Merging #2016 into master will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #2016   +/-   ##
=======================================
  Coverage   71.31%   71.31%           
=======================================
  Files         149      149           
  Lines       19866    19866           
=======================================
  Hits        14167    14167           
  Misses       5699     5699