AMD64 Debian root 3.x:

http://buildbot.python.org/all/#/builders/27/builds/1067

======================================================================
FAIL: test_thread_time (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/root/buildarea/3.x.angelico-debian-amd64/build/Lib/test/test_time.py", line 536, in test_thread_time
    self.assertGreaterEqual(stop - start, 0.020)  # machine busy?
AssertionError: 0.019943113999999984 not greater than or equal to 0.02

The following test failed:

        # thread_time() should include CPU time spent in current thread...
        start = time.thread_time()
        busy_wait(0.100)
        stop = time.thread_time()
        self.assertGreaterEqual(stop - start, 0.020)  # machine busy?

def busy_wait(duration):
    deadline = time.monotonic() + duration
    while time.monotonic() < deadline:
        pass

Do we really have to have functional tests on Python block functions? These tests seem very fragile...

I proposed the PR 8265 to remove these fragile tests.

I chose to keep time.process_time() functional test. We can remove it later if it starts failing on a buildbot.

New changeset d6345def68d3a0227253da26213dadb247f786db by Victor Stinner in branch 'master':
bpo-33723: Fix test_time.test_thread_time() (GH-8267)
https://github.com/python/cpython/commit/d6345def68d3a0227253da26213dadb247f786db

New changeset 4ada0cff5315627eae8208adf0209b479f497b8d by Miss Islington (bot) in branch '3.7':
bpo-33723: Fix test_time.test_thread_time() (GH-8267)
https://github.com/python/cpython/commit/4ada0cff5315627eae8208adf0209b479f497b8d

I reopen the issue since I got a very similar failure of test_time.test_process_time() on my laptop: the busy loop took 15.9 ms
whereas the test expects at least 20 ms. I wrote PR 8358 to only require 15 ms instead.

New changeset e78dace8dcb23c371df19c9add65895adf436995 by Victor Stinner in branch 'master':
bpo-33723: Fix test_time.test_process_time() (GH-8358)
https://github.com/python/cpython/commit/e78dace8dcb23c371df19c9add65895adf436995

New changeset a7a6eac966dd52db1762c8f455c1e208df36feb4 by Victor Stinner (Miss Islington (bot)) in branch '3.7':
bpo-33723: Fix test_time.test_process_time() (GH-8358) (GH-8362)
https://github.com/python/cpython/commit/a7a6eac966dd52db1762c8f455c1e208df36feb4

I changed the minimum time from 20 ms to 15 ms in test_time.test_process_time(), in Python 3.7 and master branches.

New failure on AppVeyor:

FAIL: test_thread_time (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\projects\cpython\lib\test\test_time.py", line 553, in test_thread_time
    self.assertLess(stop - start, min_time)
AssertionError: 0.015625 not less than 0.015

Oh. In my previous attempt, I changed the minimum time from 20 ms to 15 ms. But I didn't notice that one check uses assertLess() rather than assertGreaterEqual(). So I wrote PR 10724 to change the maximum from 15 ms to 30 ms in test_thread_time().

New changeset 65c216e74f7957006ef7653b7e2afe83007c45ce by Victor Stinner in branch 'master':
bpo-33723: Fix test_time.test_thread_time() (GH-10724)
https://github.com/python/cpython/commit/65c216e74f7957006ef7653b7e2afe83007c45ce

New changeset 5350dd1b50e60882a2da6d53ed27e02d2b698f2e by Miss Islington (bot) in branch '3.7':
bpo-33723: Fix test_time.test_thread_time() (GH-10724)
https://github.com/python/cpython/commit/5350dd1b50e60882a2da6d53ed27e02d2b698f2e

I close again the bug.

A test failed again on AppVeyor! In the 3.7 branch, on this test:

def busy_wait(duration):
    deadline = time.monotonic() + duration
    while time.monotonic() < deadline:
        pass
(...)
    def test_thread_time(self):
        (...)
        # bpo-33723: A busy loop of 100 ms should increase thread_time()
        # by at least 15 ms, but less than 30 ms in other threads.
        # Tolerate 15 and 30 ms because of the bad resolution
        # of the clock on Windows (around 15.6 ms).
        min_time = 0.015
        max_time = 0.030
        busy_time = 0.100

        # thread_time() should include CPU time spent in current thread...
        start = time.thread_time()
        busy_wait(busy_time)
        stop = time.thread_time()
        self.assertGreaterEqual(stop - start, min_time)   # <===== HERE =====
        (...)

FAIL: test_thread_time (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\projects\cpython\lib\test\test_time.py", line 550, in test_thread_time
    self.assertGreaterEqual(stop - start, min_time)
AssertionError: 0.0 not greater than or equal to 0.015

IMHO the root issue is that time.thread_time() and time.monotonic() clocks are just... unrelated... on Windows.

This time, I give up and wrote a PR to remove the "busy loop" tests, sorry! PR 10773.

New changeset 48498dd57f79ab1d061c754ad6a2ebe1a7172b0e by Victor Stinner in branch 'master':
bpo-33723: Remove busy loop from test_time (GH-10773)
https://github.com/python/cpython/commit/48498dd57f79ab1d061c754ad6a2ebe1a7172b0e

New changeset d46d753d152a5d01f9c454d18b1ae660509d9b16 by Miss Islington (bot) in branch '3.7':
bpo-33723: Remove busy loop from test_time (GH-10773)
https://github.com/python/cpython/commit/d46d753d152a5d01f9c454d18b1ae660509d9b16

I removed the "busy loop" tests from 3.7 and master.