Thank you for the answers and the wiki page -- its beginning to become clear.
However, I am still having problems. If I define multiple lifecycle-mapping pluginExecutionFilter, and then another in a parent pom, the effective pom only shows one of them (the first one defined in the local pom: excluding the second and the one from the parent pom). I also get a NullPointerException (see details below) -- which I assume might be a result of this discrepancy in the effective pom. I couldn't find a related bug for this...
I also attempted to add the buildhelper plugin connector which also had NPEs.
Errors running builder 'Maven Project Builder' on project 'apollo-taurus'.
java.vendor=Sun Microsystems Inc.
BootLoader constants: OS=win32, ARCH=x86_64, WS=win32, NL=en_CA
Framework arguments: -product org.eclipse.epp.package.java.product
Command-line arguments: -data c:\tools\eclipse\eclipse-indigo\workspace -os win32 -ws win32 -arch x86_64 -product org.eclipse.epp.package.java.product
On Fri, May 27, 2011 at 9:45 AM, Igor Fedorenko <igor@xxxxxxxxxxxxxx>
I wish it was that simple.
First, "execute" was not the default. m2e had none less than four
different sets of maven goals it ran during project import, project
configuration update and workspace full and incremental builds. Some of
these were configured at workspace level, some in project/.settings. On
top of that there was project-level setting to "skip"
Second, "execute" _is_ the root cause of missing resources, endless
builds and memory leaks, so making it the default basically blocks all
other possibilities to improve.
Also, m2e 0.13 is much better at detecting incompatibilities and
reporting them. So when 0.12 was kinda-sorta-working-sometimes, 0.13
creates those big red obvious error markers. We could hide the error
markers by default and let the users _guess_ what's going on. I don't
think there are too many projects that worked well in 0.12 and don't
work in 0.13, so most likely hiding the errors will make users feel
better about m2e 0.13 when comparing it with 0.12. At the same time I
believe failing fast and obviously is the right thing to do.
On 11-05-27 02:31 AM, Jochen Wiedmann wrote:
On Fri, May 27, 2011 at 6:02 AM, Igor Fedorenko<igor@xxxxxxxxxxxxxx> wrote:
Of course, there is a chance we overlooked another, simpler solution, so
if you think you know how to make m2e work reliably and efficiently with
all/many projects while failing gracefully for the rest, please bring
this up on m2e-dev and I will see how is the best to schedule this.
For example, how about default=execute, at least while developing the
metadata library? That would, IMO, spare us this discussion and lots
of similar discussions in the near future? And, if I get things right,
we'd had not a worse state than we had before. (Which we currently do,
IMO. I struggled about one hour yesterday in the evening in trying to
get my poms working. Although there was definitely another problem in
m2e as well, which I'll report later on, so the lifecycle mapping
isn't the only thing to blame.)
+1 613-592-6776 x254http://www.nanometrics.ca