Update fixtures.rst w/ finalizer order

[aizpurua23a]

Fixes #10023
Add note to clarify the difference in execution order between after-yield finalizers and request.addfinalizer finalizers.

[nicoddemus]

Thanks a lot @aizpurua23a! Left a few comments, please take a look.

[nicoddemus]

This is not a good example I think, because of course both prints will be executed sequentially (pytest cannot mess with how Python executes code within a function).

I suggest we use two fixtures:

    @pytest.fixture
    def fix_w_yield1():
        yield
        print("after_yield_1")

    @pytest.fixture
    def fix_w_yield2():
        yield
        print("after_yield_2")

And then make test_bar request both:

    def test_bar(fix_w_yield1, fix_w_yield2):
        print("test_bar")

[aizpurua23a]

Ah, I see.
The execution order with the addfinalizer method is evident, so the execution order only needs clarification in this case.
Lovely! How about now?

[nicoddemus]

I think we can keep the examples using addfinalizer as they are useful for comparision, so could you please bring them back? Thanks

[aizpurua23a]

Sure thing, re-added.

[nicoddemus]

I think we can keep the examples using addfinalizer as they are useful for comparision, so could you please bring them back? Thanks

[nicoddemus]

while operations after yield are executed sequentially.

This sentence needs tweaking if we look at the output. 👍

Finalizers are executed in FILO order, that is correct, because we can see that adding fin2 and fin1 we get back fn1, fn2 in the output.

But the same can be said for the yield fixtures: we request them in fix_w_yield1, fix_w_yield2 order, and we get fix_w_yield2 and fix_w_yield1 in the output.

That's because yield fixtures use addfinalizer behind the scenes: when the fixture executes, we register a function with addfinalizer that resumes the generator (which in turn calls the teardown code).

Not sure how to "fix" this sentence, or perhaps we can explain something along the lines of what I wrote above, if that makes things clearer.

[aizpurua23a]

Yeah, that makes sense.
I tried to explain how both examples show the similar FILO behavior, plus added the hidden usage of addfinalizer after the examples.
What do you think?

[nicoddemus]

Thanks @aizpurua23a!

cc @fish-face