This seems like a serious problem: something is really broken with the build?
I setup a jenkins job to run this in a loop against the 4.6 branch.
On Tue, Nov 12, 2013 at 11:19 PM, Apache Jenkins Server
jenk...@builds.apache.org wrote:
Build:
Robert,
I chatted with Uwe on IRC about it earlier today - this is a permgen OOM - he
thinks the issue is that Groovy is being loaded multiple times in order to
define the lucene.javadoc.url - the LUCENE-5322 changes I backported to
branch_4x yesterday are the cause.
I did not backport
oh i see. you already have a patch or you need help?
Ill keep the smoketest-loop against 4.6 running just in case, because
this jenkins job runs rarely at builds.apache.org but is really useful
at times like this :)
On Wed, Nov 13, 2013 at 12:46 PM, Steve Rowe sar...@gmail.com wrote:
Robert,
I have a patch I’m testing now, thanks.
I agree, it is useful to have more testing around a release.
Looks like you should be running your smoketest-looper with java 6 instead of
java 7 - build #3 is complaining about jar metadata.
Steve
On Nov 13, 2013, at 12:49 PM, Robert Muir
I set JAVA_HOME to a 1.6 compiler, but this still trips: I actually
think something is seriously broken with the build?
I set this up because Simon was trying to make a RC manually and
pinged me about wrong compiler being used, OOMs, etc.
[lucene-solr-46-smoker] $ ant
This isn't how its supposed to work though :)
JAVA_HOME should work. I'm 99% positive if i switch my exact same
jenkins task to a previous release branch it will pass.
On Wed, Nov 13, 2013 at 1:07 PM, Steve Rowe sar...@gmail.com wrote:
I don’t think it’s JAVA_HOME that’s consulted to produce
I don’t think it’s JAVA_HOME that’s consulted to produce the fake release, but
rather the first javac on the PATH. (That’s true for a real release too of
course.)
So you should do
$ JAVA_HOME=/usr/local/jdk1.6.0_37 PATH=$JAVA_HOME/bin:$PATH ant
-DJAVA7_HOME=/usr/local/jdk1.7.0_25
I can't find anywhere in the build that looks up JAVA_HOME (other than
smokeTestRelease.py) - through what mechanism should a JAVA_HOME sysprop affect
which javac and javadoc binaries are used to build a release?
On Nov 13, 2013, at 1:11 PM, Robert Muir rcm...@gmail.com wrote:
This isn't how
JAVA_HOME still works: its what the ant 'javac' target uses.
My problem was related to jenkins only, and I think Uwe already logged
in and fixed it. Sorry for the noise.
On Wed, Nov 13, 2013 at 2:03 PM, Steve Rowe sar...@gmail.com wrote:
I can't find anywhere in the build that looks up