Description

Eclips4

opened on Feb 29, 2024

Issue body actions

Bug report

Bug description:

./python.exe -m test -R 3:3 test_frame
Using random seed: 617081239
0:00:00 load avg: 38.69 Run 1 test sequentially
0:00:00 load avg: 38.69 [1/1] test_frame
beginning 6 repetitions. Showing number of leaks (. for 0 or less, X for 10 or more)
123:456
Xtest test_frame failed -- Traceback (most recent call last):
  File "/Users/admin/Projects/cpython/Lib/test/test_frame.py", line 353, in test_sneaky_frame_object
    self.assertIs(g.gi_frame, sneaky_frame_object)
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: <frame at 0x1063d5020, file '/Users/admin/Projects/cpython/Lib/test/test_frame.py', line 318, code f> is not <frame at 0x10637b9b0, file '/Users/admin/Projects/cpython/Lib/test/test_frame.py', line 353, code test_sneaky_frame_object>

test_frame failed (1 failure)

== Tests result: FAILURE ==

1 test failed:
    test_frame

Total duration: 146 ms
Total tests: run=21 failures=1
Total test files: run=1/1 failed=1
Result: FAILURE

CPython versions tested on:

CPython main branch

Operating systems tested on:

macOS

Linked PRs

• gh-116098: Revert "gh-107674: Improve performance of sys.settrace (GH-114986)" #116178
• gh-116098: Clean up frame allocation code and remove the invalid sneaky frame test #116687

Activity

1. Eclips4

   added

   type-bugAn unexpected behavior, bug, or errorAn unexpected behavior, bug, or error

   testsTests in the Lib/test dirTests in the Lib/test dir

   3.13only security fixesonly security fixes

   on Feb 29, 2024
2. Eclips4 commented on Feb 29, 2024

   

   Eclips4

   on Feb 29, 2024

   MemberAuthor

   More actions

   Bisected to 0a61e23
   cc @gaogaotiantian
3. gaogaotiantian commented on Feb 29, 2024

   

   gaogaotiantian

   on Feb 29, 2024

   Member

   More actions

   Okay it took me a while, but I don't believe this is a "bug" the commit brings - more like a very deligate(fragile) test is broken because it depends on some undocumented(undefined) behavior.

   I believe @brandtbucher is the author for the test. If I understand it correctly, the test requires a gc triggers in trace function - which is not guaranteed. Normally RESUME and INSTRUMENTED_RESUME triggers the gc, but RESUME_CHECK does not. I think it's intentional @markshannon ? The commit mentioned (correctly?) converts the RESUME instruction for trace to RESUME_CHECK which is not able to trigger gc anymore, thus fails the test.

   So we need to confirm a couple of things before fixing this:

   1. Should we fix the test, or the code?
   2. Is it reasonable for the test to require a gc in trace?
   3. Is it intentional that RESUME_CHECK does not trigger gc? Is that okay?
   4. Is it valid to have RESUME_CHECK for trace instead of RESUME or INSTRUMENTED_RESUME?
4. gaogaotiantian commented on Feb 29, 2024

   

   gaogaotiantian

   on Feb 29, 2024

   Member

   More actions

   If we need further discussion for this and can't come up with a fix today, I'll revert this commit by the end of today and wait for the final decision.
5. encukou

   added a commit that references this issue on Mar 1, 2024

   gh-116098: Revert "gh-107674: Improve performance of `sys.settrace` (G…

   7895a61
6. encukou commented on Mar 1, 2024

   

   encukou

   on Mar 1, 2024

   Member

   More actions

   Thank you for the revert.
   Sorry that it's your commit that got caught here. Reverting is a way to get the buildbots green so they can catch more issues, not a judgement of your work.

   Reacted by Carl Meyer
7. gaogaotiantian commented on Mar 1, 2024

   

   gaogaotiantian

   on Mar 1, 2024

   Member

   More actions

   I understand - it's important to keep the test clean otherwise the test lost its meaning. I'll work on the fix and hopefully we can come up with something and check it back in.
8. woodruffw

   added a commit that references this issue on Mar 4, 2024

   pythongh-116098: Revert "pythongh-107674: Improve performance of `sys…

   7c194b9
9. sthagen

   added a commit that references this issue on Mar 12, 2024

   pythonGH-116098: Remove dead frame object creation code (pythonGH-116687

   a53cc3f
10. vstinner

    added a commit that references this issue on Mar 20, 2024

    pythonGH-116098: Remove dead frame object creation code (pythonGH-116687

    ccd58bf
11. adorilson

    added 2 commits that reference this issue on Mar 25, 2024

    pythongh-116098: Revert "pythongh-107674: Improve performance of `sys…

    aabeaa5

    pythonGH-116098: Remove dead frame object creation code (pythonGH-116687

    5267e90
12. diegorusso

    added 2 commits that reference this issue on Apr 17, 2024

    pythongh-116098: Revert "pythongh-107674: Improve performance of `sys…

    25aa06c

    pythonGH-116098: Remove dead frame object creation code (pythonGH-116687

    6be31cb
13. Eclips4 commented on May 8, 2024

    

    Eclips4

    on May 8, 2024

    MemberAuthor

    More actions

    @gaogaotiantian can it be closed now?
14. gaogaotiantian commented on May 8, 2024

    

    gaogaotiantian

    on May 8, 2024

    Member

    More actions

    Yes. The root cause was that the test is not valid anymore and we removed the test.

    Reacted by Kirill Podoprigora
15. Eclips4

    closed this as completedon May 8, 2024

Issue actions