1. the extra openjpa-xxx and 1.1.0 directories
I believe that the openjpa-parent directory is unnecessary; that was a
glitch in my copy-and-paste. However, all the rest looks just the same
to me as people.apache.org:/www/people.apache.org/repo/m2-ibiblio-
rsync-repository. Bear in mind that the root of the repository overlay
is ~pcl/release-candidates/1.1.0/repo/m2-ibiblio-rsync-repository, so
the release-candidates/1.1.0 bit is not part of what the repo would
look like.
2. the lack of jars and poms directories in the repository
Which directories are you expecting? Could you provide a pointer to
what you're comparing to? Looking at people.apache.org:/repo/m2-
ibiblio-rsync-repository/org/apache/openejb, the structure looks
pretty similar.
3. unnecessary .sha files for .jar and .pom
4. missing .asc files for .jar and .pom
When I added the code that Wendy provided, it generated .sha files and
did not generate .asc files. I've been operating under the assumption
that her code snippet is generating satisfactory output. Do we have
any clear guidance for what output is necessary from that code?
5. extra maven-metadata.xml files
Which are the extra ones? Looking at people.apache.org/www/
people.apache.org/repo/m2-ibiblio-rsync-repository/org/apache/openejb,
I see lots of maven-metadata.xml files there as well.
Thanks,
-Patrick
On May 16, 2008, at 12:06 PM, Craig L Russell wrote:
Hi Patrick,
Thanks for doing that. The results don't match my expectations for a
properly organized maven2 repository.
I'm not an expert in maven, but this structure doesn't look like the
one you're copying into.
Could someone from maven take a look? What looks fishy to me is
1. the extra openjpa-xxx and 1.1.0 directories
2. the lack of jars and poms directories in the repository
3. unnecessary .sha files for .jar and .pom
4. missing .asc files for .jar and .pom
5. extra maven-metadata.xml files
This doesn't require another vote because the artifacts are not at
issue; the packaging is. Once the packaging is straightened out
we're good.
Craig
On May 15, 2008, at 3:16 PM, Patrick Linskey wrote:
Done. You can browse the results at http://people.apache.org/~pcl/release-candidates/1.1.0/repo/m2-ibiblio-rsync-repository
, or look at it on people.apache.org at ~pcl/public_html/release-
candidates/1.1.0/repo/m2-ibiblio-rsync-repository.
-Patrick
On May 15, 2008, at 2:13 PM, Craig L Russell wrote:
Ok,
How about doing something like this:
$ find . \ -name m2-repository -not -path "*openjpa-project*" -
exec mvn -f /tmp/maven-stage-plugin/pom.xml stage:copy -
Dsource=file://{} -Dtarget=scp://people.apache.org/home/pcl/
release-candidate/1.0.0/repo/m2-ibiblio-rsync-repository \ -
Dversion=1.0.1 \;
Then we can look at what you will do after the vote completes and
argue about whether the artifacts are correct.
Craig
On May 15, 2008, at 12:20 PM, Patrick Linskey wrote:
What we need to vote on are the (reproducible) release
artifacts: the source and binary distributions with asc and
md5 sig/checksums; and the maven artifacts that will go into
the mirrored maven repository.
Sadly, our process does not seem to support generating such an
artifact.
Then how will the maven artifacts get published?
I do not understand what you mean by "reproducible";
Reinforcing that you should be able to create the same artifacts
by running the scripts again. No manual to-and-fro should be
involved.
From [1], the last step in the release process is:
$ find . \ -name m2-repository -not -path "*openjpa-project*" -
exec mvn -f /tmp/maven-stage-plugin/pom.xml stage:copy -
Dsource=file://{} -Dtarget=scp://people.apache.org/www/
people.apache.org/repo/m2-ibiblio-rsync-repository \ -
Dversion=1.0.1 \;
My understanding is that once that script is run, the contents
are uploaded to the non-snapshot mvn repository. I.e., if I were
to run that script, things would be released to the wild. That
sounds to me like something that should happen only once the
release is approved, and not before.
So, in other words, the semi-automatic process documented at [1]
takes the output that I've linked to and pushes it to the mvn
repo directly.
