Use command rm in alias_completion

[NoahGorny]

Closes #1354

[NoahGorny]

Works, but the test always passes, as alias-completion exists early, which makes it harder to test it

[davidpfarrell]

Does it exit early because there are no completions ie:

 (( ${#completions[@]} == 0 )) && return 0

?

If that were so, then if feels like the tests above it aren't really testing anything?

[NoahGorny]

Does it exit early because there are no completions ie:

 (( ${#completions[@]} == 0 )) && return 0

?

If that were so, then if feels like the tests above it aren't really testing anything?

Yes.. Although I am sure I made it fail the test someway.. This needs to be fixed

[NoahGorny]

@davidpfarrell The completion file I loaded was not a good one, replaced with a simpler completion file and now it should work

[davidpfarrell]

In order for the test to be useful, we need a way to prove that command rm gets invoked.
A quick idea:

• Path-munge a test-specific rm that does something like mv $1 $1.rm (would probably need more logic)
• Make the alias invoke something like false
• Confirm that file.rm exists
  If the alias gets called, then file.rm will not exist.

[NoahGorny]

It's probably easier to just make the alias to mv the file, and assert that it does not exist afterwards

[davidpfarrell]

Well we want the test to confirm a few things:

1. The script section that contains rm was executed
2. The rm command was invoked
3. The rm alias was not invoked

Passing the test requires confirming some observable behavior that only exists when all of these are true.

The existing refute_output check for example, is poor because, although it can confirm if #3 fails, it can't confirm that #1 passes.

So by path-munging a custom rm script, you get the ability to create some type of observable outcome that proves it was invoked. Additionally, in this case, having the alias invoke the path-munged command could muddy the test, unless it somehow affects the observed results.

NOTE: I actually don't know if trying to accurately test this behavior is worth the effort

[NoahGorny]

@davidpfarrell take a look at #1757, I think this PR addresses this problem