Pytest isn't killed when stopping test sessions

[almetge]

Type: Bug

When using the "Cancel test run" button to stop the current test, vscode immediately behave as if the test has been canceled, but the underlying process keep running. This can lead in resource contention due to old process keeping lock over resource files.
To compare, running pytest in a command interpreter and sending CTRL-C do terminate the process properly, with pytest logging the test failure and the reason (KeyboardInterrupt).

The issue was previously described in this closed ticket: #7055
I followed the instructions in this comment: #7055 (comment)
But the issue persists.

Steps to reproduce:

Create a file test_bug.py with this content:

def test_loop_forever():
    import logging
    import os
    import time
    logger = logging.getLogger("infinite_loop")
    logger.error("My PID is: %d", os.getpid())
    while True:
        time.sleep(10)
        logger.error("Still running... PID: %d", os.getpid())

Also create a pytest.ini file with this content, to log data to a file:

[pytest]
log_format = %(asctime)s %(name)s %(levelname)s %(message)s
log_date_format = %Y-%m-%d %H:%M:%S
log_file = pytest.log
log_level = DEBUG

Now, in the "Testing" view, configure the tests with pytest, and run the created test.
A file pytest.log should be created in the current folder, open it, you should see the PID of the process running the test, as well as a repetition of it every 10 seconds.
In VSCode, in the "Testing" view, click the "Cancel test run" button. VSCode says that the test has been canceled, and no longer show the spinning progress indicator next to the test (as expected).
But, if you look either in Windows task manager, in the "Details" tab, the process with the previously printed PID still exists, and in the log file pytest.log you can see new lines being added.

Diagnostic data

Output for Python in the Output panel (View→Output, change the drop-down the upper-right of the Output panel to Python)

2025-07-18 11:57:55.823 [info] arg: --rootdir already exists in args, not adding.
2025-07-18 11:57:55.823 [info] Attempting to use temp directory for test ids file, file name: test-ids-58d69856a7e6d6ec1cff.txt
2025-07-18 11:57:55.846 [info] Environment variables set for pytest execution: PYTHONPATH=C:\Users\USER\scoop\persist\vscode\data\extensions\ms-python.python-2025.10.0-win32-x64\python_files, TEST_RUN_PIPE=\\.\pipe\python-test-results-df28fc97cdc7c1fc4d38, RUN_TEST_IDS_PIPE=C:\Users\USER\AppData\Local\Temp\test-ids-58d69856a7e6d6ec1cff.txt
2025-07-18 11:57:55.846 [info] Running pytest with arguments: C:\Users\USER\scoop\persist\vscode\data\extensions\ms-python.python-2025.10.0-win32-x64\python_files\vscode_pytest\run_pytest_script.py --rootdir=c:\TEMP\aze for workspace c:\TEMP\aze 

2025-07-18 11:57:55.846 [info] > C:\TEMP\aze\.venv\Scripts\python.exe ~\scoop\persist\vscode\data\extensions\ms-python.python-2025.10.0-win32-x64\python_files\vscode_pytest\run_pytest_script.py --rootdir=.
2025-07-18 11:57:55.846 [info] cwd: .
2025-07-18 11:58:02.292 [info] Test Result named pipe \\.\pipe\python-test-results-df28fc97cdc7c1fc4d38  cancelled
2025-07-18 11:58:02.292 [info] Test run cancelled, resolving 'TillServerClose' deferred for c:\TEMP\aze.
2025-07-18 11:58:02.292 [info] Test run cancelled, killing pytest subprocess for workspace c:\TEMP\aze

Extension version: 2025.10.0
VS Code version: Code 1.101.2 (2901c5ac6db8a986a5666c3af51ff804d05af0d4, 2025-06-24T20:27:15.391Z)
OS version: Windows_NT x64 10.0.22631
Modes:

• Python version (& distribution if applicable, e.g. Anaconda): 3.13.2
• Type of virtual environment used (e.g. conda, venv, virtualenv, etc.): Venv
• Value of the python.languageServer setting: Default

User Settings

languageServer: "Pylance"

testing
• pytestArgs: "<placeholder>"
• pytestEnabled: true

experiments
• optInto: ["pythonTestAdapter"]

Installed Extensions

Extension Name	Extension Id	Version
black-formatter	ms-	2025.2.0
cmake-tools	ms-	1.21.36
copilot	Git	1.344.0
copilot-chat	Git	0.28.5
cpptools	ms-	1.26.3
cpptools-extension-pack	ms-	1.3.1
cpptools-themes	ms-	2.0.0
debugpy	ms-	2025.10.0
even-better-toml	tam	0.21.2
hexeditor	ms-	1.11.1
isort	ms-	2025.0.0
js-debug	ms-	1.100.1
js-debug-companion	ms-	1.1.3
powershell	ms-	2025.2.0
pylint	ms-	2025.2.0
python	ms-	2025.10.0
tcl	ras	0.1.0
vscode-github-actions	git	0.27.2
vscode-js-profile-table	ms-	1.0.10
vscode-openapi	42C	4.37.2
vscode-pylance	ms-	2025.6.2
vscode-st	Ser	1.12.36
vscode-xml	red	0.29.0
vscode-yaml	red	1.18.0

A/B Experiments

vsliv368:30146709
vswsl492:30256859
binariesv615:30325510
962ge761:31349326
2e7ec940:31000449
nativeloc1:31344060
5fd0e150:31155592
dwcopilot:31170013
471b6256:31263136
6074i472:31201624
dwoutputs:31242946
9064b325:31222308
copilot_t_ci:31333650
e5gg6876:31282496
pythoneinst12:31285622
c7cif404:31314491
pythonpulldiag:31343502
996jf627:31283433
pythonrdcb7:31342333
usemplatestapi:31297334
0aa6g176:31307128
747dc170:31275177
aj953862:31281341
generatesymbolt:31295002
convertfstringf:31295003
pylancequickfixf:31350060
j6a0c355:31346307
usemarketplace:31343026
nesew2to5:31336538
agentclaude:31335815
4f60g487:31327383
nes-diff-11:31337487
992g2967-color:31348502
6abeh943:31336334
yijiwantestdri0626-c:31336931
64bie973:31349831
ji9b5146:31348712
nes-deb-050:31346716

[almetge]

Manually killing the process after cancelling the test run will generate the following lines in the Python output window:

2025-07-18 14:25:35.437 [error] Subprocess exited unsuccessfully with exit code 1 and signal null on workspace c:\TEMP\aze
2025-07-18 14:25:35.437 [error] Subprocess closed unsuccessfully with exit code 1 and signal null for workspace c:\TEMP\aze. Creating and sending error execution payload 

So it seems VSCode is still aware of those processes.

[eleanorjboyd]

will investigate thank you

[PierreLeger8]

I determine the same issue with not properly closed COM ports when I cancel a test run. With the result that I get a premission denied error during the next test run. The only solution is to restart VS Code.

[eleanorjboyd]

@karthiknadig as a windows expert could you help add some background here on what you think the fix might be?

[karthiknadig]

We have to look at what signal is being sent on cancel. I think this might be a case where we are not sending terminal after giving it a reasonable amount of time to shutdown.

[github-actions]

Because we have not heard back with the information we requested, we are closing this issue for now. If you are able to provide the info later on, then we will be happy to re-open this issue to pick up where we left off.

Happy Coding!