GH-94438: Fix RuntimeWarning for jump tests in test_sys_settrace

[gaogaotiantian]

A couple of new tests were introduced to test_sys_settrace in #111237, however, two of them triggered a RuntimeWarning:

./python -m test test_sys_settrace -m test_jump_is_not_none_forwards
Using random seed: 53646886
0:00:00 load avg: 0.44 Run 1 test sequentially
0:00:00 load avg: 0.44 [1/1] test_sys_settrace
/home/gaogaotiantian/programs/mycpython/Lib/test/test_sys_settrace.py:1947: RuntimeWarning: assigning None to 1 unbound local
  frame.f_lineno = self.firstLine + self.jumpTo

== Tests result: SUCCESS ==

This is because the jump happens before all the local variables are assigned:

    @jump_test(1, 4, [5])
    def test_jump_is_not_none_forwards(output):
        x = None
        if x is not None:
            output.append(3)
        else:
            output.append(5)

The test jumps on line 1 - before x is assigned to None. So when frame.f_lineno is being set, a warning is generated to alert the users that the unbound local variables are being set to None.

This does not affect the validity or the result of the test, but I don't think it's intended, and it is not the best practice. Especially considering that the decorator can already take care of such warnings.

In this PR, two different fixes were applied to the problematic tests to maximize the diversity:

• Add the warning to expected result
• Start from line 2 so x is assigned.

A warning checker is also introduced so this won't happen in the future - if a test is not expecting a warning, then a warning is treated as an error. This might not be the best practice for all tests globally, but in this specific case, we do have "expected warnings" and it should count.

• Issue: frame.setlineno has serious flaws. #94438

[gaogaotiantian]

cc: @savannahostrowski

[savannahostrowski]

My intention was to have both of these jump after x was defined. I wonder if we should just update to both jump on line 2 and then have a separate test case for warnings?

[gaogaotiantian]

My intention was to have both of these jump after x was defined. I wonder if we should just update to both jump on line 2 and then have a separate test case for warnings?

We can do that, but the test case with warning tests what you want as well. Your test case does not rely on the variable x - it just needs some specific bytecode in the function (if I understand your test case correctly). Jumping from line 1 will just cover all the stuff you want to test, and it gives a little bit extra coverage for jumping before a variable is initialized. You wrote the test so I will definitely change it to jumping from line 2 if you want - just share my thoughts on this.

[savannahostrowski]

That makes sense - thanks for catching this. Good learning for me too!

[brandtbucher]

Looks good, thanks! This should be backported, yeah?

[gaogaotiantian]

Looks good, thanks! This should be backported, yeah?

Yes as the original test was backported.

[miss-islington-app]

Thanks @gaogaotiantian for the PR, and @brandtbucher for merging it 🌮🎉.. I'm working now to backport this PR to: 3.12.
🐍🍒⛏🤖

[bedevere-app]

GH-111369 is a backport of this pull request to the 3.12 branch.

[gaogaotiantian]

Just to mention that the original test was backported to 3.11 as well.

[brandtbucher]

If I recall correctly, we only introduced these warnings in 3.12 (I think it coincided with the LOAD_FAST/LOAD_FAST_CHECK split). So no need to backport further than that?

[gaogaotiantian]

Yeah just checked the blame, the warning was added in 3.12.