On 11/08/2011 00:03, Steven Dolg wrote:
Oops!... I did it again
But this time I'm innocent :P
It took me some time to figure this one out since I didn't touch
anything near it.
My theory is this:
With COCOON3-69 ehcache was upgraded from 1.6.1 to 2.4.3.
Ehm, yes, I am actually guilty in this respect: I did this version
upgrade, mainly to solve some dependency issues related to SLF4J,
Commons Logging and Log4j.
Between those versions, the serialVersionUID of net.sf.ehcache.Element
changed, so serialized versions of that class cannot be loaded any more.
The build fails because the ehcache for cocoon-profiling is configured
to be "diskPersistent" and reading the old cache with the new class fails.
The build fails only sometimes, because it fails only once on each
node in the build cluster, after that the cache is rewritten with the
newer version.
This is coherent with what I've found while building C3 on different
machines: from a fresh checkout no errors, from a source tree already
built before updating, first an error, then everything works.
Should we wait until the build has spread through all nodes and
stopped failing from this or try something else?
I'm really for changing the profiling cache to non-persistent - unless
that breaks something.
Do we need to preserve profiling results between restarts?
AFAIK, there should be no issue in making profiling results volatile,
but I am really not familiar with profiling internals: who knows better?
Regards.
-------- Original-Nachricht --------
Betreff: Cocoon-trunk - Build # 49 - Unstable
Datum: Wed, 10 Aug 2011 21:22:17 +0000 (UTC)
Von: Apache Jenkins Server <jenk...@builds.apache.org>
An: steven.d...@indoqa.com
The Apache Jenkins build system has built Cocoon-trunk (build #49)
Status: Unstable
Check console output athttps://builds.apache.org/job/Cocoon-trunk/49/ to view
the results.
--
Francesco Chicchiriccò
Apache Cocoon Committer and PMC Member
http://people.apache.org/~ilgrosso/
