Hi,

On Fri, Jan 24, 2014 at 9:48 AM, Julian Reschke <[email protected]> wrote:
> a) tune the test,

Since it's a jackrabbit-core tests instead of one in the TCK, the
assumptions it makes are a bit more strict than what the spec would
require. But it would be fair for us to relax the test to allow the
getSize() to return also -1 instead of the expected number of results.

> b) tune the timeout,

A more accurate approach would probably be to raise the minimum number
of items to prefetch to something like 100, and make that limit
override the timeout (i.e. the timeout would kick in only after the
minimum prefetch count is reached). That would make the test fully
deterministic, but wouldn't work too well in practice for example for
a query that returns zero or just a few results but has to traverse
the entire repository to figure that out.

> c) mark the test is failing for now due to slow backend?

That would also be OK. Doing so would reflect the fact that the
behavior of getSize() in Oak is more relaxed than in Jackrabbit
Classic.

I'd go for c, as that's the easiest and least troublesome thing to do.

BR,

Jukka Zitting

Reply via email to