I think you're misunderstanding me: I was talking about test coverage
within _wagon_ itself. Since wagon is on a separate release cycle to
Maven core, Maven releases have to consume released versions of wagon.
This means 72h of voting and so forth before releasing a new wagon version.
I understand this is a lot like interpolation changes within
maven-project, but there is one important difference: when problems
arise with interpolation during RCs, we have a lot of flexibility WRT
fixing those issues, since it's in the Maven codebase (or else in a
dependency that has more relaxed criteria for a release).
My biggest concern here is that proxied connections are one of the most
common problems new users have with Maven; I appreciate that they're
fixed in the 1.0-beta-4 snapshot up for a vote, but the fact that
something this common wasn't tested in wagon itself before it was
released worries me. A lot. It's not appropriate for Maven's own ITs to
serve as the only test suite for wagon, because that means its much more
likely Maven releases will get stuck on this sort of issue, where we're
blocked by a wagon release before we can move forward.
Again, to me it's all about the fact that Maven's core is a consumer of
the wagon libraries, which are released on a separate cycle, for use in
(I'm assuming) several projects. The fact that we got three RCs into a
release of a Wagon consumer project says something about that consumer
project; but IMO the fact that Wagon made it through testing to get a
release out with a defect in such a common feature is much more worrisome.
I'm saying that, again IMO, we need to spend some serious time on
wagon's test suite before we attempt to upgrade the version Maven uses.
-john
Brett Porter wrote:
I understand the concerns, but let me try and explain where I'm coming
from.
On 29/07/2008, at 1:58 AM, John Casey wrote:
What really concerns me is that this release has been in the wild for
quite awhile now, and this issue is just coming to light. Plenty of
people use proxies, so it's not like it's that shady a corner of the
implementation.
Do you mean beta-3, or 2.0.10-RC? For beta-3, the problem is that it's
just impossible to test it without upgrading core, because it forces the
provider-api on you.
I guess what I'm getting at is I wonder if the test suite for wagon is
good enough to give us confidence in moving forward without a
substantial soak period in something like Maven (as could happen if we
make the change asap after 2.0.10, to give people a lot of time using
it in the run-up to 2.0.11).
Unfortunately, I'm not exactly sure how this is going to be any
different to 2.0.10. It was already in there for a month and no one
noticed this particular issue. I tested HTTP proxying, but I messed
something up in my environment - there is now an integration test to
verify it instead. I spent a lot more time on testing WebDAV and SSH as
that's where the bulk of the changes were. We did go through a
reasonable amount of testing a month ago to the point people were happy
to approve the wagon release.
I think this is quite the same as the interpolation changes - we knew
they were risky and were prepared to spend more time testing them, and
as issues have cropped up you have fixed them and moved forward beefing
up the test suite along the way.
The fixes in Wagon have been very insignificant in b3 to b4 if you want
to review them all:
http://mail-archives.apache.org/mod_mbox/maven-wagon-commits/200806.mbox/[EMAIL PROTECTED]
http://mail-archives.apache.org/mod_mbox/maven-wagon-commits/200807.mbox/[EMAIL PROTECTED]
http://mail-archives.apache.org/mod_mbox/maven-wagon-commits/200807.mbox/[EMAIL PROTECTED]
http://mail-archives.apache.org/mod_mbox/maven-wagon-commits/200807.mbox/[EMAIL PROTECTED]
I would prefer to get this in and call on users to test deployment and
proxying in their environment specifically. The RCs are the best time to
draw light to it.
I care as much as everyone else about not having any regressions. Given
the extent of the interpolation changes as well as these, this is a good
release to run through a longer series of RCs anyway, and we're in no
rush - right?
Cheers,
Brett
--
Brett Porter
[EMAIL PROTECTED]
http://blogs.exist.com/bporter/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
John Casey
Developer, PMC Member - Apache Maven (http://maven.apache.org)
Blog: http://www.ejlife.net/blogs/buildchimp/
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
