https://travis-ci.org/python/cpython/jobs/385458840

0:00:11 load avg: 22.29 [ 23/415/1] test_datetime crashed (Exit code -11)
Fatal Python error: Segmentation fault
Current thread 0x00002b52f7326400 (most recent call first):
  File "/home/travis/build/python/cpython/Lib/test/datetimetester.py", line 4630 in test_check_arg_types
  File "/home/travis/build/python/cpython/Lib/unittest/case.py", line 615 in run
  File "/home/travis/build/python/cpython/Lib/unittest/case.py", line 663 in __call__
  File "/home/travis/build/python/cpython/Lib/unittest/suite.py", line 122 in run
  File "/home/travis/build/python/cpython/Lib/unittest/suite.py", line 84 in __call__
  File "/home/travis/build/python/cpython/Lib/unittest/suite.py", line 122 in run
  File "/home/travis/build/python/cpython/Lib/unittest/suite.py", line 84 in __call__
  File "/home/travis/build/python/cpython/Lib/unittest/runner.py", line 176 in run
  (...)

Lib/test/datetimetester.py:4630:

    def test_check_arg_types(self):
        class Number:
            def __init__(self, value):
                self.value = value
            def __int__(self):
                return self.value

        for xx in [decimal.Decimal(10),    <~~~ HERE
                   decimal.Decimal('10.9'),
                   Number(10)]:
            self.assertEqual(datetime(10, 10, 10, 10, 10, 10, 10),
                             datetime(xx, xx, xx, xx, xx, xx, xx))


Hum, this bug looks like bpo-33627.

Is it a bug in the _decimal module?

On Wed, May 30, 2018 at 10:34:01PM +0000, STINNER Victor wrote:
> Fatal Python error: Segmentation fault
> Current thread 0x00002b52f7326400 (most recent call first):
> 
> Is it a bug in the _decimal module?

These kinds of errors are often completely random in the test suite:

https://bugs.python.org/issue25276
https://bugs.python.org/issue25276#msg252113

One with test_decimal, one with test_email.

How would #33627 support the theory that this is a bug in _decimal?  As I read it, #33627 is in test_complex, which supports the "unrelated random" theory.

> How would #33627 support the theory that this is a bug in _decimal?  As I read it, #33627 is in test_complex, which supports the "unrelated random" theory.

I'm not sure of anything. It's just a bet. I asked you to have a look, just in case you see something "obvious". My theory for bpo-33627:
https://bugs.python.org/issue33627#msg318304

None of these bugs make any sense :-)

> https://travis-ci.org/python/cpython/jobs/385458840

pythoninfo:

Py_DEBUG: Yes (sys.gettotalrefcount() present)
_decimal.__libmpdec_version__: 2.4.2
builtins.float.double_format: IEEE, little-endian
builtins.float.float_format: IEEE, little-endian
os.cpu_count: 48
os.environ[CC]: clang -pthread
os.uname: posix.uname_result(sysname='Linux', nodename='travis-job-python-cpython-385458840.travisci.net', release='4.4.0-112-generic', version='#135-Ubuntu SMP Fri Jan 19 11:48:36 UTC 2018', machine='x86_64')
platform.architecture: 64bit ELF
platform.platform: Linux-4.4.0-112-generic-x86_64-with-debian-jessie-sid
sys.version: 3.7.0b4+ (heads/3.7:4f53e2a, May 30 2018, 00:26:26)  [Clang 5.0.0 (tags/RELEASE_500/final)]

sysconfig[CCSHARED]: -fPIC
sysconfig[CC]: clang -pthread
sysconfig[CFLAGS]: -Wno-unused-result -Wsign-compare -g -O0 -Wall -Wstrict-prototypes -I/home/travis/multissl/openssl/1.1.0h/include -O3 -I/home/travis/multissl/openssl/1.1.0h/include -O3
sysconfig[CONFIG_ARGS]: '--with-pydebug' 'CC=clang' 'CFLAGS=-I/home/travis/multissl/openssl/1.1.0h/include -O3' 'LDFLAGS=-L/home/travis/multissl/openssl/1.1.0h/lib'
sysconfig[HOST_GNU_TYPE]: x86_64-pc-linux-gnu
sysconfig[OPT]: -g -O0 -Wall -Wstrict-prototypes
sysconfig[PY_CFLAGS]: -Wno-unused-result -Wsign-compare -g -O0 -Wall -Wstrict-prototypes -I/home/travis/multissl/openssl/1.1.0h/include -O3 -I/home/travis/multissl/openssl/1.1.0h/include -O3
sysconfig[PY_CFLAGS_NODIST]: -std=c99 -Wextra -Wno-unused-result -Wno-unused-parameter -Wno-missing-field-initializers -Werror=implicit-function-declaration
sysconfig[PY_LDFLAGS]: -L/home/travis/multissl/openssl/1.1.0h/lib -L/home/travis/multissl/openssl/1.1.0h/lib
sysconfig[Py_DEBUG]: 1
sysconfig[Py_ENABLE_SHARED]: 0

