One thing to note is that the build plan will be constructed _prior_
to any mojos executing, so modding that build plan implies you're
also getting in there ahead of the first mojo's
execution...otherwise, you don't have complete access to modify the
plan. This takes us out of the realm of standard mojo
implementations, and into that of custom components for Maven.
Currently, Maven provides very poor support for customizations like
this...even in 2.1/trunk.
We need to think about improving this somehow, IMO. For now, I can
probably provide a hook to the BuildPlanner functionality (or maybe,
the LifecycleExecutor right after the build planner is called) that
will post-process the build plan before its first use. Then, build
extensions _should_ be able to provide implementations for this hook.
Of course, all of this is only for 2.1-land...2.0.x is very, very
different in this part of the code.
-john
On Aug 15, 2007, at 2:10 AM, Jason van Zyl wrote:
Hi Brian,
Thanks for adding that rule, I have incorporated into the best
practices guide for the client I'm working with and it stands
generally I believe as something any good build should do.
There are a couple things to note just so I convey the goal, and I
think the work that John is doing with the build plan also plays a
part here.
In this particular case the enforcer with this rule should walk and
find all of the plugins without versions, including the ones in the
lifecycle that may be running implicitly for a given packaging:
1) We should be able to notify the user of all the plugin
declarations in violation
2) This information should be captured in a way by the enforcer so
that it can be used in an IDE
3) We should be able to do something with this information and not
make it onerous to the user to correct. So we might eventually
offer a way to find the latest versions, create a chunk of a
pluginManagement section, or anything else that makes it easy to
correct.
John is working on what's called the build plan, which will
eventually be the sum total of everything that _will_ execute and
its configuration. I think the enforcer being able to act on the
build plan would be very powerful. And when we fix the problem
where the raw models are getting are getting tainted then you will
have accurate information. I'm just suggesting that you talk with
John as the build plan will ultimately be one stop shopping for
what is about to happen and then we can proactively make assertions.
Just to note that one very ingenious work around for this problem
(and it is a problem unless you know about every Maven plugin you
use and define versions) is to create a profile, and within that
profile specify a plugin repository and then disable it for
releases and snapshots and this prevents the plugin version manager
from being able to resolve versions and the build dies. It doesn't
provide all the information in a nice way but does the job of
letting a build person no you have some potential instability.
Someone could do something stupid with a clean plugin release, and
if it got automatically updated would throw off your whole system.
So this locking down of plugin versions is absolutely critical in
corporate environments and should be a best practice to follow.
This new rule in the enforcer will actually make this feasible.
Thanks for whipping that up!
Jason
----------------------------------------------------------
Jason van Zyl
Founder and PMC Chair, Apache Maven
jason at sonatype dot com
----------------------------------------------------------
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---
John Casey
Committer and PMC Member, Apache Maven
mail: jdcasey at commonjava dot org
blog: http://www.ejlife.net/blogs/john
