CC: sent to serviceability-dev for general distribution
Hi Kumar:
Thanks for looking into these issues.
The cause for 6758616 failure is the Serial UID has been changes,
this is not a launcher issue, I also verified by copying the
TL's rt.jar into a stock b37 promoted build (which has the old launcher and
passed the test) and now the test fails.
I have updated the bug with the following comments:
"As the error suggests the schema of the class has changed, the test seems
to be trying to redefine the class java/lang/InterruptedException.
This appears to be the case with the fix for 6749308, which has
changed the SerialUID of various exceptions and errors, and
java/lang/InterruptedException is one of them, and a CCC
request has been filed http://ccc.sfbay/6749308, therefore it appears
that the test needs to be readjusted."
So 6758616 is a test bug, correct?
Tim: I filed a new bug 6758881 for the NoClassDefFoundError exception of
the
launcher, I think its best to fix this before we push, I will complete
the code changes
test and send it out to review tmr, I should be able to push by
tomorrow PM.
Please let me know how I should go about it.
Thanks - For 6758881, once you have your reviews, send me a pointer
to your webrev. I can pull the changes and create a changeset on
your behalf in my merge workspace. After that I will respin builds
and make them available for testing and verification.
Tim
Thanks
Kumar
btw, it would be good to have new bits with fixed NoClassDefFoundError
so we can test it.
but probably after you look at 6758616 ...
Ekaterina Pavlova wrote:
Thanks Kumar.
I have filed
6758616 Redefinition of InterruptedException failed with
JVMTI_ERROR_UNSUPPORTED_REDEFINITION_SCHEMA_CHANGED
java/java/serviceability
I am not sure the category is proper because I am not sure what
changes caused this regression.
So, feel free to change the category if needed.
-katya
Kumar Srinivasan wrote:
Katya,
Ok I have a fix for the Exception ie, it will throw the
NoClassDefFoundError,
instead of JavaRuntimeException.
After making this change, ran runnsk (from the location below), I
see only the
attach002 unexpected failure. If you can file a bug, I will get this
reviewed and
work with Tim to get this into the push repo.
Thanks
Kumar
% $ALT_OUTPUTDIR/bin/java Foo
Error: Could not find main class Foo
Exception in thread "main" java.lang.NoClassDefFoundError: Foo
at
sun.launcher.LauncherHelper.checkAndLoadMain(LauncherHelper.java:193)
Caused by: java.lang.ClassNotFoundException: Foo
at java.net.URLClassLoader$1.run(URLClassLoader.java:220)
at java.net.URLClassLoader$1.run(URLClassLoader.java:209)
at java.security.AccessController.doPrivileged(Native Method)
at java.net.URLClassLoader.findClass(URLClassLoader.java:208)
at java.lang.ClassLoader.loadClass(ClassLoader.java:325)
at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:294)
at java.lang.ClassLoader.loadClass(ClassLoader.java:270)
at
sun.launcher.LauncherHelper.checkAndLoadMain(LauncherHelper.java:190)
btw: are these tests that you are running the nsk tests ?
If so, is this accurate ? /java/svc/tools/bin/runnsk
The reason I ask is there are more launcher changes coming up.
I would like to catch issues earlier on in the development cycle.
Thanks
Kumar
Tim,
if we have time I would prefer to postpone the integration till
next week.
Is it ok?
-katya
Tim Bell wrote:
Kumar Srinivasan wrote:
I am not sure why the RedefineClasses test is failing.
If the expected NoClassDefFoundError is required
i can look into throwing that exception. I don't think
that warrants a no-go.
Katya - are you OK with this explanation?
Will you be able to issue a Serviceability PIT certificate today
(Friday)?
Or, do you feel this is a stopper, in which case we can wait and
iron out the issues next week.
Please let me know what to expect so I can plan the rest of my day-
Thanks-
Tim
Thanks
Kumar
<including Dan in case he has to look at the RedefineClasses
test failing>
But Katya has what looks like two valid tests failing:
I am seeing strange failures in jdk7b38 TL pit.
One of them is that java doesn't throw "NoClassDefFoundError"
when trying to load non existing class.
