6962419: TEST_BUG: java_io tests fails in samevm mode
kelly.ohair at oracle.com
Fri Jun 18 08:23:30 PDT 2010
Looks great to me.
Thanks to everyone for fixing tests like this.
People should keep in mind that some of the test batches in the jdk/
do NOT run in samevm mode at all, e.g. jdk_awt, jdk_beans2,
jdk_management2, jdk_nio2, jdk_nio3, jdk_rmi, jdk_security2,
and jdk_tools2. When I tried to run them in samevm mode, there were
too many problems,
so I gave up on the entire batch. Although some of the batches may
make sense to be entirely
othervm tests, like jdk_tools2. But eventually, I think it's a good
idea to mark tests
that need a dedicated VM "othervm", making it explicit.
If you look at the jdk/test/Makefile, you will see uses of
my goal was to run everything "RunSamevmBatch", someday.
(FYI... the jdk_awt, jdk_rmi, and jdk_swing tests really need a
dedicated X11 display
and are not run on a regular basis yet).
Just an aside comment.
I really hate import wildcards, seems so... inexact.
I noticed that there are GPL utilities out there that could clean
these up for us,
making the import list exact.
One I found at http://www.javafaq.nu/java-article914.html, and I know
NetBeans and IDEs
have 'Fix Imports' options, I use them all the time. But if this could
batched... seems like something to think about.
Maybe that would be a good summer intern job of figuring out how to
make the imports
explicit, and also a scheme to maintain it?
What do you think? Is it worth it?
Maybe create a database too so people could run all the tests using a
or method? Just ideas...
On Jun 18, 2010, at 7:55 AM, Alan Bateman wrote:
> This is another batch of updates to the regression tests to allow
> them be run in jtreg samevm mode. This batch covers the java.io
> tests. As expected, most of the issues are simply tests leaving
> files open, causing problems for subsequent tests running in the
> same VM. There are also several other random issues such as tests
> assuming the working directory, one test assuming it will be defined
> by the system class loader, and an issue cleaning up after a test
> that leaves a hidden file.
> The webrev with the changes is here:
> With these changes (plus yesterday's batch to fix the serialization
> tests) then all 305 java.io tests pass, for me, in both samevm and
> othervm modes on all platforms.
More information about the core-libs-dev