On 7/20/2010 2:22 PM, Kelly O'Hair wrote:
The first two batches look fine to me.
Thanks!
The last one was pretty big, and I poked around a little, looks fine
to me,
but I am not sure I gave it a very good review.
I'll wait to see if one of the original Logger WeakRef leak reviewers
chimes in here (Alan, David or Eamonn...) I'm very confident of
the tests at this point; you've probably seen all the test JPRT
jobs that I've been sending through... :-)
I think getting these tests right may be more difficult than getting
the API to work. :^{
Hence the refactoring of the infrastructure stuff into common files.
I'm getting tired of fixing this infrastructure stuff in more than
one place...
So for risking your sanity and working on these tests, you get 3 gold
stars. ;^)
Thanks! There will be more fixes coming in the test arena...
Dan
-kto
On Jul 20, 2010, at 11:35 AM, Daniel D. Daugherty wrote:
Greetings,
I have fixes for three different test bugs ready to go for OpenJDK7.
I've run these tests manually on Linux, Solaris SPARC, Solaris X86
and WinXP/Cygwin. I've run these tests through JPRT several times to
make sure they are stable (which also covers WinXP/MKS). I've been
running select tests from this set in loops for 12-24 hours on Linux,
Solaris X86 and WinXP/Cygwin to make sure they don't hiccup.
This first fix I've had for a while and it has been used in my personal
baseline testing for the last several JDK7 promotions:
http://cr.openjdk.java.net/~dcubed/6941287-webrev/0/
Yes, it could be refactored to use some of the newer logic that I added
in the fix for 6964018, but that's a task that should be applied to all
the test/sun/tools/* tests.
This second fix was intended to add diagnostic information for tracking
down some of the strange intermittent failures on Windows:
http://cr.openjdk.java.net/~dcubed/6962804-webrev/0/
However, the fix seems to have solved the intermittent failures on
Windows problem. I haven't seen these failures reproduce in my personal
baseline testing on the last several JDK7 promotions and a stress loop
using JDK7-B98 bits hasn't reproduced the failures either. I suspect
that my refactoring of the logic that handles a missing NL at the end
of a file has fixed this problem. The scaffold no longer creates a
temporary file every time a test runs; it is only created when needed
to solve the missing NL problem.
The third fix gets my recent Logger WeakReference leak tests back into
the available test mix:
http://cr.openjdk.java.net/~dcubed/6964018-webrev/0/
This fix refactors and enhances some of the test infrastructure in
test/sun/tools/common/* and changes the new Logger WeakReference leak
tests to use that infrastructure. Other tests in test/sun/tools/*
also need minor tweaks to fit into the new infrastructure.
Thanks, in advance, for any reviews.
Dan