For example:
>
/net/sqenfs-1.sfbay/export1/comp/vm/jdk/7/pit/b38/solaris-sparc/bin/java
ops
Error: Could not find main class ops
Exception in thread "main" java.lang.RuntimeException: Could
not find the main class ops
at
sun.launcher.LauncherHelper.checkAndLoadMain(LauncherHelper.java:193)
Before it was:
>
/java/re/jdk/7/promoted/all/b37/binaries/solaris-sparc/bin/java
ops
Exception in thread "main" java.lang.NoClassDefFoundError: ops
Caused by: java.lang.ClassNotFoundException: ops
at java.net.URLClassLoader$1.run(URLClassLoader.java:220)
at java.net.URLClassLoader$1.run(URLClassLoader.java:209)
at java.security.AccessController.doPrivileged(Native
Method)
at
java.net.URLClassLoader.findClass(URLClassLoader.java:208)
at java.lang.ClassLoader.loadClass(ClassLoader.java:325)
at
sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:294)
at java.lang.ClassLoader.loadClass(ClassLoader.java:270)
at
java.lang.ClassLoader.loadClassInternal(ClassLoader.java:338)
Error: Could not find the main class ops.
and
one more failure is in nsk/jvmti/AttachOnDemand/attach002a.
It fails on all platforms with
# error occured while redefining
JVMTI_ERROR_UNSUPPORTED_REDEFINITION_SCHEMA_CHANGED
# ERROR: attach002aAgent00.c, 48:
nsk_jvmti_aod_redefineClass(options, jvmti, object_class,
REDEFINED_CLASS_FILE_NAME)
See full log here:
http://sqeweb.sfbay.sun.com/nfs/comp/vm/misc/results/vm/gtee/JDK7/PIT/CORE/SERVICEABILITY/7/b38/vm/solaris-amd64/server/mixed/solaris-amd64_server_mixed_nsk.jvmti.testlist/ResultDir/attach002a/attach002a.log
The test passes with jdk7-b37.
I am going to file new bugs but I am also thinking if we
should give no-go for this PIT ...
So it looks like something is wrong with your change.
- jjh
Kumar Srinivasan wrote:
Hi Jim. et. al.,
Yes from a time line the applications main-class
is loader a bit earlier than before, with one JNI upcall vs.
3-4.
The class loading is slightly different with the new launcher.
The big improvements are the application class loading is
pushed into
java and and the boot classes need by the launcher are fetched
from the VMs BCP (katya, this is related to the other HotSpot fix
which you refered to earlier this week) loader directly vs.
going through
the application class loader, this improves startup performance.
Pushing the application class loading to java gives us better
control of diagnosing user problems and reporting the issues
with user friendly and localized messages. A classic example
is the one cited below, the new message is neat and clean vs. a
terse and unfriendly message scary message. Also its considerably
less native code in the launcher (which is a "good thing!").
Thanks
Kumar
Ekaterina Pavlova wrote:
may be these 2 bugs caused the changes ...
6742159 java tools (launcher) improve the java
launching mechanism
6755847 java tools (launcher) will trigger
assertions in debug buil
These do look like likely candidates. Kumar, what do you think?
- jjh
Jim Holmlund wrote:
I wonder if this is related to that issue I sent earlier:
Per previous email, this change causes VM test
nsk/jdi/ThreadReference/forceEarlyReturn/forceEarlyReturn002
to fail. It appears to me that the test makes an invalid
assumption about when a class is loaded, and with this new
TL, the class is loaded at a point earlier than the point
at which it is loaded in previous JDK builds/releases. If
this is true, then the new TL isn't wrong, just
different. But I wonder if this difference could have any
other negative impacts?
- jjh
'this change' means the TL pit. I looked thru the list
of CRs below and doen't see a likely culprit.
Chris, what about:
6749308 java.io, java.lang, java.util exception
classes don't specify serialVersionUID
- jjh
Ekaterina Pavlova wrote:
thanks Jim.
Btw, Jim, Tim,
I am seeing strange failures in jdk7b38 TL pit.
One of them is that java doesn't throw
"NoClassDefFoundError" when trying to load non existing
class.
