Tuscany weekly IRC chat log December 4, 2006

[ant elder <an...@gmail.com>]

Here's the chat log from yesterdays chat. Only thing discussed was the axiom
snapshot problem of the M2 release. Looks like thats probably a non-issue.

   ...ant

Session Start: Mon Dec 04 16:31:07 2006
Session Ident: #tuscany
* Now talking in #tuscany
* jboynes_ has joined #tuscany
* jboynes has quit IRC (" HydraIRC -> http://www.hydrairc.com <-")
* jboynes_ is now known as jboynes
<jboynes> hi
<jboynes> anything to chat about this morning?
<cr22rc> hi
<cr22rc> got cold here :-)
* simonnash has joined #Tuscany
<jboynes> not so cold here - especially SoCal :(
<cr22rc> any update on M2/Axis 1.1.1 ?  maybe interesting
<jboynes> ant_ rfeng are you working together on this snapshot problem?
* dkulp has left #tuscany
<rfeng> hi
<rfeng> I already did some investigation on the axis2 transitive
dependencies and had a note posted before ant's one. We're seeing the same
behavior
<jboynes> so it sounds like the snapshot may not be an issue?
<rfeng> but what I'm not very sure is how to guarantee the axis2 artifacts
are downloaded from maven1 repo instead of m2 repo
<rfeng> I did one test, if I get the artifacts from m2 repo, then it's
really a problem
* Amita has joined #tuscany
<cr22rc> with 1.1.1 all of this will go away... right? they'll remove all
snapshot refs there .. this time.
<simonnash> is there any danger that just because this works OK at the
moment, it might not mean that it will continue to work?
<jboynes> so if axis2 comes from the m2 repo then we do end up using the
snapshot?
<jboynes> but if it comes from m1 we use a released version?
<rfeng> seems so
<jboynes> ugh
<rfeng> because we also have axiom artifacts listed in the axis2 pom.xml
* kgoodson has joined #Tuscany
<jboynes> in our axis2 pom, yes?
<rfeng> yes
* ant_ is now known as ant_away
<jboynes> if we do that  and they are before the axi2 dependency, are we
guaranteed to get the released version?
<jboynes> even if we get them from the m2 repo?
<jboynes> iirc there's some priority in how deps get resolved
* jsdelfino has joined #tuscany
<jboynes> something like the dep of a dep gets resolved but not returned if
overridden
<rfeng> That's not what I found when I add axiom 1.2.1 to axis2 pom.xml
<jboynes> dar
<jboynes> darn
* halehM has joined #tuscany
<rfeng> any maven expert knows how to control the ordering of maven repo?
<simonnash> that's what I'm concerned about... it seems risky to rely on
this ordering
<jboynes> I don't think you can (I think its a Set in the code)
<rfeng> :-(
<rfeng> the mixed usage of m2 and m1 is troublesome
<jboynes> but the order should not matter if there is a "closeness" priority
<rfeng> right
<cr22rc> where is 1.1.1 of axis and will they fix this there ?
<rfeng> the root problem is axiom and it's already fixed in 1.2.1
<rfeng> but we come from the path of axis2 1.1
<jboynes> yeah but they need to update axis2 to the new one, right?
<rfeng> yes
<jboynes> which means 1.1.1 :(
<rfeng> y
<jboynes> does anyone know where that is?
<cr22rc> what i asked
<simonnash> dimas said they would decide this week what to do about that
<simonnash> dims
<cr22rc> curious: if all projects when they release ONLY referenced
non-snapshot repos in pom's than this would never happen right?
<jboynes> yes
<cr22rc> seems like a practice we should encourage ?
<rfeng> A side note, isn't it also risky to use SNAPSHOT versions of maven
plugins?
<simonnash> i had presumed that it was mistake on their part rather than
intentional
<jboynes> simonnash: that's the artifact level not the repo level
<cr22rc> right
<jboynes> if you made cr22rc's change then the build would fail as the
artifact would not be found
<simonnash> sorry i did not read you previous msg carefully enough
<jboynes> unfortuanately people add them to make it easier for users
<jboynes> rfeng: yes
<jboynes> but I don't we reference any except the assembly one
<rfeng> ok
<rfeng> It seems that running mvn command against the plugin prefix (such as
mvn eclipse:eclipse) will get the SNAPSHOT versions
<jboynes> ugh
<jboynes> nice and friendly
<rfeng> unless you configure it in the pom or plugin registry with a fixed
version
<jboynes> is there a 'latest release' version ?
<rfeng> yes, for the maven-eclipse-plugin
<jboynes> I mean that you can add to the pom
<rfeng> yes
<jboynes> I seem to remember something like that
<rfeng> If I add it, then it's fine
<jboynes> what do you add?
<rfeng> are you saying a special version as "latest release"?
<jboynes> yeah <version>LATEST
<rfeng> not case by case?
<jboynes> right
<rfeng> oh, I didn't try that
<jboynes> there's some magic for version ranges that I think applies to
plugins
* isilval_ has quit IRC ("Trillian (http://www.ceruleanstudios.com")
<jboynes> can't remember the details
<rfeng> anyway, back to the axis2 issue, how much is the risk?
