On 22/10/15 21:33, Andy Seaborne wrote:
I was working on processing Windows file names into base URIs and got
fed up with the Windows Jenkins job failing on AlarmClock tests. They
test setting a callback to happen at a specific few 10s of ms time.
These fail, maybe 50% of the time, when some other CI job is running.
When they run on an otherwise empty machine, they pass.
It seems there are multisecond pauses going on. Setting the timeouts so
large as to work on Windows gets ridiculous. There is some muttering on
the web about ScheduledThreadPoolExecutor and cycle countering so VM's
may be involved.
I have turned the test off on Windows for now.
And the Windows file names processing works.
Test reenabled with adjusting timeouts.
JENA-1060 is not related to the windows development job failing on
timeouts (Windows failing in different tests, e.g. test 3, not 5).
But with new, improved (longer) timeouts for Windows, things seem a bit
more stable. We'll see how stable over time.
Andy
