The false failures are from this bug in Hudson (right?):
http://issues.hudson-ci.org/browse/HUDSON-7836
Somehow the clover XML is corrupted by Hudson...
Anyway, I think we should find a workaround. In general we shouldn't
let false failures like this make us look bad. Sure we devs inside
know the scoop -- ignore the failure when it says all tests passed.
But to the outside world at first glance it makes our stability look
awful.
What workarounds can we take, short of shutting off the Clover build?
Mike
On Sun, Jan 23, 2011 at 4:30 AM, Uwe Schindler <u...@thetaphi.de> wrote:
> We had no more nightly build failures since hudson update that were caused
> by clover. The recent ones are not caused by that, they are caused by a
> committed "nocommit" by Shaie :-)
>
> Uwe
>
> -----
> Uwe Schindler
> H.-H.-Meier-Allee 63, D-28213 Bremen
> http://www.thetaphi.de
> eMail: u...@thetaphi.de
>
>
>> -----Original Message-----
>> From: Mark Miller [mailto:markrmil...@gmail.com]
>> Sent: Sunday, January 23, 2011 4:33 AM
>> To: dev@lucene.apache.org
>> Subject: Re: Hudson nightly build failures due to Clover
>>
>> Heh - I'm with you - I think it would be awesome to fix :) I think the
> other
>> devs think so too. I think that you prodding the issue is a smart move. We
> do
>> want to encourage users to use the nightlies - lets more people try out
> Uwe's
>> questionable code early on ;) - fake failures are a downer though - it
> probably
>> could easily affect how some users perceive the stableness. Though if you
>> are that concerned, you would hopefully read the message and realize that
>> something about the message is off - it says all tests passed. Not ideal,
> but I
>> wonder how much it *actually* hurts trunk/nightly use. I have not seen
>> much mail about it yet...
>>
>> But then again, I work at lucid too, so I'm probably reading the same
> script
>> you are ;)
>>
>> Bottom line though - everyone knows of the problem. An acceptable
>> solution has not yet been found it seems. I guess we are waiting for the
> bug
>> fix to come out. If anyone really wants it fixed before that - prob going
> to
>> have to put on your thinking cap and dive in with some concrete
>> suggestions/action.
>>
>> - Mark
>>
>> On Jan 22, 2011, at 5:40 PM, Erick Erickson wrote:
>>
>> > RANT WARNING! RANT WARNING!
>> >
>> > When did Lucid enter the picture? This has nothing to do with Lucid.
>> >
>> > As you say this may all be taken care of with the new Hudson, and that
>> > could be the end of the story, hooray!!!!! I'm perfectly willing to wait
> and
>> see if it settles out.
>> >
>> > What this *does* have to do with, from my perspective, is that Solr
>> > hasn't had a release in quite a while. There is lots of goodness in
>> > the 3_x and trunk builds. We see comments on the user's list of "get a
>> > nightly build from trunk or 3_X and try it". Which may be sound
>> > advice. But I can absolutely guarantee that a number of potential
>> > users take a single glance at the number of "failures" (even if they are
>> bogus) reported on Hudson and immediately cross Solr off their list as far
> as
>> using trunk or 3.x.
>> >
>> > It doesn't matter that 1.4.1 would report the same nonsense if it was
>> > continually built. It doesn't matter that 3.x and trunk have far
>> > better automated tests. It doesn't matter that the developers have
>> > confidence. I'm talking perception here, not underlying code quality.
>> > What matters (and I'm talking perception, remember) is that out of the
>> > last 10 3.x builds 6 have "failed", as have 5 of the last 10 trunk
>> > builds. Which makes it easy to dismiss and/or have an exaggerated sense
>> of the instability of the 3_x and trunk builds.
>> >
>> > If there were a solution that allowed us to satisfy both the
>> > developers' needs and this perception, I think we should go for it.
>> >
>> > Now, it may well be that the current situation is acceptable to the
>> > community and that our story should continue to be "be patient, we'll
>> > release sometime". But this story is getting old(er).
>> >
>> > But please don't make the mistake of dismissing stodgy corporate
>> > concerns (and I'm speaking of my experience at several companies
>> > here). They may or may not be valid from a technical perspective. It
>> > may even be that stodgy corporations wouldn't use open source software
>> > anyway. It may be that we just don't care. I'm not in a position to
>> > offer any hard evidence either way. Nor, I suspect is most anyone else
>> given the recent Maven kerfluffle....
>> >
>> > And I have no good response at all to the reply "Ok, wise guy, dive in
>> > and *make* a release happen". "I'm too busy" is a pretty lousy excuse
>> <G>...
>> >
>> > OK, rant pretty much over. It would be an easier thing to recommend
>> > trunk or 3_x if there were some commitment to a release date.
>> > Potential users of the newer branches could at least plan on using one
>> > of them with the expectation that the target would stop moving before
>> > their go-live date. But as it is some number of users will stay on 1.4.1
> for
>> lack of the ability to plan.
>> >
>> > FWIW
>> > Erick
>> >
>> > On Sat, Jan 22, 2011 at 3:37 PM, Uwe Schindler <u...@thetaphi.de> wrote:
>> > The failures from today are just test builds.
>> >
>> > Today also a new Hudson was installed... so please simply wait a few
> days
>> until it settles.
>> >
>> > If Lucid wants their customer to use nightly builds, they could setup
> ones
>> on their servers for their customers? For us Hudson mostly a test system
> to
>> check our commits. And clover is part of that.
>> >
>> > If somebody wants to install a trunk build, they should always svn
> checkout
>> and build themselves. Then they can even fix to specific rev no and can
>> always reproduce their build.
>> >
>> > Uwe
>> >
>> >
>> >
>> > "Erick Erickson" <erickerick...@gmail.com> schrieb:
>> >
>> > >I don't know what other issues you're referring to, but please,
>> > >please, please do whatever you can to remove "false failures". It's
>> > >highly disconcerting to folks we talk to on the message boards to say
>> > >"Functionality you need is in the nightly builds and you can use
>> > >them, but just ignore the errors the build reports. Really, it's OK.
>> > >Trust us".
>> > >
>> > >Putting on my corporate IT hat I'd have serious reservations about
>> > >using code that looks broken all the time (even if it's "just a build
>> > >artifact")...
>> > >
>> > >Erick
>> > >
>> > >On Fri, Jan 21, 2011 at 6:59 PM, Steven A Rowe <sar...@syr.edu> wrote:
>> > >
>> > >> Clover causes Hudson nightly builds to fail intermittently. This
>> > >> is
>> > >bad,
>> > >> because it looks like Lucene/Solr tests are failing when they are
>> > >not. But
>> > >> Clover is good, so nobody wants to turn it off.
>> > >>
>> > >> One possible solution (apologies if this has already been suggested):
>> > >make
>> > >> new nightly Clover-only Hudson builds, and remove Clover from the
>> > >existing
>> > >> nightly builds.
>> > >>
>> > >> I think that would address all of the issues, wouldn't it?
>> > >>
>> > >> Steve
>> > >>
>> >
>> > --
>> > Uwe Schindler
>> > H.-H.-Meier-Allee 63, 28213 Bremen
>> > http://www.thetaphi.de
>> >
>> > ---------------------------------------------------------------------
>> > To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For
>> > additional commands, e-mail: dev-h...@lucene.apache.org
>> >
>> >
>>
>> - Mark Miller
>> lucidimagination.com
>>
>>
>>
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional
>> commands, e-mail: dev-h...@lucene.apache.org
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
> For additional commands, e-mail: dev-h...@lucene.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
