Add more tests with per_file = False

[nettle]

Describe the task

It is easier to compare new implementation (per_file = True) with reference, old and stable implementation (per_file = False).
Besides old implementation should be tested too.

Scope

1. Add more test cases with per_file = False
2. Make sure all tests pass

Definition of Done

Ideally all tests should have old and new versions

[furtib]

We have usually added a per_file = False for every per_file = True option, but there are some exceptions:

• argument merge (Add test for CodeChecker argument merging #87) - since then, the paired feature has gone out of scope. This test should be reverted(?).
• caching (Unit test: caching #59) - There is no point in testing per_file = False caching, there is none.
• compile_flags (Add test for compile flag filtering #78) - The per_file = True option should also be removed after Use the compile_commands implementation from the monolithic rule in the per_file rule #98; the only reason it's even there is because it had its own compile_commands.json generator.

I have checked in every other test codechecker_test-s come in pairs, one per_file = True and one per_file = False

[nettle]

Thanks for the comment @furtib! I see now :)
I think I have a couple of question though:

• About caching test - why it is not applicable for old (per_file = False) case?
  I mean we should also check that we re-run analysis if a file change, shouldnt we?
• What is the purpose of argument_merge test? I just did not get it

[furtib]

The case of the caching test is an interesting issue.

My view on this is this: I would think that if any file is missing during analysis, then the legacy tests would break. But if the files are in the input field, then Bazel should be able to determine if it must be re-run. So I view the testing of caching in the False rule as testing Bazel itself.

But I do see now that I might not be completely correct in this case. I will add a test for False.

If we really want to dissect the argument_merge test. I was working on an issue (I think it was #77) and blamed the per_file script getting multiple of the same flag as the issue (it wasn't). But the feature PR (#88) and the test have been created. As I said, this feature went out of scope. Moving forward, we either remove the test or merge the PR. (In case we merge, the test won't be relevant for False tests, since they do not have default arguments)