You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@gump.apache.org by Stefan Bodewig <bo...@apache.org> on 2005/01/18 10:34:03 UTC
Maven generated descriptors and runtime="true"
Hi all,
when Jelly switched to a Maven generated descriptor, it lost all the
runtime="true" attributes on the dependencies. As a result Latka,
which uses Jelly during its tests, no longer picks up
commons-collections (and probably others).
Before I start adding the Jelly runtime dependencies left and right,
is there any way to recover the runtime attributes in the descriptor?
Cheers
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Stefan Bodewig <bo...@apache.org>.
On Wed, 19 Jan 2005, Brett Porter <br...@gmail.com> wrote:
> SPF is running - it will drop it, and I've seen it happen.
Seems to be working again - or at least did yesterday.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Brett Porter <br...@gmail.com>.
On Tue, 18 Jan 2005 12:09:11 +0100, Stefan Bodewig <bo...@apache.org> wrote:
> On Tue, 18 Jan 2005, Brett Porter <br...@gmail.com> wrote:
>
> >> If we allow commons-dev@jakarta to post to commons-dev@jakarta,
> >> then suddenly a lot of viruses will get through. How many viruses
> >> have you received that have been "sent" from yourself?
> >
> > Doesn't the mailer drop it because it wasn't sent from the
> > apache.org domain?
>
> I don't think so. I can use my apache.org address while posting from
> a different machine without problems.
>
SPF is running - it will drop it, and I've seen it happen. However, it
is on *.apache.org, not on just apache.org mail domains (which is why
you can do it for your own).
mail sent from brutus won't get through though, so that explains why
its not working now either.
- Brett
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 18 Jan 2005, Brett Porter <br...@gmail.com> wrote:
>> If we allow commons-dev@jakarta to post to commons-dev@jakarta,
>> then suddenly a lot of viruses will get through. How many viruses
>> have you received that have been "sent" from yourself?
>
> Doesn't the mailer drop it because it wasn't sent from the
> apache.org domain?
I don't think so. I can use my apache.org address while posting from
a different machine without problems.
> I don't think it is Ant that needs to get the property though - its
> junit, isn't it?
It's the VM running the JUnit tests, yes.
> If so, maven.junit.sysproperties should be being used anyway. I'll
> look into it.
Great, thanks.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Brett Porter <br...@gmail.com>.
> They should (Maven should use CVS Ant and CVS Jelly) - at some point.
Sure, but one step at a time.
> > They've gotten through in the past. We should probably change the
> > from address to general@gump.apache.org and make sure it gets
> > moderated through, I guess.
>
> IMHO we should use real email addresses or general@gump, yes.
general@gump is a real email :) Using "people" is a bit deceptive I found.
> If we allow commons-dev@jakarta to post to commons-dev@jakarta, then
> suddenly a lot of viruses will get through. How many viruses have you
> received that have been "sent" from yourself?
Doesn't the mailer drop it because it wasn't sent from the apache.org domain?
> For the Ant builds, setting ant.build.clonevm did the trick (passing
> system properties to forked VMs). Maven would have to use CVS Ant in
> order to make this work. Hmm, I don't even know whether we currently
> support system properties for <maven> at all.
I don't think it is Ant that needs to get the property though - its
junit, isn't it? If so, maven.junit.sysproperties should be being used
anyway. I'll look into it.
- Brett
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 18 Jan 2005, Brett Porter <br...@gmail.com> wrote:
>> > I didn't quite understand why something would be needed at build
>> > time, but not runtime?
>>
>> Ant or JavaCC, for example.
>
> Right, build != compile - sorry I wasn't thinking straight. These
> don't often affect Maven if the plugin is assumed to be installed.
They should (Maven should use CVS Ant and CVS Jelly) - at some point.
> However, doesn't it really make more sense to be the other way
> around? These are surely less common?
It depends on the complexity of the project. There are projects
without any runtime dependencies at all. In general, you are correct
- but the distinction has been added as an afterthought, we'd now need
to patch a whole lot of descriptors if we changed the default of the
runtime attribute.
>> I don't understand that either. They probably sit in the
>> moderation queue since commons-dev@jakarta isn't subscribed to
>> itself.
>
> They've gotten through in the past. We should probably change the
> from address to general@gump.apache.org and make sure it gets
> moderated through, I guess.
IMHO we should use real email addresses or general@gump, yes.
If we allow commons-dev@jakarta to post to commons-dev@jakarta, then
suddenly a lot of viruses will get through. How many viruses have you
received that have been "sent" from yourself?
>> OK, swing fails since the tests don't pick up the java.awt.headless
>> sysproperty.[1]
For the Ant builds, setting ant.build.clonevm did the trick (passing
system properties to forked VMs). Maven would have to use CVS Ant in
order to make this work. Hmm, I don't even know whether we currently
support system properties for <maven> at all.
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Brett Porter <br...@gmail.com>.
> > I didn't quite understand why something would be needed at build
> > time, but not runtime?
>
> Ant or JavaCC, for example.
Right, build != compile - sorry I wasn't thinking straight. These
don't often affect Maven if the plugin is assumed to be installed.
However, doesn't it really make more sense to be the other way around?
These are surely less common?
> I don't understand that either. They probably sit in the moderation
> queue since commons-dev@jakarta isn't subscribed to itself.
They've gotten through in the past. We should probably change the from
address to general@gump.apache.org and make sure it gets moderated
through, I guess.
> No, the tests [ERROR] out, and I think it is because ouf a
> ClaasNotFoundException for org.jaxen.JaxenException. It was the last
> time when I checked.
Ok, will take a look when they appear on the main list. I don't have
time tonight anyway.
