[hyades-dev] JUnit and integration, was: integration vs interfacing

[Thomas L Roche <tlroche@xxxxxxxxxx>]

Michael.Norman Tue, 21 Sep 2004 20:36:08 +0100
> I think there's one important strand that hasn't been aired. When we
> at Hyades say we support JUnit as a sample tool this means that we
> ship the jars, ensure it is all buildable under CPL or in due course
> EPL, document its use and support it for our consumers. This means
> that companies building commercial offering can extend the framework
> and/or the sample tools under a common licence, and raise defects
> against us and we are committed to fixing them. 

<disclaimer>

Well, just to put it on the table: there is another potential
alternative. I don't necessarily support it, but it should be
considered: integrating/supporting JUnit.

</disclaimer>

JUnit is about as bulletproof/battlehardened as any code available.
But 3 things often whined about on the JUnit list

http://groups.yahoo.com/group/junit/messages

are that

0 JUnit is {orphaned, unmaintained}. Beck and Gamma have indeed kinda
  abandoned it. Their opinion, shared by many, that it's "done" is not
  shared by all.

1 JUnit has bugs. Not many, but some are known. IIRC JB Rainsberger is
  talking with Beck and Gamma about them.

2 "JUnit would be so much better with" <your feature here>. See item
  0. An Official Maintainer would hafta dope-slap such persons
  regularly; OTOH there might be the occasional Great Idea.

IIRC something occasionally whined about on the Hyades list is Hyades'
lack of visibility. "Norman Conquest"-style integration of JUnit would
give Hyades "visibility up the wazoo" :-)

Is this possible? Eclipse does have close ties to both Beck and Gamma,
esp the latter. They might be _happy_ to handoff to "someone
responsible" (such as Your August Persons :-) Is this feasible?
Personally I suspect little real work would be involved, other than
monitoring the inevitable bugzilla posts and administering regular
dope-slaps. Is this desirable? I am mentioning, not advocating :-)
YMMV.

> Subject to fixing the CPL/EPL disconnect I think we are reasonably
> safe for Junit but if we move to a broader support of these
> derivatives we may end up with some support headaches.

JUnit and JUnit extensions are certainly qualitatively different WRT
support. IMHO interfacing JUnit extensions is the Way To Go (as
argued in the previous thread).