How to assert exit code 1 ?

[limboShadow]

I used @assert-fail to assert exit code 1 When something run fail, but it does't work.
This is my example:

source bach.sh
test-something-fail(){
    run_something_fail
    @assert-fail
}

run_something_fail(){
    exit 1
}

And I got the error:

1..1
not ok 1 - something fail


# -----
# All tests: 1, failed: 1, skipped: 0

So, what's wrong with this? what should I do？

[chaifeng]

There are two ways.

The first one is using test-ASSERT-FAIL- prefix for the testing function. This prefix tells Bach Testing Framework that this testing function should fail. In this case, we don't need to write the corresponding asserting function.

test-ASSERT-FAIL-something-fail(){
    run_something_fail
    @assert-fail
}

But I do recommend using another way, using separate testing and asserting functions

test-something-fail-1(){
    run_something_fail
}
test-something-fail-1-assert(){
    @false
}

[chaifeng]

The root cause is this function

run_something_fail(){
    exit 1 # should be `return 1`
}

Because of exit 1, the second line of code in the function test-ASSERT-FAIL-something-fail will never be executed. So @assert-fail will never be executed.

#!/usr/bin/env bash

source ./bach.sh
test-something-fail-1(){
    run_something_fail
}
test-something-fail-1-assert(){
    @false
}

test-something-fail(){
    run_something_fail
    @assert-fail
}

run_something_fail(){
    return 1 # here
}

[limboShadow]

Thank you your answer！I can run my unit test correctly after modifying exit .
But The return only can use in a function or sourced script that mean sometimes we have to use the exit when error occurs.
How to fix the situation in Bach Testing Framework？or this situation beyonds the scope of unit test ？

[chaifeng]

I don't recommend mocking the exit command. But we can still do it with Bach.

#!/usr/bin/env bash

source ./bach.sh
test-something-fail-1(){
    @mock exit 1
    run_something_fail
}
test-something-fail-1-assert(){
    blabla
    @dryrun exit 1 # We use this API `@dryrun` explicitly since the `exit` command is not mocked
    should-not-be-executed
}

test-something-fail(){
    run_something_fail
}

test-something-fail-assert(){
    blabla
    exit 1
    should-not-be-executed
}

run_something_fail(){
    blabla
    exit 1 # here
    should-not-be-executed
}