Maybe if would be safer to run tests on buildbots to check more SQLite versions, before merging this PR.

Oh, it may also help if someone could test manually with SQLite 3.32.

Oh, it may also help if someone could test manually with SQLite 3.32.

Already did that, and it passes! 3.32.1 FWIW

Already did that, and it passes! 3.32.1 FWIW

Cool. Thanks!

This still fails. I've fixed CheckFuncNonDeterministic locally but not pushed it yet.

Yeah, the Ubuntu job of Azure Pipelines fails. It uses "sqlite3.sqlite_version: 3.11.0" according to its test.pythoninfo. Test failure:

FAIL: CheckFuncNonDeterministic (sqlite3.test.userfunctions.FunctionTests)
Non-deterministic functions cannot be used with partial indices
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/vsts/work/1/s/Lib/sqlite3/test/userfunctions.py", line 288, in CheckFuncNonDeterministic
    self.assertIn("non-deterministic functions prohibited", str(cm.exception))
AssertionError: 'non-deterministic functions prohibited' not found in 'functions prohibited in partial index WHERE clauses'

Honestly, I don't think that we should test the error message. Please simply remove the test the error message. IMHO excepting an exception is enough.

Yeah, the Ubuntu job of Azure Pipelines fails. It uses "sqlite3.sqlite_version: 3.11.0" according to its test.pythoninfo.

Right. Deterministic functions were allowed in partial indices in sqlite v3.15.0 (2016-10-14). I'll find another way of testing this, then.

Honestly, I don't think that we should test the error message. Please simply remove the test the error message. IMHO excepting an exception is enough.

Agreed. Error messages are likely to change between versions.

@vstinner I suggest just keeping the old tests for SQLite pre v3.15.0. From the sqlite release history:

2014-02-03 (3.8.3) […] Added SQLITE_DETERMINISTIC […] providing applications with the ability to create new functions that can be factored out of inner loops when they have constant arguments.

Seven minor versions later:

2016-10-14 (3.15.0) […] Allow deterministic SQL functions in the WHERE clause of a partial index.

I'll push the change right away.

Thank you!
I'll start the buildbots so Victor can continue the review when he gets to the computer.

🤖 New build scheduled with the buildbot fleet by @encukou for commit b969ae5 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

Thank you, @encukou!

Some failures might not be related to this change.

AMD64 Fedora Stable Clang Installed PR, PPC64LE Fedora Stable Clang Installed PR and s390x Fedora Clang Installed PR and x86 Gentoo Installed with X PR failed because of https://bugs.python.org/issue38488 regression (setuptools/chmod) which is unrelated.

AMD64 FreeBSD Non-Debug PR failures (test_asyncio, test_venv) are unrelated.

AMD64 Windows8.1 Refleaks PR failure (test_socket) is unrelated.

(this comment is unrelated to test_sqlite)

AMD64 Fedora Stable Clang Installed PR, PPC64LE Fedora Stable Clang Installed PR and s390x Fedora Clang Installed PR and x86 Gentoo Installed with X PR failed because of https://bugs.python.org/issue38488 regression (setuptools/chmod) which is unrelated.

I reverted the change which introduced this issue: see https://bugs.python.org/issue38488

AMD64 FreeBSD Non-Debug PR failures (test_asyncio, test_venv) are unrelated.

test_asyncio: test_sock_client_racing() failed, see https://bugs.python.org/issue30064. I created #20485 (merged) to skip the test.

The test_venv is also related to https://bugs.python.org/issue38488 that I reverted: I created https://bugs.python.org/issue40808

AMD64 Windows8.1 Refleaks PR failure (test_socket) is unrelated.

0:41:43 load avg: 6.17 [206/424/1] test_socket failed (41.9 sec) -- running: test_multiprocessing_spawn (41 min 20 sec), test_regrtest (5 min 37 sec)
beginning 6 repetitions
123456
Warning -- threading._dangling was modified by test_socket
  Before: {<weakref at 0x000000B0E9AE4530; to '_MainThread' at 0x000000B0E9935E10>}
  After:  {<weakref at 0x000000B0EC3A1D10; to '_MainThread' at 0x000000B0E9935E10>, <weakref at 0x000000B0EC3A1D70; to 'Thread' at 0x000000B0EBEDD820>, <weakref at 0x000000B0EC3A1DD0; to 'Thread' at 0x000000B0EBEDD690>, <weakref at 0x000000B0EC3A1E30; to 'Thread' at 0x000000B0EBEDD730>, <weakref at 0x000000B0EC3A1E90; to 'Thread' at 0x000000B0EBEDD780>} 
test test_socket failed -- Traceback (most recent call last):
  File "D:\buildarea\pull_request.ware-win81-release.refleak\build\lib\test\test_socket.py", line 369, in _tearDown
    self.wait_threads.__exit__(None, None, None)
  File "D:\buildarea\pull_request.ware-win81-release.refleak\build\lib\contextlib.py", line 124, in __exit__
    next(self.gen)
  File "D:\buildarea\pull_request.ware-win81-release.refleak\build\lib\test\support\threading_helper.py", line 101, in wait_threads_exit
    gc_collect()
NameError: name 'gc_collect' is not defined

This one looks like a combo:

• test_socket leaks a thread
• test.support.threading_helper.wait_threads_exit() has a regression: see bpo-40275: Fix a calling bug in threading_helper #20479

Thanks for the review, @berkerpeksag & @vstinner !

Thanks for the PR!

Thanks @erlend-aasland for the PR, and @berkerpeksag for merging it 🌮🎉.. I'm working now to backport this PR to: 3.9.
🐍🍒⛏🤖

Thanks @erlend-aasland for the PR, and @berkerpeksag for merging it 🌮🎉.. I'm working now to backport this PR to: 3.8.
🐍🍒⛏🤖

GH-20512 is a backport of this pull request to the 3.8 branch.

GH-20513 is a backport of this pull request to the 3.9 branch.

Thanks for fixing the encoding/copyright line ;-)