test_listcomps fails when run with -We

[AlexWaygood]

Bug report

Bug description:

python -m test test_listcomps causes SyntaxWarnings to be emitted on main, and fails when run with -We:

Without -We:

0:00:00 Run tests sequentially
0:00:00 [1/1] test_listcomps
<string>:4: SyntaxWarning: 'set' object is not callable; perhaps you missed a comma?
<string>:6: SyntaxWarning: 'set' object is not callable; perhaps you missed a comma?
<string>:6: SyntaxWarning: 'set' object is not callable; perhaps you missed a comma?

== Tests result: SUCCESS ==

1 test OK.

Total duration: 230 ms
Total tests: run=49
Total test files: run=1/1
Result: SUCCESS

With -We:

======================================================================
ERROR: test_comp_in_try_finally (test.test_listcomps.ListComprehensionTest.test_comp_in_try_finally) (scope='module')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\Users\alexw\coding\cpython\Lib\test\test_listcomps.py", line 122, in _check_in_scopes
    exec(newcode, newns)
  File "<string>", line 4
SyntaxError: 'set' object is not callable; perhaps you missed a comma?

======================================================================
ERROR: test_comp_in_try_finally (test.test_listcomps.ListComprehensionTest.test_comp_in_try_finally) (scope='class')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\Users\alexw\coding\cpython\Lib\test\test_listcomps.py", line 122, in _check_in_scopes
    exec(newcode, newns)
  File "<string>", line 6
SyntaxError: 'set' object is not callable; perhaps you missed a comma?

======================================================================
ERROR: test_comp_in_try_finally (test.test_listcomps.ListComprehensionTest.test_comp_in_try_finally) (scope='function')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\Users\alexw\coding\cpython\Lib\test\test_listcomps.py", line 122, in _check_in_scopes
    exec(newcode, newns)
  File "<string>", line 6
SyntaxError: 'set' object is not callable; perhaps you missed a comma?

----------------------------------------------------------------------

The problematic test is test_listcomps.ListComprehensionTest.test_comp_in_try_finally, which was added in d52c448 last week

CPython versions tested on:

CPython main branch

Operating systems tested on:

Windows

Linked PRs

• gh-109182: Fix SyntaxWarning in test_listcomps.py #109183
• gh-109182: Fix and improve tests for gh-108654 #109189
• [3.12] gh-109182: Fix and improve tests for gh-108654 (GH-109189) #109271

[serhiy-storchaka]

It is not the only issue with these tests:

1. test_comp_in_try_except only tests that the variable is finally restored after leaving the scope. It does not test that it is restored in the error handler.
2. test_comp_in_try_except silences all exceptions by using a bare except. It can hide an error in the testing code.
3. As well it does not ensure that an exception is not raised in one of cases.
4. test_comp_in_try_finally emits a SyntaxWarning, as was originally reported.
5. test_comp_in_try_finally does not check that the variable is restored in the finally block.
6. finally block has two paths. test_comp_in_try_finally only tests one of them. The path without exception is not tested.
7. test_comp_in_try_finally only tests the code in the function scope (because it wraps it in a function). module and class scopes are not tested.

I do not know about test_exception_in_post_comp_call because I do not know why it was added at first place. Does it test some case not covered by existing tests? Is there something special with this code that can make the variable not restoring immediately after successful evaluation of the comprehension?

[carljm]

@serhiy-storchaka thanks for reviewing and improving the tests! Will comment on your PR.

test_exception_in_post_comp_call is a regression test for a bug that I created in an early version of #108659 resulting in stack underflow when unwinding an exception raised inside a call occurring with the comprehension result still on the stack. It was surfaced by another test in the CPython test suite, but I wanted this case included in the listcomps tests.