[python-committers] AppVeyor and Travis-CI

[Terry Reedy]

On 5/16/2018 2:31 AM, Victor Stinner wrote:
> Oh, maybe I saw them. Are you talking about:
> 
> "test_multiprocessing_forkserver: TestIgnoreEINTR.test_ignore() fails
> on Travis CI"
> https://bugs.python.org/issue33531
> 
> "test_asyncio: test_subprocess test_stdin_broken_pipe() failure on Travis CI"
> https://bugs.python.org/issue33532
> 
> (I just created these two issues.)

Those are the two test files that failed, though I believe there were 
more test methods that failed in test_asyncio.

> 2018-05-16 2:22 GMT-04:00 Victor Stinner <vstinner at redhat.com>:
>> It seems like the job has been rescheduled.

I later restarted just the Travis test, so as to not toss the OK 
Appveyor test and have to wait another 4 hours to get it to pass again.

  I cannot see the failure
>> ("394 tests OK"). Would you mind to open a bug report?
>>
>> Do you only see these failures in the 3.6 branch?

No.  However, it happened again on my next 3.6 only patch.  I restarted 
just Travis and it passed.  The failure rate the last 24 hours has been 
about 1/3.

>> 2018-05-15 16:36 GMT-04:00 Terry Reedy <tjreedy at udel.edu>:
>>> I am getting repeated bogus failures, completely unrelated to a trivial
>>> patch, more often than not, on Travis-CI
>>> For instance,
>>> https://travis-ci.org/python/cpython/jobs/379349709
>>> 2 tests failed:
>>>      test_asyncio test_multiprocessing_forkserver
>>> 1 test altered the execution environment:
>>>      test_importlib
>>>
>>> The failures repeated on retest.  Can someone turn off the bad test methods?