Describe the bug

The run =<expected-exit-status> convenience syntax (documented here) is broken, and does not work.

I believe anyone relying on the currently-documented syntax to assert a run command's exit status has been inadvertently testing bats' breakage, not their underlying command's success or failure.

To Reproduce

Running bats with the attached test file below should produce 4 test failures, according to the docs. Instead, it produces 2 passes and 2 failures. The correctly-failing-failures demonstrate that the alternative run -<expected-exit-status> syntax does work as expected.

$ bats foo.bats 
foo.bats
 ✓ This test should fail, but will succeed, #1
   ## Command that 'run' ran was: '=0 false'
 ✓ This test should fail, but will succeed, #2
   ## Command that 'run' ran was: '=1 true'
 ✗ This test should fail, and will fail, #1
   (in test file foo.bats, line 25)
     `run -0 false' failed, expected exit code 0, got 1
 ✗ This test should fail, and will fail, #2
   (in test file foo.bats, line 29)
     `run -1 true' failed, expected exit code 1, got 0

4 tests, 2 failures

Notice that the reporting line that shows BATS_RUN_COMMAND for each of the incorrectly-passing tests demonstrates that the =<exit-status> is injected as a prefix to the actual command the user wanted to test.

And because (to a reasonable approximation) ~100.00% of machines lack a command named =0/=1/etc, the fact that run (by design) masks the exit code of the erroneously-missing =0/=1 command, anyone using this syntax to test either success or failure will not have been testing -- or even executing -- the command they thought they were testing, since #479 was merged in October 2021.

#!/usr/bin/env bats

# Testing with this file should result in 4x failing tests.
# Instead, it will result in 2x failing and 2x passing tests.

report() {
  echo "# ## Command that 'run' ran was: '$BATS_RUN_COMMAND'" >&3
}

# 2x incorrectly passing tests

@test "This test should fail, but will succeed, #1" {
  run =0 false
  report
}
@test "This test should fail, but will succeed, #2" {
  run =1 true
  report
}

# 2x correctly failing tests, exactly the same as the previous 2 tests, but
# with "=<N>" changed to "-<N>"

@test "This test should fail, and will fail, #1" {
  run -0 false
  report
}
@test "This test should fail, and will fail, #2" {
  run -1 true
  report
}

Expected behavior

I expected 4 test failures.

Environment (please complete the following information):

• Bats Version 1.6.0 @ 5c7a96727535a91fcef50b22783f0a54e10821eb
• OS:

Distributor ID:	Debian
Description:	Debian GNU/Linux 11 (bullseye)
Release:	11
Codename:	bullseye

• Bash version: GNU bash, version 5.1.4(1)-release (x86_64-pc-linux-gnu)

Additional context

I have opened PR #579 to change the "Writing Tests" documentation to stop pointing users to the old syntax, and to use the current syntax.