: We just made it random always, so it will fail consistently regardless
: of JRE implementation.

Right, which means people who:
 * have an app that uses Lucene
 * have tests that use the Lucene test-framework
 * upgrade lucene w/o changing their app or their JVM

...could now get failures.

: Adding CHANGES.txt to test changes seems like a slippery slope, the

But this wasn't a "test change", this was a change to the published 
test-framework.

: behavior and APIs here is undefined, we can and should break it at any
: time we want without any hesitation or notice, and it should be
: assumed that all things are random :)

How is the API undefined?  Starting with 3.1 the framework was specificly 
pulled out into it's own jar that was publicly advertised for lucene 
users to use...

http://lucene.apache.org/java/3_3_0/api/test-framework/index.html
   "The Lucene Test Framework is used by Lucene as the basis for its 
    tests. The framework can also be used for testing third-party code 
    that uses the Lucene API."

I agree that the back-compat on classes in test-framework should be loose, 
and we should be allowed to break compat anytime it helps improve our test 
coverage, but we should at least *mention* in CHANGES when we break it.

right?

-Hoss

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to