Repository navigation
test_largefile hangs on x86-64 MacOS Intel ASAN NoGIL buildbot #159608
Description
Since 2 days, test_largefile started to hang randomy on x86-64 MacOS Intel ASAN NoGIL buildbot. It's strange that test_largefile and test_faulthandler #159606 started to fail for 2 days on this specific ASAN buildbot.
cc @itamaro
x86-64 MacOS Intel ASAN NoGIL 3.14: https://buildbot.python.org/#/builders/1729/builds/1754
1:21:16 load avg: 6.39 [2/2/2] test_largefile worker non-zero exit code (Exit code 1) Re-running test_largefile in verbose mode (...) test_seekable (test.test_largefile.PyLargeFileTest.test_seekable) ... ok test_truncate (test.test_largefile.PyLargeFileTest.test_truncate) ... ok test_it (test.test_largefile.TestCopyfile.test_it) ... ok test_it (test.test_largefile.TestSocketSendfile.test_it) ... Timeout (0:35:00)! Thread 0x000070000f10c000 [Thread-1 (run)] (most recent call first): File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/test/test_largefile.py", line 233 in run File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/threading.py", line 1024 in run File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/threading.py", line 1082 in _bootstrap_inner File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/threading.py", line 1044 in _bootstrap Thread 0x00007ff84b416a00 (most recent call first): File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/socket.py", line 396 in _sendfile_use_sendfile File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/socket.py", line 493 in sendfile File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/test/test_largefile.py", line 251 in test_it File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/test/test_largefile.py", line 183 in wrapper File "/Users/buildbot/buildarea/3.14.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/unittest/case.py", line 615 in _callTestMethod (...)
3.14 started to fail recently:
- latest green build: https://buildbot.python.org/#/builders/1729/builds/1752
- first test_largefile failure: https://buildbot.python.org/#/builders/1729/builds/1754 (latest build)
3.15:
main:
- fail: https://buildbot.python.org/#/builders/1366/builds/8253
- fail: https://buildbot.python.org/#/builders/1366/builds/8255
- fail: https://buildbot.python.org/#/builders/1366/builds/8258
- fail: https://buildbot.python.org/#/builders/1366/builds/8260
The 3.15 traceback is different:
(...) test_seekable (test.test_io.test_largefile.PyLargeFileTest.test_seekable) ... ok test_truncate (test.test_io.test_largefile.PyLargeFileTest.test_truncate) ... ok test_it (test.test_io.test_largefile.TestCopyfile.test_it) ... ok Timeout (0:35:00)! Thread 0x00007ff84b416a00 (most recent call first): File "/Users/buildbot/buildarea/3.15.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/test/support/os_helper.py", line 374 in unlink File "/Users/buildbot/buildarea/3.15.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/test/test_io/test_largefile.py", line 51 in tearDownClass File "/Users/buildbot/buildarea/3.15.itamaro-macos-intel-aws.macos-with-brew.asan.nogil/build/Lib/unittest/suite.py", line 318 in _tearDownPreviousClass (...)
Activity
vstinner commented on Oct 17, 2026
@itamaro: Something is wrong on this macOS machine.
See for example this recent build: https://buildbot.python.org/#/builders/1729/builds/1754
The "clean" step hangs on find build -type f -a ! -name '*.gc??' -exec rm -f {} ';'.
The "git gc" had to be interrupted after 20 minutes (whereas the command should take less than 1 minute).
The machine has an uptime of 400 days. Maybe it would be interesting to just reboot it?
itamaro commented on Oct 17, 2026
Thanks for pinging me, I'll reboot it on Sunday
itamaro commented on Oct 18, 2026
the machine has been recycled and reprovisioned.
the few runs since it restarted look good, so I'll close this issue (and the other one) for now. if the issues come back, we can reopen the issue.