If you try to run HotSpot debug build when the current working directory
can't be determined (e.g., it's been removed), the JVM crashes.

Steps to reproduce the problem:

In shell 1:

    $ mkdir Foo
    $ cd Foo

In shell 2:

    $ rmdir Foo

Back in shell 1:

    $ $Deployed/JDK-1.5.0/bin/java_g -version
    Warning: Cannot open log file: hotspot.log
    Warning: Forcing option -XX:LogFile=/tmp//hs_pid3449.log
    Error occurred during initialization of VM
    java.lang.Error: Properties init: Could not determine current working directory.
    Dumping core....
    Abort

Of course, dbx is no help in this situation:

    $ dbx $Deployed/JDK-1.5.0/bin/java_g core
    fatal error: Cannot find run directory for dbx -- Error 0

and anyway, since there's no current directory, there's no core file to debug:

    $ ls -l core
    core not found

The java.lang.Error comes from

    /java/re/jdk/1.5.0/latest/ws/j2se/src/solaris/native/java/lang/java_props_md.c

and the stack trace looks like it would be:

    getcwd(buf, sizeof(buf)) in java_props_md.c
    java_props_t *sprops = GetJavaProperties(env); in System.c
    initProperties(props); in System.java
    JavaCalls::call_static(.... initializeSystemClass_name(), ....) in thread.cpp
    call_initializeSystemClass(CHECK_0); in thread.cpp
    jint result = Threads::create_vm((JavaVMInitArgs*) args); in jni.cpp
    r = ifn->CreateJavaVM(pvm, (void **)penv, &args); in java.c

I'm not sure who is being surprised by the exception and dumping
core. The product build puts out the error message and then seems
to exit cleanly, so I'm lowering the priority.