python / python/cpython

TSAN reports multiple data races, SEGV errors, and thread leaks when running the full test suite

Open
#140,209 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

tests type-feature
Dominant language
Python
Stars
77.2k
Forks
35.9k
PR merge metrics
PR metrics pending

Description

Bug report

Bug description:

Description:

When running the full test suite (./python -m test) on a debug build of the main branch configured with ThreadSanitizer (TSAN) and the experimental JIT, the test run terminates with a FAILURE result.

The attached logs show thousands of warnings from TSAN, including:

  1. Massive numbers of data race warnings, primarily during the parallel initialization of sub-interpreters.
  2. Several DEADLYSIGNAL (SEGV) reports, causing processes to abort.
  3. Multiple thread leak warnings related to multiprocessing tests.
  4. Failures in several core tests, including test_datetime, test_faulthandler, test_pyrepl, and test.test_multiprocessing_forkserver.test_manager.
Build & Environment
  • CPython Version: 3.15.0a1+ (heads/main:7ac94fcb1d0, Oct 16 2025)
  • Platform: Linux-6.12.48+deb13-cloud-amd64-x86_64-with-glibc2.41
  • Compiler: Clang 19.1.7
  • Configuration command:
    CC=clang CXX=clang++ ./configure --with-thread-sanitizer --with-pydebug --enable-experimental-jit=yes
    
  • Build command:
    make -j$(nproc)
    
  • Test command:
    ./python -m test -j$(nproc)
    
Summary of Failures

The test suite concludes with the following result:

== Tests result: FAILURE ==

4 tests failed:
    test.test_multiprocessing_forkserver.test_manager test_datetime
    test_faulthandler test_pyrepl
Key Issues Reported by ThreadSanitizer

The full logs will be attached to this report, but here are the primary categories of errors observed.

1. Data Races during Sub-interpreter Initialization

Thousands of data race warnings are reported when multiple threads attempt to initialize global state concurrently. This is especially prominent in test_datetime, which fails with an exit code from TSAN.

  • Example Data Race Warning:
    ==================
    WARNING: ThreadSanitizer: data race (pid=1111680)
      Write of size 8 at 0x56174f42e490 by thread T2:
        #0 _PyExc_InitTypes /root/cpython/main/Objects/exceptions.c:4428:32
        #1 pycore_init_types /root/cpython/main/Python/pylifecycle.c:732:9
        ...
    
      Previous write of size 8 at 0x56174f42e490 by thread T1:
        #0 _PyExc_InitTypes /root/cpython/main/Objects/exceptions.c:4428:32
        #1 pycore_init_types /root/cpython/main/Python/pylifecycle.c:732:9
        ...
    
      Location is global '_PyExc_ReferenceError' of size 416 at 0x56174f42e300 (python+0xa57490)
    ...
    SUMMARY: ThreadSanitizer: data race /root/cpython/main/Objects/exceptions.c:4428:32 in _PyExc_InitTypes
    ==================
    
  • Resulting test_datetime Failure:
    test test_datetime failed -- Traceback (most recent call last):
    ...
    AssertionError: Process return code is 66
    command line: ['/root/cpython/main/python', '-X', 'faulthandler', '-I', '-c', "if True:\n from concurrent.futures import InterpreterPoolExecutor\n\n def func():\n import _datetime\n ..."]
    

2. SEGV Errors and Process Aborts

Multiple tests trigger fatal segmentation faults, which are caught and reported by TSAN.

  • Example SEGV Report:
    ThreadSanitizer:DEADLYSIGNAL
    ==1101487==ERROR: ThreadSanitizer: SEGV on unknown address 0x00000010ceaf (pc 0x7f05f669195c bp 0x000000000000 sp 0x7ffde6ad06a0 T1101487)
    ==1101487==The signal is caused by a READ memory access.
        #0 <null> <null> (libc.so.6+0x9495c)
        #1 raise <null> (libc.so.6+0x3fcc1)
        #2 raise <null> (python+0x103503)
        #3 faulthandler_raise_sigsegv /root/cpython/main/./Modules/faulthandler.c:1158:5
        ...
    SUMMARY: ThreadSanitizer: SEGV (/lib/x86_64-linux-gnu/libc.so.6+0x9495c) 
    ==1101487==ABORTING
    

3. Thread Leaks

TSAN reports that threads created during multiprocessing tests are not properly joined, leading to resource leaks.

  • Example Thread Leak Warning:
    ==================
    WARNING: ThreadSanitizer: thread leak (pid=1106511)
      Thread T2 'Thread-2 (handl' (tid=1106562, finished) created by thread T1 at:
        #0 pthread_create <null> (python+0xfaa35)
        #1 do_start_joinable_thread /root/cpython/main/Python/thread_pthread.h:281:14
        ...
    SUMMARY: ThreadSanitizer: thread leak (/root/cpython/main/python+0xfaa35) in pthread_create
    ==================
    
CPython versions tested on:

3.15, CPython main branch

Operating systems tested on:

Linux

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the report with the listed configure, build, and ./python -m test commands. Read the TSAN traces in Objects/exceptions.c, Python/pylifecycle.c, Modules/faulthandler.c, and Python/thread_pthread.h, then isolate the failures in test_datetime, test_faulthandler, test_pyrepl, and test.test_multiprocessing_forkserver.test_manager. Done requires a scoped cause and a verified resolution for the reported races, crashes, and leaks.

Written by the indexing model from the issue text.

Assessment

Tech stack
linux, python
Domain
operating-systems, testing-qa
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
15/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.