Bug ID: JDK-6524009 Java 7 b07 throws core dump if creating COF 2.0 file by javatest 3.2.2 harness

Type: Bug
Component: hotspot
Sub-Component: compiler
Affected Version: 6u4,7

Priority: P3
Status: Resolved
Resolution: Cannot Reproduce
OS: solaris,solaris_10
CPU: sparc

Submitted: 2007-02-12
Updated: 2013-04-30
Resolved: 2013-04-30

Other
hs25Resolved

JDK 7 b07 Solaris 10 sparc 4v
javatest 3.2.2 fcs (b09) from JCK 6a package.
Problem doesn't appear with the latest javatest 4.1 b04

Normal (using javatest 3.2.2 b03 and JDK 6u1 b03)
[stt-robot@stt-32]$ /set/stt/jdk_promotions/JDK6u1/b03/binaries/solaris-sparcv9/jdk1.6.0_01/bin/java -Xmx512m -cp /set/stt/gtee/scripts/javatest/javatest.jar com.sun.javatest.cof.Main -o /set/stt/jck_workspace/jdk7/b07/runtime/runtime-solaris10-Niagara-server-Xcomp-Xincgc-Xcheckjni/cof20.xml /set/stt/jck_workspace/jdk7/b07/runtime/runtime-solaris10-Niagara-server-Xcomp-Xincgc-Xcheckjni/workDir


crash (using javatest 3.2.2 b09 and JDK 7 b07)

[stt-robot@stt-32]$ /set/stt/jdk_promotions/JDK7.0/b07/binaries/solaris-sparcv9/jdk1.7.0/bin/java -Xmx512m -cp /set/stt/gtee/scripts/javatest/javatest.jar com.sun.javatest.cof.Main -o /set/stt/jck_workspace/jdk7/b07/runtime/runtime-solaris10-Niagara-server-Xcomp-Xincgc-Xcheckjni/cof20_byjdk7.xml /set/stt/jck_workspace/jdk7/b07/runtime/runtime-solaris10-Niagara-server-Xcomp-Xincgc-Xcheckjni/workDir
#
# An unexpected error has been detected by Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0xfed4711c, pid=21822, tid=32
#
# Java VM: Java HotSpot(TM) Server VM (1.7.0-ea-b07 mixed mode)
# Problematic frame:
# V  [libjvm.so+0x54711c]
#
# An error report file with more information is saved as hs_err_pid21822.log
#
# If you would like to submit a bug report, please visit:
#   http://java.sun.com/webapps/bugreport/crash.jsp
#

The paths to the inputs for reproducing the bug no longer work. If the files are moved, or if the bug still reproduces, please reopen/refile with new information.
30-04-2013
Can you reproduce in JDK8? If not please close.
26-04-2013
EVALUATION This appears to be a latent bug with the changes for 6339956. Split_if is trying to split a phi and is visiting it's users. One of the users is a FastLockNode and split_up is called on the FastLockNode. The FastLockNode is a subclass of CmpNode and the code in split_up assumes that the user of a CmpNode is a Bool. After locks were converted to macro nodes the user of the FastLockNode became LockNode. In debug mode we assert because the is_Bool fails on the LockNode. We probably need to put a Bool in front of the FastLock so that it follows that the rule that CmpNodes are used by BoolNodes. Alternatively we could handle this specially in split_if but I'm not sure about that. After macro expansion we'll have a Bool in front of the FastLock so it couldn't hurt to create it earlier. Here's the relevant piece of the graph. 1448 Phi === 138 1444 1435 228 110 110 [[ 1616 1651 1824 1674 2833 1651 2881 5524 1615 1843 5274 2947 5473 3070 5231 2991 5439 5420 5215 3016 5405 4388 5350 3045 1616 2703 5605 1616 2423 1766 3109 5136 1651 3129 5031 5005 1626 3289 3320 3357 3372 3389 1793 1766 1782 3512 3528 3562 3576 3588 3616 4915 3741 4814 4786 3817 3848 3887 3903 3920 2582 1626 1851 4048 4064 4098 4112 4124 4152 4774 4283 4760 4324 4726 4340 4354 5549 1803 5624 1616 2328 2339 2272 4418 4438 4581 4564 4485 4516 4545 2143 2165 1936 4710 1916 2247 2362 2547 1880 1722 1722 1674 1674 1736 1736 1736 1651 ]] #com/sun/javatest/TestResult:NotNull * Oop:com/sun/javatest/TestResult:NotNull * !orig=[1588],[2304] !jvms: COFTest::fillTestCases @ bci:93 1615 FastLock === _ 1448 148 [[ 1616 ]] !jvms: TestResult::getSection @ bci:-1 COFTest::fillTestCases @ bci:130 1616 Lock === 5902 1 1591 8 1 ( 1448 148 1615 10 1448 209 1590 1 1 1 1 1 1 1448 1590 1 1 148 1448 ) [[ 1617 1619 ]] void ( java/lang/Object:NotNull *, rawptr:BotPTR, bool ) TestResult::getSection @ bci:-1 COFTest::fillTestCases @ bci:130 !jvms: TestResult::getSection @ bci:-1 COFTest::fillTestCases @ bci:130
14-02-2007
EVALUATION Steps to reproduce that does not include /set were included to comments section at submission, please find them below: ==================================================================================================================== steps to reproduce: run the following script on Niagara system: /net/stt-13.russia/export/stt/jdk_promotions/JDK7.0/b04/binaries/solaris-sparcv9/jdk1.7.0/bin/java -Xmx512m -cp /net/stt-13.russia/export/stt/jck_promotions/6a/beta/b05/binaries/JCK-runtime-6a/lib/javatest.jar com.sun.javatest.cof.Main -o /net/stt-13.russia/export/stt/jck_workspace/jdk7/b07/runtime/runtime-solaris10-Niagara-server-Xcomp-Xincgc-Xcheckjni/cof20_byjdk7.xml /net/stt-13.russia/export/stt/jck_workspace/jdk7/b07/runtime/runtime-solaris10-Niagara-server-Xcomp-Xincgc-Xcheckjni/workDir ====================================================================================================================
13-02-2007
EVALUATION This may be a duplicate of 6486391. Please include the full hs_err log file with bug reports. The short log isn't very helpful. Also, could you please include a command line for the test case that doesn't use /set? That doesn't work in our domain.
13-02-2007