For example:
>
/net/sqenfs-1.sfbay/export1/comp/vm/jdk/7/pit/b38/solaris-sparc/bin/java
ops
Error: Could not find main class ops
Exception in thread "main" java.lang.RuntimeException:
Could not find the main class ops
at
sun.launcher.LauncherHelper.checkAndLoadMain(LauncherHelper.java:193)
Before it was:
>
/java/re/jdk/7/promoted/all/b37/binaries/solaris-sparc/bin/java
ops
Exception in thread "main" java.lang.NoClassDefFoundError:
ops
Caused by: java.lang.ClassNotFoundException: ops
at
java.net.URLClassLoader$1.run(URLClassLoader.java:220)
at
java.net.URLClassLoader$1.run(URLClassLoader.java:209)
at
java.security.AccessController.doPrivileged(Native Method)
at
java.net.URLClassLoader.findClass(URLClassLoader.java:208)
at
java.lang.ClassLoader.loadClass(ClassLoader.java:325)
at
sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:294)
at
java.lang.ClassLoader.loadClass(ClassLoader.java:270)
at
java.lang.ClassLoader.loadClassInternal(ClassLoader.java:338)
Error: Could not find the main class ops.
something is wrong with this build.
Do you know what changes result in such behavior?
Jim Holmlund wrote:
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/jdk/test/com/sun/jdi/SimulResumerTest.java
I added that to the CR also. Sorry about that.
- jjh
Ekaterina Pavlova wrote:
Hi Jim,
what tests could be used to verify:
6751643 java debugger
ThreadReference.ownedMonitors() can return null
thanks,
-katya
Tim Bell wrote:
Webrevs: http://lonepeak.sfbay/PIT/b38/Agones-MASTER
PIT builds done by JPRT job ID
2008-10-07-051427.tbell.Agones
executable bits (nightly build number 281)
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/linux-i586
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/linux-x64
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/solaris-i586
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/solaris-sparc
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/solaris-sparcv9
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/solaris-x64
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/windows-i586
/net/bonsai.sfbay/w/builds/jdk/7/pit/b38/windows-x64
Tests:
Under /net/bonsai.sfbay/w/builds/jdk/7/pit/b38/
jdk/test
and
langtools/test
bundles:
are in
/net/jprt-web.sfbay/jprt/archive/2008/10/2008-10-07-051427.tbell.Agones/bundles
linux_i586_2.4-product.zip
linux_x64_2.4-product.zip
solaris_i586_5.10-product.zip
solaris_sparc_5.10-product.zip
solaris_sparcv9_5.10-product.zip
solaris_x64_5.10-product.zip
windows_i586_5-product.zip
windows_x64_5.2-product.zip
Please issue PIT certificates (or PIT waivers) for the
following areas
by noon (Pacific) on Friday the 10th of October (or
sooner if
possible). Our putback slot is PM on Friday. Please CC:
[EMAIL PROTECTED] when sending certificates.
Security: YES
Networking: YES
Core Libraries: YES
Tools: YES
JMX: YES
jconsole: NO
Serviceability: YES
security/pit and networking/docs/pit/ jobs are in the
process of
submitting themselves now.
In this round, we have:
6749308 java classes_lang java.io, java.lang,
java.util exception classes
6621697 java classes_net Problem with file
test/sun/net/www/http/ChunkedI
6739920 java classes_net java 6u4~ use larger C heap
if there are many th
6746836 java classes_net java.net exception classes
don't specify serialV
6750364 java classes_net Error in test for 6744329
6752807 java classes_net java.net test
ADatagramSocket.sh is broken
6469266 java classes_secu Integrate Apache XMLSec
1.4.2 into JDK 7
6420151 java compiler need to improve byfile
compile policy to elimina
6500343 java compiler compiler generates bad code
when translating con
6735840 java compiler Need a type/symbol visitor
printer
6739427 java compiler -Xlint:processing not
recognized as an option
6747671 java compiler -Xlint:rawtypes
6753718 java compiler raw type warnings should be
suprresed during a l
6751643 java debugger
ThreadReference.ownedMonitors() can return null
6748546 java javadoctool javadoc API should be
classloader-friendly
6748601 java javadoctool javadoc API should allow
varargs use
6581243 java sunserviceta Service Tag and Product
Registration Support in
4459600 java tools java -jar fails to run
Main-Class if classname f
6742159 java tools (launcher) improve the java
launching mechanism
6755847 java tools (launcher) will trigger
assertions in debug buil
6750935 jmx classes The expected
NotCompliantMBeanException is not t
6751872 jmx doc MXBeanMappingFactory example
says "implements" w
