On Solaris the "Signal Dispatcher" will assert in non-product builds if the semaphore function that it waits on (sema_wait) is interrupted (EINTR).

# To suppress the following error report, specify this argument
# after -XX: or in .hotspotrc: SuppressErrorAt=/os_solaris.cpp:2020]
#
# An unexpected error has been detected by HotSpot Virtual Machine:
#
# Internal Error (/export/Users/alanb/ws/append_ws/src/os/solaris/vm/os_solaris.cpp, 2020 [ Patched ]), pid=19606, tid=7
#
# Java VM: Java HotSpot(TM) Server VM (1.6-internal-debug mixed mode)
#
# Error: assert(ret == 0,"sema_wait() failed")
# An error report file with more information is saved as hs_err_pid19606.log
#
# If you would like to submit a bug report, please visit:
# http://java.sun.com/webapps/bugreport/crash.jsp
#
Current thread is 7
Dumping core ...
java version "1.6.0-internal"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.6.0-internal)
Java HotSpot(TM) Server VM (build 1.6-internal-debug, mixed mode)
Abort(coredump)

I've created a small test case to demonstrate the issue. The test cases uses fork(2) to launch a command. Obviously fork1 is the right call to make. The not nice thing about fork is that interrupts all blocked threads. The test case is attached in test.zip. The commands to demonstrate it are:

unzip test.zip
export LD_LIBRARY_PATH=.
java Test

If you want to re-build the test case there's a bld.sh script included.