<rfeng> We have the maven1 repo listed in the parent/pom.xml
<jboynes> sca/parent?
<rfeng> In my case, it seems that we always get the m1 artifacts for axis2
<rfeng> it's in the root parent/pom.xml
<rfeng> pom/parent ...
<jboynes> ok
<jboynes> does that help or hinder?
<rfeng> it's also very frustrating that only axis2 1.1 artifacts (not axiom,
...) are published to http://people.apache.org/repository
<lresende> are these issues related to axis2/axiom also affecting trunk code
? i'm getting build failure around databind-jaxb
<jboynes> rfeng: why is that frustrating?
<rfeng> I have no idea which m1 repo should be used
<jboynes> ah ok
<rfeng> it's not in one place (official m1 repo,
http://people.apache.org/repository and ws zones for woden)
<rfeng> kind of ugly
<rfeng> lresende, no
<jboynes> and it /should/ be in
http://people.apache.org/repo/m1-ibiblio-rsync-repository/
<jboynes> (well, not woden but the rest)
<rfeng> referencing
http://people.apache.org/repo/m1-ibiblio-rsync-repository/ is also
problematic as I found in the trunk
<jboynes> you should never reference that, it gets replicated to
http://www.ibiblio.org/mave
<jboynes> you should never reference that, it gets replicated to
http://www.ibiblio.org/maven
<rfeng> for example, there're some commons-xxx m1 artifacts and they get
pulled in and the transitive dependencies break
<rfeng> agree
<jboynes> but as it's a replica then it should break equally as well
<jboynes> :)
<rfeng> any idea about if the m1 artifacts get automatically converted into
m2 repo?
<jboynes> there used to be some conversion, I'm not sure if that is still
going on
<jboynes> I'd suggest asking on repository@a.o
<rfeng> I had some e-mail exchanges with the axiom 1.2.1 RM, and he only
uploaded stuff into the
http://people.apache.org/repo/m1-ibiblio-rsync-repository/ but I started to
see m2 artifacts
<rfeng> ok
<jboynes> that sounds like "yes it is still being converted" :)
<rfeng> and the pom file is missing, for example,
http://repo1.maven.org/maven2/org/apache/ws/commons/axiom/axiom-impl/1.2.1/
<jboynes> :(
<jboynes> I didn't see a [vote] thread for that - was there one?
<rfeng> no, there was only a vote to have the 1.2.1 release
<jboynes> odd
<simonnash> jboynes, sorry but I'm not sure what is "that".  which vote did
not happen?
<jboynes> axiom 1.2.1
<simonnash> oh you mean the actual publication of it?
<jboynes> yes, I saw a vote to do the release but nothing on the actual
content they were publishing
<simonnash> ok, i understand
* Amita has quit IRC
<jboynes> aiui are options are to continue to wait for axis2 1.1.1 or to
figure out if we can ignore the snapshot
<jboynes> rfeng: do you agree?
* bertlamb_ has quit IRC (Read error: 110 (Connection timed out))
<simonnash> i agree those are the options
* dkulp has joined #tuscany
* isilval_ has joined #tuscany
<jboynes> ok
* isilval_ has quit IRC (Client Quit)
* ant_away is now known as _ant
<_ant> sorry, got stuck on a call for al lthis chat
* jmarino has joined #tuscany
* bertlamb_ has joined #tuscany
<simonnash> ant, we we talking about various options for the axiom SNAPSHOT
issue.  can you review the log?
<_ant> so i'm trying to quickly catch up, did there something wrong with
what i posted about the snapshot jars not being used?
<simonnash> i think you need to read all the discussion.  it got a bit
detailed.  are you sure you were picking up from m2 repo or was it maybe
from m1 repo?
<_ant> i was using whatever is in the M2 branch
<_ant> why does M1/M2 matter if its not used in the classpath?
<simonnash> not M2 brach, m2 maven repo.  apparently the M2 branch can pick
up some parts of axis from the m1 repo.  you need to read all the
discussion.
<_ant> i've skimmed it all now i think, so rfeng, are you saying you have
had the situation where the snapshot are actually on the classpath?
<rfeng> yes, if the axis2 artifacts are pulled in from m2 repo
<rfeng> then the transitive dependencies work :-(
<_ant> i've had them in my local M2 repo and they're not used, is that
different?
<rfeng> can you check the axis2 artifacts in your local maven repo to see if
they are from m1 repo?
<_ant> ok, how do you tell?
<rfeng> reading the pom file
<rfeng> It seems that by default m1 artifacts are pulled in and no
transitive deps happen, then it works (meaning the axiom SNAPSHOT version
are not used)
<_ant> which artifacts, the axiom or axis2 ones?
<rfeng> axis2
<_ant> well currently the axis2 pom has    xmlns="
http://maven.apache.org/POM/4.0.0"
<_ant> but i'mnot sure if thats what i had when running the tests
<rfeng> oh
* dkulp has left #tuscany
<_ant> can i delete the m1 repos from tuscany pom and try to recreate?
