Jean-Marc's idea is interesting. The tests in this 'flaky' category should succeed, say, once in 5 consecutive builds. If particular test fails consistently after a checkin, we should investigate further.
On Wed, Apr 3, 2013 at 4:38 AM, Jean-Marc Spaggiari <[email protected] > wrote: > Or even always have them running in this new category but if one of > them fails,not tag the build as a fail? That way we can continue to > keep an eye on them but they are not turning the builds red? > > 2013/4/3 Andrew Purtell <[email protected]>: > > Could we introduce a junit category for flakey tests such that they don't > > run unless we activate them with some flag to maven? > > > > On Tuesday, April 2, 2013, Stack wrote: > > > >> On Tue, Apr 2, 2013 at 6:20 PM, Andrew Purtell <[email protected] > <javascript:;>> > >> wrote: > >> > >> > > - HBASE-7897 Add support for tags to Cell Interface > >> > > >> > I think one of us over here could pick this up, let me ask. > >> > > >> > > - Unit test fixes > >> > > >> > I have some stuff due by the 12th. After that I aim for green trunk > and > >> > 0.94 builds on 54.241.6.143, including the Hadoop 2 ones. Will put on > my > >> > hip boots. > >> > > >> > > >> I like Jimmy's suggestion from another thread where we just back out all > >> flakey tests and roll them in one at a time into a build that is usually > >> blue rather than always red so we can figure which tests and code > >> destabilize. > >> > >> St.Ack > >> > > > > > > -- > > Best regards, > > > > - Andy > > > > Problems worthy of attack prove their worth by hitting back. - Piet Hein > > (via Tom White) >
