On Jun 19, 2009, at 2:49 PM, Ian Boston wrote:
On 19 Jun 2009, at 21:09, Vincent Siveton wrote:
2009/6/19 Paul Lindner <lind...@inuus.com>:
Okay, sounds good then.
For the truly bleeding edge should we have hudson publish
timestamped
snapshots somewhere?
Personally, I am not in favour to automate the snapshot deployment:
human action ensures that the code does what it is supposed to do.
Cheers,
Vincent
We do already have snapshots going out from hudson, but I agree, if
they are broken in a subtle way and go out to the repo its a pain
for everyone depending on them.
I don't trust humans :-) If people are finding problems with the
builds that aren't caught by automated tests, it would be nice if they
at least file bugs to improve the test suite (if not actually make the
improvements themselves). A list of holes in the tests would actually
be a nice way to encourage more contributions from the community.
Maybe this is already happening, but I only see a handful of issues in
JIRA related to improving test coverage, and they're mostly pretty
broad (although SHINDIG-1031 looks like a good starting point).
It may also be worthwhile to consider more frequent tagged releases...
call them betas, or milestones... the point is that it will give
people a more stable version to depend on than a snapshot, and a more
current version than 1.0-incubating. I'm not sure what Apache rules
are, but I would hope that as long as the release isn't announced or
packaged up for download then it shouldn't require a vote to release a
monthly or biweekly milestone.
-- Tim
