Some C1 changes that went in JDK1.4 build 75 broke hprof cpu=times option
and VM seg fault in CodeEmitInfo::add_registers_to_oop_map() where
_local_mapping is NULL.

Running hprof on the Jmol application fails with "oopmap not found"
assertion using JDK1.4 build 74, which is a known problem
that Dave Cox is working on (see 4478223).

Using JDK1.4 build 75, now it fails much earlier while compiling Java
method. To reproduce,

> cd /home/zzhang/bugs/footprint
> /re/promoted/jdk/1.4/beta_refresh/b75/binaries/solsparc/bin/java_g -Xrunhprof:cpu=times -cp jmol-0.2/jars/jmol.jar -Djmol.home=jmol-0.2 jmol.Jmol

The stack trace is:
  ---- called from signal handler with signal 11 (SIGSEGV) ------
  [14] LocalMapping::is_cache_reg(0x0, 0xf37810e4, 0x0, 0x19, 0x1c6630, 0xfe04b7e0), at 0xfe12b298
  [15] CodeEmitInfo::add_registers_to_oop_map(0xe8fb4, 0x1c88a4, 0x16, 0x1, 0x1, 0x0), at 0xfe146078
  [16] CodeEmitInfo::compute_debug_info(0xe8fb4, 0x1000000, 0x16, 0x9, 0x0, 0x1c8894), at 0xfe144d78
  [17] CodeEmitInfo::oop_map(0xe8fb4, 0x1000000, 0x3, 0x1c6f98, 0xa, 0x0), at 0xfe144bf0
  [18] CodeEmitInfo::record_debug_info(0xe8fb4, 0x183f8c, 0xe4, 0x1, 0x0, 0x0), at 0xfe1450cc
  [19] LIR_Assembler::add_debug_info(0xf37815fc, 0xe4, 0xe8fb4, 0x1, 0xf9cb052c, 0x0), at 0xfe182134
  [20] LIR_Assembler::add_call_info(0xf37815fc, 0xe4, 0xe8fb4, 0xe8fb4, 0x0, 0x0), at 0xfe1820ac
  [21] LIR_Assembler::emit_delay(0xf37815fc, 0xe9140, 0x0, 0x1c6628, 0x1c662b, 0x1c6624), at 0xfe19e584
  [22] LIR_OpDelay::emit_code(0xe9140, 0xf37815fc, 0x0, 0x0, 0x1c6630, 0xfe04b7e0), at 0xfe173374
  [23] LIR_CodeGen::block_do(0xf37815d4, 0x1c2a10, 0x0, 0x1, 0x1, 0x0), at 0xfe1707fc
  [24] BlockList::iterate_forward(0x1c6608, 0xf37815d4, 0x1c2924, 0x0, 0x0, 0xffffffff), at 0xfe039220
  [25] BlockCollection::iterate_forward(0xe85b8, 0xf37815d4, 0xf37815fc, 0xf3781634, 0x1e3100, 0x0), at 0xfe14104c
  [26] Compilation::emit_code_body(0xf37818b4, 0xf37817ac, 0x0, 0x0, 0x0, 0x0), at 0xfe13e0ac
  [27] Compilation::compile_java_method(0xf37818b4, 0xf37817ac, 0x4, 0x1c12ec, 0x0, 0x1842d8), at 0xfe13e578
  [28] Compilation::compile_method(0xf37818b4, 0x183eec, 0x4, 0x0, 0x0, 0x0), at 0xfe13ef78
  [29] Compilation::Compilation(0xf37818b4, 0xf37819b8, 0xbb974, 0xffffffff, 0x183f6c, 0x1), at 0xfe13f2b0
  [30] Compiler::compile_method(0x1830d8, 0xf37819b8, 0x0, 0xbb974, 0xffffffff, 0x0), at 0xfe033fc8
  [31] CompileBroker::invoke_compiler_on_method(0x1c0b10, 0xf78061d8, 0x6, 0x0, 0x0, 0x0), at 0xfdf807b8
  [32] CompileBroker::compiler_thread_loop(0x0, 0x5, 0x0, 0xff38e000, 0xf3781e10, 0xf3781e00), at 0xfdf7fe90
  [33] compiler_thread_entry(0x183d18, 0x183d18, 0x1, 0x0, 0x0, 0x0), at 0xfde65460
  [34] JavaThread::thread_main_inner(0x183d18, 0xcb2f0, 0x6, 0x0, 0x0, 0x0), at 0xfde601b8
  [35] JavaThread::run(0x183d18, 0x0, 0x0, 0x0, 0x0, 0x0), at 0xfde60050
  [36] _start(0x183d18, 0xff38f6a0, 0x1, 0x1, 0xff38e000, 0x0), at 0xfdd53a4c

It seg fault in CodeEmitInfo::add_registers_to_oop_map()
when calling _local_mapping->is_cache_reg(rinfo) but
_local_mapping is NULL.

void CodeEmitInfo::add_registers_to_oop_map(OopMap* map) {
  const RInfoCollection* reg_oops = this->register_oops();
  assert(reg_oops != NULL, "null check");
  int frame_size = frame_map()->framesize();
  int arg_count = frame_map()->oop_map_arg_count();
  for (int i = 0; i < reg_oops->length(); i++) {
    RInfo rinfo = reg_oops->at(i);
    if (!_local_mapping->is_cache_reg(rinfo)) {
      // cache registers have already been added
      OptoReg::Name rn = frame_map()->register_regname(rinfo);
      map->set_oop(rn, frame_size, arg_count);
    }
  }
}