As I pointed out, I think that there is room for improving this
process. I would rather, however, cut out the mvn artifacts from
the 1.1.0 release than spend the next days rushing through trying
to learn about mvn's world view when it comes to further mvn
automation.
this was all generated in exactly the same manner as prior
(approved) OpenJPA releases, with the exception that we're now
generating md5 and asc files for more of the artifacts.
And we have had problems with subsequent maven artifact
postings, which I want to fix by having more oversight earlier.
It was my understanding that the only issue that we've
encountered in the previous releases was the lack of signed
artifacts. I believe that the work that I've done addresses that.
As I mentioned above, since our process deploys directly to the
mvn repo, I know of no way to validate that that's the case.
Further, I am not scared of directly (manually) updating the mvn
repo; I was under the impression that it was the artifacts, and
not the process, that needed voting / approval.
I see lots of m2-repository contents that would seem to conform
to what maven expects, but I'm no maven expert:
http://openjpa.apache.org/builds/1.1.0/apache-openjpa/m2-repository/org/apache/openjpa/apache-openjpa/1.1.0/
These directories are not proper maven repositories, so a subset
of these artifacts then need to be copied to completely
different directories. How will this be done? Why can't they be
copied "now" to a temporary maven repo so they can be vetted?
I have no idea what "proper maven repositories" look like.
How about this: let's get rid of all the mvn artifacts from the
1.1.0 release, and focus instead just on the binary and source
zips.
-Patrick
[1] http://openjpa.apache.org/releasing-openjpa.html
On May 15, 2008, at 12:04 PM, Craig L Russell wrote:
Hi Patrick,
On May 15, 2008, at 11:52 AM, Patrick Linskey wrote:
What we need to vote on are the (reproducible) release
artifacts: the source and binary distributions with asc and
md5 sig/checksums; and the maven artifacts that will go into
the mirrored maven repository.
Sadly, our process does not seem to support generating such an
artifact.
Then how will the maven artifacts get published?
I do not understand what you mean by "reproducible";
Reinforcing that you should be able to create the same artifacts
by running the scripts again. No manual to-and-fro should be
involved.
this was all generated in exactly the same manner as prior
(approved) OpenJPA releases, with the exception that we're now
generating md5 and asc files for more of the artifacts.
And we have had problems with subsequent maven artifact
postings, which I want to fix by having more oversight earlier.
I see lots of m2-repository contents that would seem to conform
to what maven expects, but I'm no maven expert:
http://openjpa.apache.org/builds/1.1.0/apache-openjpa/m2-repository/org/apache/openjpa/apache-openjpa/1.1.0/
These directories are not proper maven repositories, so a subset
of these artifacts then need to be copied to completely
different directories. How will this be done? Why can't they be
copied "now" to a temporary maven repo so they can be vetted?
Craig
http://openjpa.apache.org/builds/1.1.0/openjpa-examples/m2-repository/org/apache/openjpa/openjpa-examples/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-jdbc/m2-repository/org/apache/openjpa/openjpa-jdbc/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-kernel/m2-repository/org/apache/openjpa/openjpa-kernel/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-lib/m2-repository/org/apache/openjpa/openjpa-lib/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-persistence-jdbc/m2-repository/org/apache/openjpa/openjpa-jdbc/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-persistence/m2-repository/org/apache/openjpa/openjpa-persistence/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-slice/m2-repository/org/apache/openjpa/openjpa-slice/1.1.0/
http://openjpa.apache.org/builds/1.1.0/openjpa-xmlstore/m2-repository/org/apache/openjpa/openjpa-xmlstore/1.1.0/
Does this help?
-Patrick
On May 15, 2008, at 10:32 AM, Craig L Russell wrote:
Hi Patrick,
On May 15, 2008, at 10:10 AM, Patrick Linskey wrote:
Hi,
Nice job, but need a bit more to review.
Thanks! I'm assuming that you mean that you can't vote +1
now, but that you'll have time between now and Monday
evening, right?
Right.
There's a requirement that all artifacts have an md5
checksum in addition to the asc signature.
