commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Oliver Heger <>
Subject Re: [all] Nightlies are back...mostly
Date Thu, 06 Jul 2006 20:06:04 GMT

thank you very much for your work! Finally I can see, which test fails 
for [configuration].

Hm, it's a test dealing with Applets. I guess, you run the tests in a 
non GUI environment, where creating an Applet is not allowed? I modified 
the test and hope that the problem is now solved.


Phil Steitz wrote:
> Nightlies are back up and running for most components.  I still think
> maven2 + continuum is the way to go long term, but to get the
> nightlies back up, I hacked together a little bash script based on
> Craig's and set it up to run as a crontab on
> It would be great if people could have a look at the nightlies
> themselves as well as the script, which I checked into commons-build
> as ""  I have no shame regarding my limited bash
> (or Java, actually ;-) skills, so feel free to make suggestions for
> improvement or submit patches.  Be careful checking in the script,
> though, as it is executed via a crontab wrapper that svn ups it before
> executing it on  You can test it locally by
> changing the config to make your local host the deploy host.  You have
> to gen an ssh key (if you don't already have one) and trust it locally
> to do this.
> The script reads lists of components from commons-build txt files like
> "nightly_sandbox_maven_list.txt" and executes maven clean dist for the
> ones in the maven lists and ant clean dist for those in the ant lists
> (after svn up).  I split them up based on whether or not I could get
> "maven clean dist" to work.  The ant lists consist of components for
> which the maven build did not work, but the ant build did. I have left
> a few things in the maven list even though they fail, because the
> reason seems trivial (e.g. clover license, or missing sun jars - I
> will fix the second).  It svn ups the lists before executing, so if
> you want to move components from ant -> maven, or vice-versa, just
> make the changes to the files and check them in.  Remember to delete
> as well as add; otherwise the script will run both.
> Logs for each component build are for the moment being written out to
> under dated directories.  I didn't want to clutter the normal nightly
> location with logs, but that could be changed if we like having them
> there.
> I ommitted dormant components.  If anyone feels strongly that we
> should be doing nightlies for any of these, the script can be patched
> to do that.
> There are no sigs or hashes.  Another easy patch, but since this all
> runs as me, I don't want to sign.  Adding hashes is no problem.  Some
> of the maven builds do this already.  Best would probably be to have
> the maven builds all do it and the nightly script just copy out.  I am
> OK either way, but the script needs to be patched in either case to
> scp the hashes.
> More warts:
> 1) The files are not group writable on the deploy host
> (  I set the umask in the script, but somehow the
> permissions are not getting set right.  Any pointers / patches on how
> to get the wrapper or script itself to do this correctly would be
> appreciated.
> 2) Ubuntu on vmbuild seems to have no zip command, so the ant builds
> are not getting zips generated.  I guess I could break down and fork
> Ant again to do the zips, but maybe there is an easier way?
> 3) While it would be easy to get jar:deploy done from within the maven
> loop, that seems like running with scissors.  Also, what would be
> deployed would be a SNAPSHOT that just over-wrote the previous (I
> think). This exposes another wart - though the names are munged to be
> dated, the internally bundled jars are not.  For anyone wanting to
> play with this, be forwarned that Ubuntu rename is picky - you need to
> use the perl regexp form like the script does.  Could be there is a
> simple maven way to get this right.
> Thanks in advance for feedback / patches / review of the output and
> help getting the failing builds restored.
> Phil
> ---------------------------------------------------------------------
> To unsubscribe, e-mail:
> For additional commands, e-mail:

To unsubscribe, e-mail:
For additional commands, e-mail:

View raw message