qpid-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steve Vinoski <vino...@iona.com>
Subject Re: [java] client tests hanging
Date Fri, 10 Nov 2006 14:48:38 GMT
On Nov 10, 2006, at 3:28 AM, Marnie McCormack wrote:
> As you all know we're currently getting organised for M1.

Yep, and I've been working to make maven ready for it. Brian asked  
the question about moving to maven for M1 in another thread last  
night, and I'm surprised to see no answers to it this morning.

> At things stand
> there are several open JIRAs for this work. Most of the changes you  
> are
> seeing are directly related to the legal/license requirements for  
> releasing,
> with some others being derived from these changes.

This isn't accurate. I don't think the reorganization of tests that  
occurred earlier this week was necessary for legal/license requirements.

All I'm saying in my email below is that 1) it would have been nice  
to have gotten advance notice that the tests were going to be  
reorganized, so that I wouldn't have wasted time moving the old stuff  
around on my branch, and 2) the only thing keeping maven from being  
ready for M1 is the need for further improvement to the test structure.

> Martin, Rajith & I have been doing the work for the release, but  
> everyone is
> most welcome (indeed encouraged) to contribute to the leg work on  
> this front
> - via the JIRAs (and any new ones you think we need). We're hoping to
> complete the restructuring required for release later today. It's not
> exciting, but we're getting there.

And as I've noted several times, if anyone wants to pitch in and help  
on the maven front, that would be great as well.

> For unrelated changes to the tests that you think a good idea -  
> probably
> best to raise JIRAs with details and progress them after M1 is  
> complete to
> minimise flux.

These aren't unrelated test changes -- they're required for maven to  

> Plenty to be done :-)

Indeed, but let's not make a flurry of immediate changes in the name  
of M1 based on assumptions that may or may not be accurate. You seem  
to be assuming that maven is not in the M1 picture, while a number of  
the rest of us are not making that assumption, as indicated by  
Brian's other email thread.


> Marnie
> On 11/10/06, Steve Vinoski <vinoski@iona.com> wrote:
>> On Nov 9, 2006, at 10:04 AM, Martin Ritchie wrote:
>> > - Reorganised unit tests to test.unit
>> >    + This allows ant build system to show what is being built and
>> > allow easier detection of failure
>> BTW, despite the reorganization of the unit tests earlier this week,
>> the client tests still suffer a fundamental flaw: they depend on test
>> classes under the broker directory. In general, unit tests, for
>> purposes of isolation, should have as few dependencies as possible,
>> so this arrangement is less than ideal. It's also specifically bad
>> from a maven POV because by default maven builds each subproject's
>> unit tests separately. Such isolation enforcement via standardized
>> project directory structure is another nice feature of maven.
>> The client org.apache.qpid.test.unit.ack.DisconnectAndRedeliverTest
>> class appears to be a big offender in this regard, for example. This
>> is one of the issues that's been giving me fits on the maven branch.
>> Over there, I've been trying to move tests that aren't really unit
>> tests out of the subproject test subdirectories and into a new top-
>> level subproject under qpid/java called systests. The unannounced
>> test reorganizations this week have not helped the maven work,
>> because it means I have to pick up the new structure from the trunk
>> and get it over on the branch, determine yet again which tests are
>> actual unit tests and which aren't, and try again to move everything
>> to where it needs to be. Please don't do any more moves in the trunk
>> like that for now. Rather, if you'd like to help move things around,
>> let's do so on the maven branch, as this issue is really the only
>> blocker remaining to having maven working.
>> thanks,
>> --steve

View raw message