gh-136438: Make sure test_generated_cases pass with all optimization levels

[efimov-mikhail]

Now tests pass with all combinations of -OO and --without-doc-strings.

Before:

======================================================================
FAIL: test_missing_override_failure (test.test_generated_cases.TestGeneratedAbstractCases.test_missing_override_failure)
----------------------------------------------------------------------
AssertionError: 'case OP: {\n JitOptRef out;\n [104 chars] }' != ''
- case OP: {
- JitOptRef out;
- out = sym_new_not_null(ctx);
- stack_pointer[-1] = out;
- break;
- }
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/home/sikko/projects/cpython/Lib/test/test_generated_cases.py", line 2040, in test_missing_override_failure
with self.assertRaisesRegex(AssertionError, "All abstract uops"):
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: "All abstract uops" does not match "'case OP: {\n JitOptRef out;\n [104 chars] }' != ''
- case OP: {
- JitOptRef out;
- out = sym_new_not_null(ctx);
- stack_pointer[-1] = out;
- break;
- }
"

• Issue: Multiple tests fail with -OO #136438

[efimov-mikhail]

Cc @Eclips4

[Eclips4]

Though I'm not sure, from an architectural point of view, that raising AssertionError manually is appropriate. I think we should change it to ValueError and handle it in the tests. WDYM?

[efimov-mikhail]

I'm agree that raising ValueError would be better. My first attempt was just to make as few changes to code as possible.

[Eclips4]

I guess it's ok to change it in this PR. Catching an AssertionError is semantically incorrect :)

[Eclips4]

Thank you Mikhail.