Those should all be generated. In fact, it looks like we even
have md5 checksums of the signatures!
In the link you sent out
A candidate build for OpenJPA 1.1.0 is available at:
http://openjpa.apache.org/builds/1.1.0/downloads/
there are only the source and binary distribution files with
asc sigs, no md5 and no maven artifacts.
Since we plan to put the jar files into the global maven
repo, the artifacts should be on the download site for
review. The jar files also need md5 and asc signatures.
Everything can be found at people.apache.org/www/
openjpa.apache.org/builds/1.1.0.
Everything including the kitchen sink. But there are many
files and plain junk in that location that it's impossible for
me to review.
Also, I think it's worth noting that there are definite
improvements that we could make in our staging / delivery
process. I'd appreciate it if we didn't hold 1.1.0 hostage
for those changes. My understanding from Wendy's comments
about the last release process was that we needed more
signatures; I believe that the new release is sufficiently-
signed. If there are other improvements that we could make,
I'm all for making them, but would rather see non-
showstopping issues get logged and addressed in 1.1.1 etc.
The problem is that with all the stuff in the builds/1.1.0
directory it's not possible to make sense of it. Browsing the
builds/1.1.0, there are several things that look like maven
artifacts but they're in the wrong place, e.g. openjpa-jdbc/m2-
repository/org/apache/openjpa/openjpa-jdbc/1.1.0 which has an
extra 1.1.0 directory and has extra files e.g. -rw-rw-r-- 1
pcl openjpa 32 May 14 22:36 openjpa-
jdbc-1.1.0.jar.asc.md5 (generally it's not required to
checksum a signature;-).
So there's a non-trivial step between the artifacts that
you've pointed to and the actual release that we propose to
mirror to the world. And there's no PMC oversight possible for
that non-trivial step.
What we need to vote on are the (reproducible) release
artifacts: the source and binary distributions with asc and
md5 sig/checksums; and the maven artifacts that will go into
the mirrored maven repository.
Craig
-Patrick
On May 15, 2008, at 9:52 AM, Craig L Russell wrote:
Hi Patrick,
Nice job, but need a bit more to review.
There's a requirement that all artifacts have an md5
checksum in addition to the asc signature.
Since we plan to put the jar files into the global maven
repo, the artifacts should be on the download site for
review. The jar files also need md5 and asc signatures.
Craig
On May 14, 2008, at 11:37 PM, Patrick Linskey wrote:
OpenJPA Developers-
A candidate build for OpenJPA 1.1.0 is available at:
http://openjpa.apache.org/builds/1.1.0/downloads/
Please review these artifacts and signatures, and vote
whether we
should release them as Apache OpenJPA version 1.1.0.
Release notes
for this release are included in the artifact, or can be
browsed at:
http://svn.apache.org/repos/asf/openjpa/branches/1.1.0/openjpa-project/RELEASE-NOTES.html
The Apache Release Audit Tool has been run on the release,
and no
missing licenses were found with the exceptions listed in the
exclusion section of the "rat-maven-plugin" configuration
in http://
svn.apache.org/repos/asf/openjpa/branches/1.1.0/pom.xml .
In accordance with http://www.apache.org/foundation/
voting.html#ReleaseVotes , three +1 votes will be
sufficient to
approve the release for publication. While it is not
possible to veto
a release, the vote will remain open for the standard 3 day
period
(ending at 11:30pm Pacific on Monday 5/19) in order to
allow people to
thoroughly review the release and perform whatever
additional testing
they desire and raise any concerns or objections.
A vote of "+1" means you approve of the release for
publication, "-1"
means you do not approve, and a "+0" or "-0" means you are
neutral.
Thanks in advance for your diligence in helping to ensure
that the
quality of the OpenJPA 1.1.0 release reflects the high
quality of all
of its contributors!
-Patrick
--
Patrick Linskey
202 669 5907
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!
--
Patrick Linskey
202 669 5907
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!
--
Patrick Linskey
202 669 5907
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!
--
Patrick Linskey
202 669 5907
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!
--
Patrick Linskey
202 669 5907
Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!
--
Patrick Linskey
202 669 5907
