Thanks Alan. Other failures haven't shown up thus far so for now we'll just
address these ones (it improves the pass rate somewhat :) ).
Could I get a second review from someone in serviceability?
Thanks,
David
On 19/09/2011 6:57 PM, Alan Bateman wrote:
David Holmes wrote:
This a change to a bunch of serviceability tests (shell scripts that
launch the various j* tools (jps, jstatd, jstack etc)) that I'd like to
push through the TL JDK repo.
The changes were done by Carlos Lucasius but I'm acting as his "sponsor"
for getting these pushed.
webrev: http://cr.openjdk.java.net/~dholmes/7012206/webrev/
Summary: for correct operation the tools and/or the VM they target must be
running with UsePerfData enabled. This VM option is enabled in Java SE by
default, but is disabled in Java SE Embedded by default. To allow the
tests to be used regardless of the UsePerfdata setting they are augmented
to explicitly turn it on.
There has been some prior internal debate around how "best" to deal with
this issue and the resulting changes, while somewhat repetitive, are the
simplest approach to take.
There is one test - jps/jps-Vvml_2.sh - that can not pass with such a fix
because it is actually trying to test the jps output when no arguments (VM
or application) are passed to the target VM. So for that test I've just
added a comment.
This one reminds me that we need to go over all our shell tests so that they
pass $TESTVMOPTS through to all VMs that they create. Otherwise we aren't
always testing what we think we are testing.
The changes in the webrev look fine to me but I would have expected to see
tests other than the jvmstat stats, for example the tests in
com/sun/tools/attach and sun/management/jmxremote/bootstrap.
-Alan.
