On Apr 2, 2016, at 18:19, Tres Seaver <tseaver@palladion.com> wrote:

On 04/02/2016 07:00 PM, Zope tests summarizer wrote:

...

[1] FAILED winbot / ZODB_dev py_265_win32 https://mail.zope.org/pipermail/zope-tests/2016-April/093154.html

[2] FAILED winbot / ZODB_dev py_265_win64 https://mail.zope.org/pipermail/zope-tests/2016-April/093155.html

[3] FAILED winbot / ZODB_dev py_270_win32 https://mail.zope.org/pipermail/zope-tests/2016-April/093152.html

[4] FAILED winbot / ZODB_dev py_270_win64 https://mail.zope.org/pipermail/zope-tests/2016-April/093153.html

All four are the same failure:

- ------------------------------ %< ---------------------------------------- Failure in test c:\buildslave\zodb_dev_py_265_win32\build\src\ZODB\historical_connections.t xt

- ---------------------------------------------------------------------- File "c:\buildslave\zodb_dev_py_265_win32\build\src\ZODB\historical_connections. txt", line 220, in historical_connections.txt Failed example: [c in conns1 for c in conns2] Expected: [False, False, True, True] Got: [False, False, False, True]

- ------------------------------ %< ----------------------------------------

I've seen this locally, occasionally, and it's failed like this on the buildbot before too. This test is relying on time.time() and time.sleep() to do its thing: >>> db2.setHistoricalTimeout(.01) >>> time.sleep(.1) >>> conns2[2].close(); conns2[3].close() .... >>> conns1 = [db2.open(before=serial) for i in range(4)] As far as I can see (but I'm not a doctest expert), these functions aren't monkey-patched in this test the way that time.time, at least, is patched in, say, src/ZODB/DemoStorage.test (below). Perhaps they should be? I can open a copy-and-paste PR to do so. .. We need to mess with time to prevent spurious test failures on windows >>> now = 1231019584.0 >>> def faux_time_time(): ... global now ... now += .1 ... return now >>> import time >>> real_time_time = time.time >>> if isinstance(time,type): ... time.time = staticmethod(faux_time_time) # Jython ... else: ... time.time = faux_time_time

...

[5] FAILED winbot / ztk_10 py_254_win32 https://mail.zope.org/pipermail/zope-tests/2016-April/093156.html

I wish I could work out how to get the 'pyh_254' tests turned off altogethe r.

+1. And also for the ZODB_dev py_265 tests too, because ZODB no longer supports 2.6 (https://github.com/zopefoundation/ZODB/pull/48). Jason

On Apr 3, 2016, at 07:25, Tres Seaver <tseaver@palladion.com> wrote:

Hmm, that looks like it could work -- when I looked at the test yesterday, I wondered if there were a race condition going on due to load on the machine.

https://github.com/zopefoundation/ZODB/pull/51 Now all the builds *consistently* fail with the original error ([False,False,False,True]). That's progress, I guess? More comments on the PR. Jason