Test fixes for RedHat

[spevans]

• Process: Make xdgTestHelper into a self-contained file

• Build as part of DarwinCompatibilityTests.

• Add TestBundle.xdgTestHelperURL() to return the URL of the
  executable.

• Process: When setting the current directory for a sub-process,
  set the PWD environment variable to this directory if the
  default environment is used. This matches Darwin's behaviour.

• Use xdgTestHelper as an external test program for TestProcess,
  taking options for getcwd() and getenv("PWD").

• SR-7748: Use /tmp as a test directory and run it through realpath()
  to test against the result of getcwd(). On macOS, /tmp is a symlink
  to private/tmp.

• SR-7747: FileManager.attributesOfFileSystem(forPath:) may return
  .systemNumber as 0 on some systems (eg tmpfs on Redhat). Remove
  this invalid test.

[spevans]

@swift-ci please test

[davezarzycki]

1. I think test_current_working_directory() is trying to verify that the caller's current working directory doesn't change when running a task in a different working directory. Therefore, changing the task's working directory to the root directory is at greater risk of being a "no op" test. I'd just drop the "pwd" assert and add a comment to the test about what the goal of the test is.
2. I'd also add a comment to the code that initializes systemNumber to note that tmpfs on Linux can return zero.

[spevans]

test_current_working_directory() is actually testing both the child process uses .currentDirectoryPath correctly and that the parent process retains its working directory. Although some extra testing has exposed a possible issue where PWD should be set in the environment so I will fix this and update the tests.

For systemNumber I don't think a comment is actually required, the test was based on the wrong assumption that it could never be 0.

[davezarzycki]

Seems reasonable

[spevans]

@swift-ci please test

[parkera]

This seems ok, but it's somewhat ironic that bundle can't find this helper itself. That's what it's supposed to be there fore in the first place.

[millenomi]

The issue here is that we are not building xdgTestHelper in a bundle directory structure. It may be worth it to figure that out.

[spevans]

I'll aim to do that in a follow-up PR as it is complicated by the fact that there are 3 separate targets being built (Linux, macOS and the Darwin Compatibility testing). However it should act as a good test of the Bundle class so it is definitely worth doing.

[spevans]

@davezarzycki Are you able to give this a test on redhat?

[millenomi]

@swift-ci please test

[davezarzycki]

I haven't tested them, but they look fine.

[spevans]

@swift-ci please test and merge

[spevans]

@swift-ci please test and merge