<rfeng> I have this:
<rfeng> <project>
<rfeng>   <pomVersion>3</pomVersion>
<rfeng>   <groupId>org.apache.axis2</groupId>
<rfeng>   <artifactId>axis2-kernel</artifactId>
<rfeng>   <name>Apache Axis 2.0 - Kernel</name>
<rfeng>   <currentVersion>1.1</currentVersion>
<rfeng> so my version seems to be m1
<rfeng> can you try to run the helloworldws build with
addExtensionDependencies=true?
<rfeng> and see what jars are pulled in the WAR?
<_ant> ok
<_ant> you mean loadExtensionDependencies right?
<rfeng> y
<_ant> ok, I don't get the axiom snaphot in the war, just the 1.2 jars
<_ant> do you get the snapshot ones in the war? What do others see? I'm
worried I just have someconfig diff thing and leading us on a wild goose
chase
* kgoodson is now known as kgoodson_away
<rfeng> I didn't try the m2 artifacts yet with the war plugin
<rfeng> how did you get the axis2 maven2 artifacts, randomly?
<_ant> did nothing special so i guess randomly. let me delete them and see
where it gets them from...
<rfeng> ok
<rfeng> you can also check if the axiom-impl pom is m1 or m2
<rfeng> in the build, you may see warnings like invalid pom ...
<rfeng> if the pom is from m1
<_ant> Downloading:
http://repo1.maven.org/maven2/org/apache/axis2/axis2-kernel/1.1/axis2-kernel-1.1.jar
<_ant> Downloading:
http://repo1.maven.org/maven2/org/apache/ws/commons/axiom/axiom-api/1.2/axiom-api-1.2.jar
<_ant>   <modelVersion>4.0.0</modelVersion>
<_ant>   <artifactId>axiom-impl</artifactId>
<_ant>   <name>Axiom Impl</name>
<_ant>   <version>1.2</version>
<rfeng> hmm, you're getting everything from m2 repo
<rfeng> so it proves to be random :-(
<_ant> not so randon for me, AFAICT i've never had them actually used
<_ant> could anyone else try and see what they get?
<rfeng> I'm getting m1 artifacts, I can try again
<rfeng> It's interesting that I got the pom from the ws.zone (m1) and jar
from the m2 repo (repo1.maven.org/maven2)
<rfeng> ha, they removed the jars from
http://people.apache.org/repository/org.apache.axis2 but the poms are still
there
<rfeng> what a mess
<rfeng> so really it's a mixture of m1 and m2
* bertlamb_ has quit IRC (Read error: 110 (Connection timed out))
<_ant> but do you ever get the SNAPSHOT jars ending up in the WAR?
* jsdelfino has quit IRC (Read error: 110 (Connection timed out))
<rfeng> I'll try
<_ant> hows it going, have you or anyone been able to try this yet and get
axiom snapshot jars in the war?
* simonnash has quit IRC (Read error: 104 (Connection reset by peer))
<rfeng> _ant?
<_ant> hi
<rfeng> I tried
<rfeng> and it seems that luckily :-( due to a problem in the m2 pom  in
axis2 and it's still treated as an inavlid pom
<rfeng> so the transitive dependencies are not activated and our runtime
seems to be fine
<_ant> right, i've seen msgs about the invalid pom
<rfeng> here's the detail msg
<rfeng> [WARNING] POM for 'org.apache.axis2:axis2-kernel:pom:1.1:runtime' is
invalid. It
<rfeng>  will be ignored for artifact resolution. Reason: Failed to validate
POM
<rfeng> [DEBUG] Reason: Failed to validate POM
<rfeng> [DEBUG]
<rfeng> Validation Errors:
<rfeng> [DEBUG] 'dependencies.dependency.version' is missing for
org.apache.ws.commons.n
<rfeng> eethi:neethi
<rfeng> [DEBUG]
<rfeng> This section in the axis2-kernel pom is invalid
<rfeng>     <dependency>
<rfeng>       <groupId>org.apache.ws.commons.neethi</groupId>
<rfeng>       <artifactId>neethi</artifactId>
<rfeng>     </dependency>
<rfeng> in the parent pom, it defines org.apache.neethi in
<dependencyManagement>
<rfeng> net-net, axis2-kernel-1.1 has an invalid pom and transitive
dependency is not activated (which unfortunately good news to us)
<_ant> sounds good to me
<rfeng> the problem was fixed in axis2 trunk
<rfeng> let's summarize these and assume it's safe to get m2 out
<rfeng> do you want to send out the message?
<jboynes> is there any way the version can be inherited?
<rfeng> let me check the db-axiom
<_ant> why don't you so its clear its not just me who sees its working
<jboynes> e.g. if some other parent defines it in a <depManagement> element?
<rfeng> seems not
<rfeng> I'll do some final check and then send a msg saying the
SNAPSHOTversion of axiom-api won't be used at runtime
<_ant> cool. thx rfeng
* _ant is now known as ant_away