> OK, swing fails since the tests don't pick up the java.awt.headless
> sysproperty.[1]
Some of these were not actually being built under gump previously I
don't think, hence the reason they fail now and not before :)
- Brett
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 18 Jan 2005, Brett Porter <br...@gmail.com> wrote:
> I didn't quite understand why something would be needed at build
> time, but not runtime?
Ant or JavaCC, for example.
>> If that helps. I'm not sure. We still have a few cases of
>> jelly-tags builds that passed in the Ant incarnation but fail now
>> that we use Maven.
>
> which ones? There haven't been any mails,
I don't understand that either. They probably sit in the moderation
queue since commons-dev@jakarta isn't subscribed to itself.
> and http://brutus.apache.org/gump/public/project_todos.html doesn't
> show them.
Should do so shortly.
dom4j failed because tests failed and so we didn't try to build any
parts of jelly for a few days. I changed dom4j's descriptor and the
first public run that had picked it up has just finished a few minutes
ago. You can already see the builds I mean by looking into the build
log of "public".
> I'm guessing its because other dependent projects are failing?
No, the tests [ERROR] out, and I think it is because ouf a
ClaasNotFoundException for org.jaxen.JaxenException. It was the last
time when I checked.
>> I guess - but that's not backed by any facts yet - that the
>> problem is that CVS dom4j needs Jaxen at runtime, but Maven's junit
>> plugin ignores Jaxen even if it is on the classpath.
>
> "ignores Jaxen" - this shouldn't be the case, though it won't be
> used as the default XML parser I don't think.
If it is not listed as a dependency, the junit plugin won't add it to
the classpath running the tests.
> Anyway, no point stabbing in the dark - if I see a failer, I can
> look at addressing it.
OK, swing fails since the tests don't pick up the java.awt.headless
sysproperty.[1]
Some others fail because of not finding beanutils or BSF
Jetty also has different problems[2] (ports being in use and missing
commons-codec now needed by commons-httpclient).
Stefan
Footnotes:
[1] http://brutus.apache.org/~bodewig/TEST-org.apache.commons.jelly.swing.TestSwingTags.txt
[2] http://brutus.apache.org/~bodewig/TEST-org.apache.commons.jelly.jetty.TestJettyHttpServerTags.txt
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Brett Porter <br...@gmail.com>.
I didn't quite understand why something would be needed at build time,
but not runtime?
> If that helps. I'm not sure. We still have a few cases of jelly-tags
> builds that passed in the Ant incarnation but fail now that we use
> Maven.
which ones? There haven't been any mails, and
http://brutus.apache.org/gump/public/project_todos.html
doesn't show them. I'm guessing its because other dependent projects
are failing?
> I guess - but that's not backed by any facts yet - that the
> problem is that CVS dom4j needs Jaxen at runtime, but Maven's junit
> plugin ignores Jaxen even if it is on the classpath.
"ignores Jaxen" - this shouldn't be the case, though it won't be used
as the default XML parser I don't think. Anyway, no point stabbing in
the dark - if I see a failer, I can look at addressing it.
> quite a bit of work since it also happened to the tags. I'll restore
> the ones for jelly and add more to the tags as the need arises.
Probably best as it was probably cut and pasted anyway, so this should
be more accurate.
- Brett
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Stefan Bodewig <bo...@apache.org>.
On Tue, 18 Jan 2005, Brett Porter <br...@gmail.com> wrote:
> sorry, the docs aren't very clear on runtime.
Probably so.
<depend project="foo" runtime="true"/>
means "I need foo at runtime, not only at build time". runtime
defaults to false.
<depend project="foo" inherit="runtime"/>
means I also need to depend on all dependencies of foo, those that foo
has marked as runtime="true".
> so none of the deps are transitive, unless they are given at run
> time?
Yes.
> I guess another alternative is to switch latka to its maven build?
If that helps. I'm not sure. We still have a few cases of jelly-tags
builds that passed in the Ant incarnation but fail now that we use
Maven. I guess - but that's not backed by any facts yet - that the
problem is that CVS dom4j needs Jaxen at runtime, but Maven's junit
plugin ignores Jaxen even if it is on the classpath.
> Anyway, feel free to restore them to the descriptor (just look at
> the last commons.xml cvs history that had jelly),
quite a bit of work since it also happened to the tags. I'll restore
the ones for jelly and add more to the tags as the need arises.
Cheers
Stefan
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org
Re: Maven generated descriptors and runtime="true"
Posted by Brett Porter <br...@gmail.com>.
sorry, the docs aren't very clear on runtime.
so none of the deps are transitive, unless they are given at run time?
I guess I'm just trying to figure the reason to pass it around before
adding more to the jelly metadata.
I guess another alternative is to switch latka to its maven build?
Anyway, feel free to restore them to the descriptor (just look at the
last commons.xml cvs history that had jelly), and after we've resolved
if they are needed, I can make sure they persist through the generated
descriptor.
- Brett
On Tue, 18 Jan 2005 10:34:03 +0100, Stefan Bodewig <bo...@apache.org> wrote:
> Hi all,
>
> when Jelly switched to a Maven generated descriptor, it lost all the
> runtime="true" attributes on the dependencies. As a result Latka,
> which uses Jelly during its tests, no longer picks up
> commons-collections (and probably others).
>
> Before I start adding the Jelly runtime dependencies left and right,
> is there any way to recover the runtime attributes in the descriptor?
>
> Cheers
>
> Stefan
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
> For additional commands, e-mail: general-help@gump.apache.org
>
>
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@gump.apache.org
For additional commands, e-mail: general-help@gump.apache.org