OK, so this has to be CVCS (svn and so on) specific?
I didn't check if the release-plugin commits + push each time. If this is
so, it should be fixed so that this cannot happen.

I would say that though this is a valid technical concern, it's very
unlikely to happen in real life.

Let's take an example with a well known CI server, and specifically with
svn:

* You start a release. At the end of the release:prepare, the
release-plugin commits the poms just before copying the path to the tags
directory, then reverting the poms to the next snapshot development version.
* Unfortunately, the CI kicks in just there, just after that commit, and
before the trunk comes back to snapshot.
* In Jenkins (and Hudson), there's a default delay of some seconds,
historically to support VCS like CVS that wouldn't commit files in an
atomic manner
* So, Jenkins is gonna wait those seconds before actually starting the
build. Then it asks the VCS once more if a new change occurred. So, the
probability to encounter this issue is even lower...

So, all in all, if that's the issue you're talking about. Not sure it's
worth to worry a lot :-).

Btw, if you use a repo manager with a staging area, then at worse you're
gonna create a staging repo from the CI build in addition to your release
build. So, in the end you won't be able to push many releases under the
same names and your release should still be able to proceed.

HTH



2014-02-15 16:51 GMT+01:00 Thomas Broyer <[email protected]>:

> On Sat, Feb 15, 2014 at 4:18 PM, Baptiste Mathus <[email protected]>
> wrote:
>
> > Hi Dan,
> > Not sure what you mean. You say "CI", are you taking about a specific
> > server? If Jenkins for example, wouldn't it then be more a Jenkins user
> ml
> > question?
> >
> > And I don't see how a snapshot build could interfere with another build
> > with a release version.
> > Could you give details about your issue so that we can help you?
> >
>
> mvn release:prepare does a first commit with the POM modified with the
> no-SNAPSHOT version, then another commit with the POM modified with the
> "next" SNAPSHOT version.
> If a build is triggered for the first commit and basically does a "mvn
> install" or, worse, a "mvn deploy", then it'll end up deploying your
> release version, the very same that "mvn release:perform" will deploy too
> (very same version information, not necessarily the same artifact: might
> not deploy javadoc and sources for example, or the artifact might be
> slightly different because you have plugins in a profile triggered only on,
> or never on, release builds).
> I think that what Dan referred to as a "CI snapshot build" is the kind of
> build triggered by a commit (which generally builds snapshots).
>
>
> >
> > Cheers
> > Le 15 févr. 2014 07:01, "Dan Tran" <[email protected]> a écrit :
> >
> > > Hello
> > >
> > >
> > > It is possible that while release:prepare cutting the tag and CI for
> > > snapshot build wakeup at the same time.  Is there a way to prevent
> this?
> > > like a a profile to fail the build if the version happen to be a
> release
> > > version.  Ie is there a way to detect this in a profile?
> > >
> > > I currently have to keep remind my self to turn off CI snapshot build
> > while
> > > release is in progress, and too many to remember.
> > >
> > > Advice is greatly appreciated
> > >
> > > -Dan
> > >
> >
>
>
>
> --
> Thomas Broyer
> /tɔ.ma.bʁwa.je/ <http://xn--nna.ma.xn--bwa-xxb.je/> <
> http://xn--nna.ma.xn--bwa-xxb.je/>
>



-- 
Baptiste <Batmat> MATHUS - http://batmat.net
Sauvez un arbre,
Mangez un castor !

Reply via email to