Chris Santerre <[EMAIL PROTECTED]> writes:
> We do test against multiple corpa. This typo would pass --lint. It
> would only hit on rare occasions, which is why no one had heard of it
> until now. But the ability for autoupdates and fast fixes makes this
> work quite well. We love the GA runs, but sometimes things just have
> to come out a little faster.
Sorry, I think I came off a bit too bluntly, but I didn't mean to imply
that this is anything beyond the reach of SARE or anyone else. Nor do I
want to imply that we want to be exclusive of new rule development when
it comes to the infrastructure we've developed for testing:
- SVN lets everyone update easily and automatically (svn update once a
night) and it has other benefits in the development process
- multiple people can run their respective corpora on each update
(rsync up to submit results, rsync down to collate, see our nightly
results collated at http://www.pathname.com/~corpus/DETAILS.new
... due to be moved to an ASF machine once someone gets around to it)
- everyone can iterate and improve very quickly
SA testing actually moves really fast and it generates really good rules
that we *could* use the very next morning (we run tests at 0900 UTC
which is the lowest ebb of the developer's development activity). The
problem is that there are only 2 or 3 of us working on test rules and
we're also working on the perl code at the same time, so basically, it's
a pack of 2 or 3 fast people, not half a dozen. Well, one more now that
Matt Yackley a committer. :-)
Daniel
--
Daniel Quinlan anti-spam (SpamAssassin), Linux,
http://www.pathname.com/~quinlan/ and open source consulting
