commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From robert burrell donkin <>
Subject Re: [pool] Announcing Release Candidate 1 for Pool 1.3
Date Wed, 22 Mar 2006 20:51:43 GMT
On Tue, 2006-03-21 at 23:55 -0500, Sandy McArthur wrote:
> On 3/21/06, Niall Pemberton <> wrote:


> > 4) The links on the changelog and file activity report are invalid. This is
> > a bug in the maven plugins, which may be resolved in a later version. One
> > way to get round this is to change to url in the repository element of
> > project.xml to point to the root subversion directory, rather than commons
> > pool (and since your using a custom "cvs-usage" report it shouldn't matter):
> >
> > <repository>
> >    <url></url>
> > </repository>
> Setting the url like you suggested fixes the changelog links but
> doesn't help the file activity report. Actually both reports are wrong
> because they don't go back far enough. I think this is related to the
> cvs to svn conversion. How about just removing those two reports until
> the next release?

there are doubts about the accuracy of these reports for dates before
the date of the last import from CVS. so i'm happy to turn them off for

> > 6) Theres a checkstyle config file in pool, but it doesn't appear to be used
> > (and isn't included in the source distro).
> That may be because when you include the checkstyle report it lists
> about 5,000 style errors. Using the checkstyle.xml from IO brings that
> number down to 510, almost all of which are JavaDoc related. When I
> was working on the Pool 1.3 branch I mostly focused on bug fixes and
> didn't try to touch more than was needed. The trunk for pool 2 has
> more extensive clean ups. I suggest the checkstyle be left out until
> the 2.0 release.


> > I ran the tests using maven with JDK 1.3.1_04  and JDK 1.4.2_05 on Windows
> > XP and they passed OK. On W2K the tests passed with JDK 1.4, but failed for
> > JDK 1.3 (TestGenericKeyedObjectPool failed on testEviction2). However my W2K
> > m/c is a bit flakey so I would ignore this unless someone else verifies it.
> This would be a result of the test expecting more determinist behavior
> from the thread scheduler than the JVM guarantees. Odds are if you
> re-run the tests they'll pass most of the time. I've made a good
> effort to make the unit tests in the trunk all be deterministic while
> I haven't put that effort into the 1.3 branch.

a note to that effect might (in more diplomatic language) would make a
good addition to the release notes.

- robert

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message