| [jakartaee-tck-dev] Tuning for running the TCKs against GlassFish... |
The [1] branch contains some TCK tuning changes that I worked on last year but never got around to testing much. It looks like [2] took 15 hours to run which sounds right for how long it takes to run two Full Platform TCK runs at the same time (an unrelated TCK run was also going for compiling with JDK11).
I'm not really sure if the com.sun.ts.tests.ejb30.lite.basic.singleton
failures are legitimate but assume they probably are. If the
failures are caused by the tuning changes [1], we could try
breaking the ejb30.lite.basic.singleton tests into multiple tests
groups (e.g. ejb30.lite.basic.singleton.common,
ejb30.lite.basic.singleton.concurrency,
ejb30.lite.basic.singleton.dependson, ejb30.lite.basic.singleton.lifecycle)
as we really shouldn't need a 10GB virtual machine (vm) to run
tests on (or more than one CPU core per vm since we are running
single user testing).
Scott
[1]
https://github.com/scottmarlow/jakartaee-tck/tree/tuning_request2
[2]
https://ci.eclipse.org/jakartaee-tck/job/jakartaee-tck-scottmarlow/job/tuning_request2/2/#showFailuresLink