Hi Aled,
All good points.
1. Ideally, the PR queue should be emtpy.
2. Good question. However I assume there will be enough time to iron out
the snags before the next release. The best time to start this is
probably right after a release (0.7 or 0.8)
3. Would it be possible to have an (optional) migration catalogue/map
for entity names?
Actually your point #3 is interesting. I was talking to some guys a
while ago about using URIs (URN) or something instead of the confusing
class names, kinda like:
name: simple-appserver
location: localhost
services:
- type: brooklyn:.webapp:0.7.0:JBoss7Server
instead of:
- type: brooklyn.entity.webapp.jboss.JBoss7Server
... which would resolve to the appropriate class. That is introducing a
resolution mechanism.
Hadrian
On 07/14/2015 08:26 AM, Aled Sage wrote:
Hi, thanks for that.
I'd like us to treat this with caution for when we make the change. For
example:
1. All open PRs will likely be affected (i.e. potentially needing
significant rebasing).
2. Are our regression tests good enough, particularly with the impact
on persisted state.
Any improvements to do first?
3. What backwards compatibility guarantees do we want to give, e.g. of
fully qualified names used in YAML blueprints.
We should have a think about (2) and (3), to ensure we do the right prep
before the rename.
---
I hope that the refactor can be done relatively quickly in a modern IDE.
(And searching the *.yaml files). Assuming that is true, then it may be
worth doing a *dry run* and seeing what tests etc fail. Those could
either be improved in the existing master (i.e. before the rename) so
that they won't fail, or just note the required fixes so that we can do
it quickly.
We then may be able to plan ahead, to say that the rename will be done
on a particular day. We could aim to merge other PRs before that, and
everyone would know to expect major merge conflicts for anything we did
that particular day.
Does that seem feasible/sensible?!
---
I also propose we change to 1.0.0-SNAPSHOT in preparation for all of the
renames.
I'll propose that again in a separate e-mail thread for discussion, as
this one is already diverging from its original subject!
Aled
On 14/07/2015 12:59, Ciprian Ciubotariu wrote:
I can start working on the package rename on a fork of the master branch,
which can be rebased when integrating pending PRs. If we integrate
this branch
often back into master we can avoid blocking development.
On Tuesday 14 July 2015 10:07:30 Aled Sage wrote:
+1 for changing package name *after* the 0.7.0 release.
I propose:
* Release 0.7.0 with the existing package name
(so that users have a stable (not milestone) release that they can
use with confidence with minimal effort).
* Work towards a 1.0:
o clear any PR backlog
o change the package names in master
o document and thoroughly test the upgrade process from 0.7.0
to 1.0
(explicitly *not* support any other upgrade paths, such as
0.7.0-M2 to 1.0, etc)
o see what other features / fixes the community demand/want in a
1.0 release.
---
For version number, I also favour sticking with 0.7.0.
Of the two options for that:
1. backport changes to the 0.7.x branch, or
2. delete + recreate that to incorporate all code that is currently in
master, but reverting version number to 0.7.0...
I prefer the backport option, if that's not too hard technically.
Aled
On 14/07/2015 09:13, Richard Downer wrote:
On the subject of changing the Java package name (which I forgot to
mention) - yes this is something that will need to be done and we have
discussed this previously. The decisions was that the in-development
version of 0.7.0 would *not* have this change, but that it would be
done
for 0.8.0.
0.7.0 has been brewing for a long time and I don't think there's the
appetite to do this right now. It's a major change, it would break our
user's code (which would be unfair on those that have been using
0.7.0-M2)
and I've no doubt it would take weeks to fully shake out all the
bugs from
that kind of change. Meanwhile our users are clamouring for a 0.7.0
release as soon as we can :-)
Once we've got 0.7.0 released then we'll kick off 0.8.0 development
proper
and have a discussion about what changes we want to include.
Richard.
On Tue, 14 Jul 2015 at 09:03 Richard Downer <[email protected]> wrote:
Hadrian, thanks very much for your efforts on this release blocker.
I'm now back from vacation - I think that Sam is very busy this
week so
I'll put the Release Manager hat back on again and get an rc1 done
today.
Along with Svet I'm also in favour of option 1 - we have a release
branch
ready and we've already prepped some downstream customers to
prepare for
0.7.0-final, so I'm not keen to change the version. I'm assuming that
apart
from the 0.7.0/0.8.0 difference, your changes should backport
cleanly to
the release branch.
Should I be mistaken about the difficulty of this, then I will
change my
mind :-)
Thanks again Hadrian and everyone else who has been looking at this
problem.
Richard.
On Tue, 14 Jul 2015 at 08:58 Svetoslav Neykov <
[email protected]> wrote:
I am for 1). As far as I know there's already a release branch
with some
backports. Another option is:
3. Create a 0.7.0 release off of the current master by changing
back the
version.
Changing the package name should be done sooner rather than later,
but
requires *very* careful planning due to:
* The need to keep backwards compatibility for existing
persistence
stores. I'd prefer to have
https://github.com/apache/incubator-brooklyn/pull/738 <
https://github.com/apache/incubator-brooklyn/pull/738> merged before
attempting it.
