[prev in list] [next in list] [prev in thread] [next in thread] 

List:       apache-logging-general
Subject:    Re: [VOTE] Release log4j-1.3a8 official
From:       Mark Womack <mwomack () apache ! org>
Date:       2006-02-06 16:59:29
Message-ID: 284ea76b0602060859g627f182cm6f4d2840c9d15dee () mail ! google ! com
[Download RAW message or body]

On Saturday, I instrumented the TimedLocationWatchdog code to see if I
could get a snapshot of what was going wrong, but of course after
doing that, the test never failed, after repeated tries.  It occured
to me last night that it could be related to a difference in JDK 1.3
vs JDK 1.4 (I was doing everything in 1.3).  So, I will be at it again
this evening.

However, I don't think we should stop the alpha8 release for this.  I
will create a bug to track the issue (if there isn't already an
apporpriate one).  But previously the test was just left out of the
main test suite.  At least now there is a test that can fail
(intermittently).  It is a new feature, not on a critical path for
most users, and all the other parts of the alpha8 release need to get
into the user community in general for review.  If this was beta or
rc, I'd have a much different opinion.

-Mark

On 2/1/06, Curt Arnold <carnold@apache.org> wrote:
>
> On Feb 1, 2006, at 1:44 PM, Mark Womack wrote:
>
> > Well, it is a good nit.  This particular test doesn't always fail
> > though.  Locally on my machine it failed once, and after looking at
> > the code, I ran it again and it worked.  My guess is that it has
> > something to do with the copying of the config file not changing the
> > date so that the watchdog triggers or conceiveably a bug in the
> > FileWatchdog code someplace.
> >
> > There is something similar that I have mentioned related to the
> > TimeBasedRolling scheme as well, though it does not seem to show up in
> > the Gump radar.  I get it fairly often locally.
> >
> > -Mark
>
> Gump is not consistently failing, but it isn't a desirable practice
> to be issuing releases while Gump is failing or immediately after
> Gump starts passing.  The test was recently added at which time they
> would pass on Windows but fail on most Unix platforms.  I modified
> them to get them to pass consistently on my boxes and apparently pass
> inconsistently on Gump.   I do not think it reflects a regression in
> the code base, but either the fragility of the test or a bug that has
> been latent in the code for some time.
>
> Omitting the test would not change the distribution since the unit
> tests are not included.  It would only silence Gump from reminding us
> that we have either a fragile test or a latent bug.  I think
> releasing an alpha under these conditions, while undesirable, is
> acceptable.
>

[prev in list] [next in list] [prev in thread] [next in thread]