oven-sh
Execution order of nested tests is counter intuitive
What version of Bun is running?
1.1.4+fbe2fe0c3
What platform is your computer?
Darwin 23.3.0 arm64 arm
What steps can reproduce the bug?
create test file test.spec.ts in bun project
import { describe, it, expect } from 'bun:test';
describe('A', () => {
it('a', async () => {
expect(true).toBe(true);
});
it('b', async () => {
expect(true).toBe(true);
});
describe('1', () => {
it('a', async () => {
expect(true).toBe(true);
});
it('b', async () => {
expect(true).toBe(true);
});
});
describe('2', () => {
describe('1', () => {
it('a', async () => {
expect(true).toBe(true);
});
it('b', async () => {
expect(true).toBe(true);
});
});
describe('2', () => {
it('a', async () => {
expect(true).toBe(true);
});
it('b', async () => {
expect(true).toBe(true);
});
it('c', async () => {
expect(true).toBe(true);
});
});
describe('3', () => {
it('a', async () => {
expect(true).toBe(true);
});
it('b', async () => {
expect(true).toBe(true);
});
it('c', async () => {
expect(true).toBe(true);
});
});
});
describe('3', () => {
it('a', async () => {
expect(true).toBe(true);
});
it('b', async () => {
expect(true).toBe(true);
});
});
});
run
$ bun test
What is the expected behavior?
I expect the test execution with outer describe blocks being executed first.
...
test.spec.ts:
✓ A > a [0.01ms]
✓ A > b
✓ A > 1 > a [0.23ms]
✓ A > 1 > b [0.02ms]
✓ A > 2 > 1 > a [0.01ms]
✓ A > 2 > 1 > b [0.01ms]
✓ A > 2 > 2 > a [0.01ms]
✓ A > 2 > 2 > b [0.21ms]
✓ A > 2 > 2 > c
✓ A > 2 > 3 > a
✓ A > 2 > 3 > b [0.01ms]
✓ A > 2 > 3 > c
✓ A > 3 > a
✓ A > 3 > b
...
What do you see instead?
Instead I see an execution order that seems to follow no logic.
...
test.spec.ts:
✓ A > 1 > a [0.23ms]
✓ A > 1 > b [0.02ms]
✓ A > 2 > 1 > a [0.01ms]
✓ A > 2 > 1 > b [0.01ms]
✓ A > 2 > 2 > a [0.01ms]
✓ A > 2 > 2 > b [0.21ms]
✓ A > 2 > 2 > c
✓ A > 2 > 3 > a
✓ A > 2 > 3 > b [0.01ms]
✓ A > 2 > 3 > c
✓ A > 3 > a
✓ A > 3 > b
✓ A > a [0.01ms]
✓ A > b
...
One could argue that inner describe blocks should be executed first which does not make sense to me.
But here on A > 1 level it goes from outer to inner and on A level it executes the inner level A > 1-3 first and then the outer block A > a,b
Additional information
Assuming that inner tests build on top of outer tests they should also fail if a parent test fails. When I tell the runner to exit after the first failed test I can safe valuable runtime.
Imho this is not a duplicate of #5738 since it is independent from setup and teardown hooks.
having the same issue, working with a mocked queue is extremely difficult right now, on one describe i send a message through a service and check that it's sent, then in other describe i clear the mocks to do it again, the second describe clears the mock before i can check it on the firs one
Fixed by @pfgithub in #22534, and will be included in Bun v1.2.23 later tonight
