I had an email from last week responding to your first message that I
didn't hit send on. it said :
I like it! Seems like it does no harm, too :)
I've been trying to collect the various examples/tests that's we've been
getting for DRLVM in src/test. I assume now we can convert those over?
I think this solved HARMONY-1048 too :)
More inline :
Alexander Kleymenov wrote:
Geir,
I'm responding to your comment on HARMONY-2283 report in this thread.
On 11/26/06, Geir Magnusson Jr (JIRA) <[EMAIL PROTECTED]> wrote:
[
http://issues.apache.org/jira/browse/HARMONY-2283?page=comments#action_12453441
]
Geir Magnusson Jr commented on HARMONY-2283:
--------------------------------------------
I'm playing with it. Things I want to do :
1) make it so that the jasmin jar is optional, or make it so that it's
fetched automatically, dependably
Now it is implemented using common fetching mechanism: jasmin.jar is
downloaded with other dependencies during executing of "update"
target. So before running of "reg.test" target "ant update" target
should be launched.
That wasn't my experience with the patch - I had to provide a location
for it so the jar could be fetched, which meant that there was no auto
way of doing it.
IOW, as far as I understand it now, if I committed the patch as is, it
would break the build.
I've modified things so that
a) JASMIN isn't a dependency in setup.xml. I was trying to create a
parallal "optional dependency" list and re-factor the ant code, but it's
yecchy.
b) reg.test isn't automatically run - IOW, people will have to run
manually for now.
So now, I feel comfortable that I can commit w/o breakage, and we can
move forward from this point.
2) when running, h000 fails :
run-test:
[junit] Running
org.apache.harmony.drlvm.tests.regression.h0000.DirectByteBufferTest
[junit] HMYEXEL062E Internal VM error: Failed to create Java VM
[junit] FAILED to invoke JVM.
Haven't looked into it yet.
This failure is expected - DRLVM can not find native library by path
specified by java.library.path system property (while RI can).
LOL. How do we fix?
geir
Another problem with this test is using of JUnit test format for DRLVM
testing. Now the test fails during JVM invocation but it is not marked
as failed because JVM return 0 result code (as and for successful
JUnit test execution). What can be proposed for such a test is a new
test format returning some predefined status code in case of
successful execution and another predefined value in case of expected
failure. The resulting code different from predefined can be
considered as unexpected failure.
Thank you,
Alexander
