Andrew John Hughes wrote:
2009/12/7 Kelly O'Hair <kelly.oh...@sun.com>:
OpenJDK6 changes for JDI. I'm ready to push.
I would like approvals from Joe, Jim, Alan, Dan, and Tim.
In the process of trying to get good test results running the openjdk6
debugger (JDI) and jvmti tests (com/sun/jdi and demo/jvmti), I kept running
into fixes made in jdk7 that were not in jdk6.
Since this functionality and apis are still jdk6, I took the liberty
of porting all of the openjdk7 JDI changes and JDI test changes to openjdk6.
These changes have all been reviewed for jdk7, and with a few minor
exceptions (i.e. -target 7 -source -7 in a few places) the resulting
files in openjdk6 now match openjdk7.
Please speak up if you have objections to this.
The openjdk6 com/sun/jdi tests are stable now, however for some reason
I am unable to run them with 'jtreg -samevm', but all indications are
that this is unrelated to the JDI tests themselves (the same tests run
fine with -samevm in openjdk7). The jtreg -samevm runs fail with a jtreg
error:
Error. Error while cleaning up threads after test
and it is unpredictable as to what testcase it will happen with.
I'll investigate this again after the openjdk6 version of hotspot
is updated (I think it has 14.0 right now), I'm suspecting a hotspot fix
in 14.2 regarding threads might solve this (6772683, but that's a guess).
Here is the complete webrev:
http://cr.openjdk.java.net/~ohair/openjdk6/jdk6-serviceability-update/webrev/
One changeset contains several bugs because the changes to the files
overlapped so much. I tried to selectively bring in changesets from jdk7
as I needed them, only realizing toward the end that I really needed all
changesets related to JDI to get it right.
Here is the complete bug list:
6263966: TEST: com/sun/jdi/ClassesByName2Test.java has a race
6432567: PIT : com/sun/jdi/BadHandshakeTest.java fails due to
java.net.ConnectException
6529758: JVMTI Waiters demo crashes. Double free.
6614052: jhat fails to read heap dump >2GB
6646613: ClassPrepareRequest.addSourceNameFilter() does not behave as
documented
6700889: Thread resume invalidates all stack frames, even from other threads
6701700: MonitorInfo objects aren't invalidated when the owning thread is
resumed
6725543: Compiler warnings in serviceability native code
6730273: TEST: JDI_REGRESSION test Solaris32AndSolaris64Test.sh fails if
-XX:+UseCompressedOops is used
6737900: TEST: Some JDI regression tests timeout on slow machines
6751643: ThreadReference.ownedMonitors() can return null
6787605: OpenSolaris doesn't have /usr/ucb/ps so ShellScaffold fails
6855180: Fix classfile version check in java_crw_demo
6855551: java -Xrunhprof crashes when running with classes compiled with
targed=7
6888927: Fix jdk jtreg tests to indicate which ones need othervm, allow for
use of samevm option
6892742: Improve root set used by jhat
6893426: ShellScaffold.sh fails on Solaris 10 update releases: /usr/bin/id:
illegal option -- u
6899444: Fix demo/jvmti tests so they can run in jtreg samevm mode, cleanup
problemlist
6902323: Fix testcase sun/tools/native2ascii/NativeErrors.java
6902325: Fix testcase sun/tools/jhat/HatHeapDump1Test.java
6902667: Fix JT_HOME not working from env in jdk/test/Makefile
6903102: 3/3 fixes in nightly testing version of ShellScaffold.sh need to be
committed
6904183: Fix jdk/test/com/sun/jdi tests to run with -samevm
6905705: Fix broken exit code values in jdk/test/Makefile
6906210: Fix another minor typo in test/Makefile
-kto
Are you sure this webrev was created against OpenJDK6 correctly?
Well, I looks like I did, did a 'hg pull -u' and ran webrev.
There are changes to:
Cdiffs Udiffs Sdiffs Frames Old New Patch Raw
test/java/awt/KeyboardFocusmanager/TypeAhead/TestDialogTypeAhead.java
72 lines changed: 7 ins; 57 del; 8 mod; 416 unchg
Cdiffs Udiffs Sdiffs Frames Old New Patch Raw
test/java/util/logging/LoggingDeadlock2.java
186 lines changed: 157 ins; 5 del; 24 mod; 60 unchg
Cdiffs Udiffs Sdiffs Frames Old New Patch Raw
test/sun/security/tools/keytool/StartDateTest.java
which aren't part of the bug IDs above but were changed by recent
pushes to OpenJDK6. Is this a merge changeset mixed in with the JDI
changes?
It is due to a merge changeset, I did not change those files in
my changesets. These come from rev 495de07832ca already in openjdk6,
and also d7d0e90c9f72 (StartDateTest.java) already in openjdk.
I haven't used a webrev on this many changsets before, seems like a bug
to me in webrev.
-kto