> https://travis-ci.org/python/cpython/jobs/385458840

$ clang --version
clang version 5.0.0 (tags/RELEASE_500/final)
Target: x86_64-unknown-linux-gnu
Thread model: posix
InstalledDir: /usr/local/clang-5.0.0/bin

./configure --with-pydebug

test_datetime passes here with nearly identical conditions as on the CI (Ubuntu 14.04, clang, same random seed):

Using random seed 987845
Run tests in parallel using 4 child processes
0:00:01 load avg: 1.70 [  1/415] test_html passed
0:00:01 load avg: 1.88 [  2/415] test_unicode_file_functions passed
0:00:02 load avg: 1.88 [  3/415] test_rlcompleter passed
0:00:03 load avg: 1.88 [  4/415] test_xml_etree passed
0:00:03 load avg: 1.88 [  5/415] test_codecencodings_kr passed
0:00:03 load avg: 1.88 [  6/415] test_dummy_thread passed
0:00:04 load avg: 1.88 [  7/415] test_source_encoding passed
0:00:05 load avg: 1.88 [  8/415] test_code passed
0:00:05 load avg: 1.88 [  9/415] test_curses passed
0:00:06 load avg: 1.88 [ 10/415] test_sys_setprofile passed
0:00:07 load avg: 2.05 [ 11/415] test_getopt passed
0:00:08 load avg: 2.05 [ 12/415] test_atexit passed
0:00:09 load avg: 2.05 [ 13/415] test_shlex passed
0:00:13 load avg: 2.13 [ 14/415] test_epoll passed
0:00:14 load avg: 2.13 [ 15/415] test_fileinput passed
0:00:26 load avg: 2.42 [ 16/415] test_embed passed
0:00:30 load avg: 2.54 [ 17/415] test_resource passed
0:00:31 load avg: 2.54 [ 18/415] test_augassign passed
0:00:36 load avg: 2.77 [ 19/415] test_gc passed -- running: test_io (33 sec)
0:00:38 load avg: 2.77 [ 20/415] test_multibytecodec passed -- running: test_io (35 sec)
0:00:38 load avg: 2.77 [ 21/415] test_contextlib passed -- running: test_io (35 sec)
0:00:39 load avg: 2.77 [ 22/415] test_code_module passed -- running: test_io (36 sec)
0:00:50 load avg: 2.80 [ 23/415] test_select passed -- running: test_unicodedata (36 sec), test_io (47 sec)
0:00:51 load avg: 2.90 [ 24/415] test_posixpath passed -- running: test_unicodedata (37 sec), test_io (48 sec)
0:01:05 load avg: 3.07 [ 25/415] test_capi passed -- running: test_unicodedata (50 sec), test_io (62 sec)
0:01:06 load avg: 3.07 [ 26/415] test_string passed -- running: test_unicodedata (52 sec), test_io (63 sec)
0:01:08 load avg: 3.06 [ 27/415] test_unicodedata passed (52 sec) -- running: test_io (65 sec)
0:01:09 load avg: 3.06 [ 28/415] test_iter passed -- running: test_io (66 sec)
0:01:09 load avg: 3.06 [ 29/415] test_largefile passed -- running: test_io (66 sec)
0:01:10 load avg: 3.06 [ 30/415] test_xxtestfuzz passed -- running: test_io (67 sec), test_datetime (30 sec)
0:01:10 load avg: 3.06 [ 31/415] test_pkg passed -- running: test_io (67 sec), test_datetime (31 sec)
0:01:11 load avg: 3.06 [ 32/415] test_datetime passed (30 sec) -- running: test_io (67 sec)

> test_datetime passes here with nearly identical conditions as on the CI (Ubuntu 14.04, clang, same random seed):

I tried to reproduce the issue on a Ubuntu Trusty *VM* using clang 5.0, using the same random seed: I failed to reproduce the bug.

I ran test_numeric_tower with GCC USBAN, but there is no warning (I'm not sure that I enabled correctly USBAN!).

I tested test_numeric_tower in Valgrind: no obvious memory error.

A wild theory: Because of the new pervasive includes (-I/home/travis/multissl/openssl/1.1.0h/include -O3 -I/home/travis/multissl/openssl/1.1.0h/include) some module picks up a wrong header.


But I just rebuilt https://travis-ci.org/python/cpython/jobs/385458840 and a different random seed is fine.


If the includes are responsible, I'd expect the crash to be independent of the random seed.

I didn't see the bug since at least one month. I close the issue.