* Existing 3rd party blueprints better still be usable after the
change, though this doesn't seem possible.
Svet.
On 14.07.2015 г., at 5:40, Hadrian Zbarcea <[email protected]>
wrote:
My proposed changes to unblock the release are now complete [1].
They
are done however on the master, which leaves us with one of the
following
options (if I am not missing something)
1. Backport to a 0.7.0-Mxxx branch
2. Abandon 0.7.0-Mxxx and release at 0.8.0
While either work, I somewhat prefer moving on with #2 as the
simplest
solution.
Another thought is that migrating packages from brooklyn.* to
org.apache.brooklyn.* will be necessary at some point. I suspect best
would
be to do it now before a release.
Thoughts?
Hadrian
[1] https://github.com/apache/incubator-brooklyn/pull/737
On 06/27/2015 09:09 PM, Hadrian Zbarcea wrote:
This is not an easy one and imho would require some community
choice
before implementing a solution.
1. To be able to release downstream-parent, it would have to
have the
proper configuration, specifically for the release and gpg maven
plugins, that comes actually from the org.apache:apache:17 parent.
2. Consequently, the downstream parent should have either
org.apache:apache:17 or even better
org.apache.brooklyn:brooklyn-parent
as a parent.
3. The downstream-parent is only used in the quickstart archetype.
There is questionable value in having a downstream-parent that
users
would have to change anyway if it caries the apache scp and release
configurations that wouldn't apply for a user's project.
The only 2 solutions I can think of are to:
a. Get rid of the downstream parent and move all the necessary
incantations in the quickstart archetype.
b. Transform the downstream-parent (and maybe come up with a better
name
for it) into a <scope>import</scope> pom [1].
I think this is a blocker for the release.
Thoughts?
Hadrian
[1]
https://maven.apache.org/guides/introduction/introduction-to-dependency->
>>> mechanism.html#Importing_Dependencies>>>
On 06/27/2015 04:05 AM, Andrea Turli wrote:
Thanks Hadrian,
I've also found this one while googling for another project
[1], so
either
Apache parent or nothing should fix the issue.
HTH,
Andrea
[1]:
http://central.sonatype.org/pages/apache-maven.html#deprecated-oss-paren
t
On Sat, 27 Jun 2015 at 05:58 Hadrian Zbarcea <[email protected]>
wrote:
First thing, the <parent> for the brooklyn-downstream-parent
should
not be:
<parent>
<groupId>org.sonatype.oss</groupId>
<artifactId>oss-parent</artifactId>
<version>9</version>
</parent>
but the apache parent ultimately. I think this should completely
resolve
the problem. It's a bit late here to test, I'll do it tomorrow.
Cheers,
Hadrian
On 06/26/2015 11:35 PM, Hadrian Zbarcea wrote:
I did try a dryRun myself and did encounter a problem with the
brooklyn-downstream-parent, but of a different nature
"'parent.relativePath' points at wrong local POM", but I suspect
there
more issues there. From my experience releasing other
projects, I
try to
first remove relevant branches from my local maven repo before
preparing
a release.
I will look at it during the weekend. Somebody should revert the
version
back from 0.8.0-SNAPSHOT though.
Cheers,
Hadrian
On 06/26/2015 04:53 PM, Richard Downer wrote:
So we got all the source code lined up today, and the release
branch
made.
Everything was going very promisingly until I tried to close
the
Nexus
repository to publish the artifacts and got a rule violation
error.
I'll have a look at fixing the problem and re-starting the
release
on
Monday (unfortunately I won't have any availability to look at
this
over
the weekend).
In the meantime if anyone is looking for something to do
over the
weekend,
the exact failure Nexus reported was:
Missing Signature:
'/org/apache/brooklyn/brooklyn-downstream-parent/0.7.0-incubating/brookl
yn-downstream-parent-0.7.0-incubating.pom.asc'>>>
does not exist for
'brooklyn-downstream-parent-0.7.0-incubating.pom'.
Everything else has a .pom.asc except downstream-parent so it
seems
there's
something special about this project.
Richard.
--
Cloudsoft Corporation Limited, Registered in Scotland No: SC349230.
Registered Office: 13 Dryden Place, Edinburgh, EH9 1RP
This e-mail message is confidential and for use by the addressee
only.
If
the message is received by anyone other than the addressee, please
return
the message to the sender by replying to it and then delete the
message
from your computer. Internet e-mails are not necessarily secure.
Cloudsoft
Corporation Limited does not accept responsibility for changes
made to
this
message after it was sent.
Whilst all reasonable care has been taken to avoid the
transmission of
viruses, it is the responsibility of the recipient to ensure that the
onward transmission, opening or use of this message and any
attachments
will not adversely affect its systems or data. No responsibility is
accepted by Cloudsoft Corporation Limited in this regard and the
recipient
should carry out such virus and other checks as it considers
appropriate.
