Allocate VM stack outside of ThreadLocalState

Having a large array inside the VirtualMachine (and thus the
ThreadLocalState) seems to crash the linker on Mac. It's also just
sort of wonky.

Probably ideally we'd have a more sophisticated setup here with
mmap and guard pages and growth and yada yada, but this works ok.
There is a definite suboptimality in allocating the entire max
stack size all at once, but it's not _too_ big so maybe it's ok.
It also means that the GC will walk the entire stack for pointers,
including the inactive part, but Boehm is probably not capable
of doing something more complex anyway. Or if it is it's wizardry.

include/clasp/gctools/threadlocal.h

@@ -90,7 +90,7 @@ struct VirtualMachine {
   // grow and etc.
   static constexpr size_t MaxStackWords = 65536;
   bool           _Running;
-  core::T_O*     _stackBottom[MaxStackWords];
+  core::T_O**    _stackBottom;
   size_t         _stackBytes;
   core::T_O**    _stackTop;
   core::T_O**    _stackGuard;
@@ -112,7 +112,8 @@ struct VirtualMachine {
 
   void enable_guards();
   void disable_guards();
-
+
+  void startup();
   inline void shutdown() {
     this->_Running = false;
   }
@@ -375,6 +376,8 @@ struct VirtualMachine {
     void pushObjectFile(llvmo::ObjectFile_sp of);
     void popObjectFile();
     inline DynamicBindingStack& bindings() { return this->_Bindings; };
+
+    void startUpVM();
 
     ~ThreadLocalState();
   };

src/core/mpPackage.cc

@@ -242,6 +242,7 @@ void start_thread_inner(uintptr_t uniqueId, void* cold_end_of_stack) {
 #endif
   gctools::ThreadLocalStateLowLevel thread_local_state_low_level(cold_end_of_stack);
   core::ThreadLocalState thread_local_state;
+  thread_local_state.startUpVM();
   my_thread_low_level = &thread_local_state_low_level;
   my_thread = &thread_local_state;
 //  printf("%s:%d entering start_thread  &my_thread -> %p \n", __FILE__, __LINE__, (void*)&my_thread);

src/gctools/boehmGarbageCollection.cc

@@ -458,6 +458,7 @@ void startupBoehm(gctools::ClaspInfo* claspInfo ) {
   GC_get_stack_base(&gc_stack_base);
   GC_register_my_thread(&gc_stack_base);
 #endif
+  thread_local_stateP->startUpVM();
   core::global_options = new core::CommandLineOptions(claspInfo->_argc, claspInfo->_argv);
 
 #ifndef SCRAPING

src/gctools/threadlocal.cc

@@ -159,8 +159,11 @@ VirtualMachine::VirtualMachine() :
     ,_unwind_counter(0)
     ,_throw_counter(0)
 #endif
-{
+{}
+
+void VirtualMachine::startup() {
   size_t stackSpace = VirtualMachine::MaxStackWords*sizeof(T_O*);
+  this->_stackBottom = (T_O**)GC_MALLOC_UNCOLLECTABLE(stackSpace);
   this->_stackTop = this->_stackBottom+VirtualMachine::MaxStackWords-1;
 //  printf("%s:%d:%s vm._stackTop = %p\n", __FILE__, __LINE__, __FUNCTION__, this->_stackTop );
   size_t pageSize = getpagesize();
@@ -203,6 +206,7 @@ VirtualMachine::~VirtualMachine() {
 #if 1
   this->disable_guards();
 #endif
+  GC_FREE(this->_stackBottom);
 }
 
 
@@ -427,6 +431,10 @@ void ThreadLocalState::popObjectFile() {
   SIMPLE_ERROR("There were no more object files");
 }
 
+void ThreadLocalState::startUpVM() {
+  this->_VM.startup();
+}
+
 ThreadLocalState::~ThreadLocalState() {
 }