Repository navigation
test_frame fails when running with -R 3:3 argument #116098
Description
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: FAILURECPython versions tested on:
CPython main branch
Operating systems tested on:
macOS
Activity
Eclips4 commentedon Feb 29, 2024
Bisected to 0a61e23
cc @gaogaotiantian
gaogaotiantian commentedon Feb 29, 2024
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
tracefunction - which is not guaranteed. NormallyRESUMEandINSTRUMENTED_RESUMEtriggers the gc, butRESUME_CHECKdoes not. I think it's intentional @markshannon ? The commit mentioned (correctly?) converts theRESUMEinstruction fortracetoRESUME_CHECKwhich is not able to trigger gc anymore, thus fails the test.So we need to confirm a couple of things before fixing this:
trace?RESUME_CHECKdoes not trigger gc? Is that okay?RESUME_CHECKfortraceinstead ofRESUMEorINSTRUMENTED_RESUME?gaogaotiantian commentedon Feb 29, 2024
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.
gh-116098: Revert "gh-107674: Improve performance of `sys.settrace` (G…
encukou commentedon Mar 1, 2024
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.
gaogaotiantian commentedon Mar 1, 2024
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.
pythongh-116098: Revert "pythongh-107674: Improve performance of `sys…
pythonGH-116098: Remove dead frame object creation code (pythonGH-116687
pythonGH-116098: Remove dead frame object creation code (pythonGH-116687
pythongh-116098: Revert "pythongh-107674: Improve performance of `sys…
pythonGH-116098: Remove dead frame object creation code (pythonGH-116687
pythongh-116098: Revert "pythongh-107674: Improve performance of `sys…
pythonGH-116098: Remove dead frame object creation code (pythonGH-116687
Eclips4 commentedon May 8, 2024
@gaogaotiantian can it be closed now?
gaogaotiantian commentedon May 8, 2024
Yes. The root cause was that the test is not valid anymore and we removed the test.