deps: upgrade v8 to 4.2.77.18

This commit applies a secondary change in order to make `make test` pass cleanly, specifically re-disabling post-mortem debugging in common.gypi. PR-URL: #1506 Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>
nodejs · petkaantonov · Apr 17, 2015 · Apr 18, 2015 · Apr 18, 2015 · Apr 18, 2015
commit 01652c7709c7f5019cef1df9cdc859943db6e203
diff --git a/common.gypi b/common.gypi
@@ -20,16 +20,17 @@
     # Enable disassembler for `--print-code` v8 options
     'v8_enable_disassembler': 1,
 
+    # Disable support for postmortem debugging, continuously broken.
+    'v8_postmortem_support%': 'false',
+
     # Don't bake anything extra into the snapshot.
     'v8_use_external_startup_data%': 0,
 
     'conditions': [
       ['OS == "win"', {
         'os_posix': 0,
-        'v8_postmortem_support%': 'false',
       }, {
         'os_posix': 1,
-        'v8_postmortem_support%': 'true',
       }],
       ['GENERATOR == "ninja" or OS== "mac"', {
         'OBJ_DIR': '<(PRODUCT_DIR)/obj',

diff --git a/deps/v8/AUTHORS b/deps/v8/AUTHORS
@@ -61,7 +61,7 @@ Jan de Mooij <jandemooij@gmail.com>
 Jay Freeman <saurik@saurik.com>
 James Pike <g00gle@chilon.net>
 Jianghua Yang <jianghua.yjh@alibaba-inc.com>
-Joel Stanley <joel.stan@gmail.com>
+Joel Stanley <joel@jms.id.au>
 Johan Bergström <johan@bergstroem.nu>
 Jonathan Liu <net147@gmail.com>
 Kang-Hao (Kenny) Lu <kennyluck@csail.mit.edu>

diff --git a/deps/v8/include/v8-version.h b/deps/v8/include/v8-version.h
@@ -11,7 +11,7 @@
 #define V8_MAJOR_VERSION 4
 #define V8_MINOR_VERSION 2
 #define V8_BUILD_NUMBER 77
-#define V8_PATCH_LEVEL 15
+#define V8_PATCH_LEVEL 18
 
 // Use 1 for candidates and 0 otherwise.
 // (Boolean macro values are not supported by all preprocessors.)

diff --git a/deps/v8/include/v8.h b/deps/v8/include/v8.h
@@ -932,6 +932,24 @@ class V8_EXPORT EscapableHandleScope : public HandleScope {
   internal::Object** escape_slot_;
 };
 
+class V8_EXPORT SealHandleScope {
+ public:
+  SealHandleScope(Isolate* isolate);
+  ~SealHandleScope();
+
+ private:
+  // Make it hard to create heap-allocated or illegal handle scopes by
+  // disallowing certain operations.
+  SealHandleScope(const SealHandleScope&);
+  void operator=(const SealHandleScope&);
+  void* operator new(size_t size);
+  void operator delete(void*, size_t);
+
+  internal::Isolate* isolate_;
+  int prev_level_;
+  internal::Object** prev_limit_;
+};
+
 
 /**
  * A simple Maybe type, representing an object which may or may not have a
@@ -1004,24 +1022,6 @@ class ScriptOrigin {
   Handle<Integer> script_id_;
 };
 
-class V8_EXPORT SealHandleScope {
- public:
-  SealHandleScope(Isolate* isolate);
-  ~SealHandleScope();
-
- private:
-  // Make it hard to create heap-allocated or illegal handle scopes by
-  // disallowing certain operations.
-  SealHandleScope(const SealHandleScope&);
-  void operator=(const SealHandleScope&);
-  void* operator new(size_t size);
-  void operator delete(void*, size_t);
-
-  internal::Isolate* isolate_;
-  int prev_level_;
-  internal::Object** prev_limit_;
-};
-
 
 /**
  * A compiled JavaScript script, not yet tied to a Context.

diff --git a/deps/v8/src/compiler/x64/instruction-selector-x64.cc b/deps/v8/src/compiler/x64/instruction-selector-x64.cc
@@ -812,22 +812,12 @@ void InstructionSelector::VisitTruncateInt64ToInt32(Node* node) {
   Node* value = node->InputAt(0);
   if (CanCover(node, value)) {
     switch (value->opcode()) {
-      case IrOpcode::kWord64Sar: {
+      case IrOpcode::kWord64Sar:
+      case IrOpcode::kWord64Shr: {
         Int64BinopMatcher m(value);
-        if (m.right().IsInRange(1, 32)) {
+        if (m.right().Is(32)) {
           Emit(kX64Shr, g.DefineSameAsFirst(node),
-               g.UseRegister(m.left().node()),
-               g.UseImmediate(m.right().node()));
-          return;
-        }
-        break;
-      }
-      case IrOpcode::kWord64Shl: {
-        Int64BinopMatcher m(value);
-        if (m.right().IsInRange(1, 31)) {
-          Emit(kX64Shl32, g.DefineSameAsFirst(node),
-               g.UseRegister(m.left().node()),
-               g.UseImmediate(m.right().node()));
+               g.UseRegister(m.left().node()), g.TempImmediate(32));
           return;
         }
         break;
@@ -836,9 +826,7 @@ void InstructionSelector::VisitTruncateInt64ToInt32(Node* node) {
         break;
     }
   }
-  // Otherwise truncation from 64-bit to 32-bit is a no-nop, as 32-bit
-  // operations just ignore the upper 64-bit.
-  Emit(kArchNop, g.DefineAsRegister(node), g.Use(value));
+  Emit(kX64Movl, g.DefineAsRegister(node), g.Use(value));
 }
 
 

diff --git a/deps/v8/src/heap/gc-idle-time-handler.h b/deps/v8/src/heap/gc-idle-time-handler.h
@@ -125,6 +125,9 @@ class GCIdleTimeHandler {
   // That is the maximum idle time we will have during frame rendering.
   static const size_t kMaxFrameRenderingIdleTime = 16;
 
+  // Minimum idle time to start incremental marking.
+  static const size_t kMinIdleTimeToStartIncrementalMarking = 10;
+
   // If we haven't recorded any scavenger events yet, we use a conservative
   // lower bound for the scavenger speed.
   static const size_t kInitialConservativeScavengeSpeed = 100 * KB;

diff --git a/deps/v8/src/heap/heap.cc b/deps/v8/src/heap/heap.cc
@@ -104,6 +104,8 @@ Heap::Heap()
       allocation_timeout_(0),
 #endif  // DEBUG
       old_generation_allocation_limit_(initial_old_generation_size_),
+      idle_old_generation_allocation_limit_(
+          kMinimumOldGenerationAllocationLimit),
       old_gen_exhausted_(false),
       inline_allocation_disabled_(false),
       store_buffer_rebuilder_(store_buffer()),
@@ -1159,8 +1161,7 @@ bool Heap::PerformGarbageCollection(
     // Temporarily set the limit for case when PostGarbageCollectionProcessing
     // allocates and triggers GC. The real limit is set at after
     // PostGarbageCollectionProcessing.
-    old_generation_allocation_limit_ =
-        OldGenerationAllocationLimit(PromotedSpaceSizeOfObjects(), 0);
+    SetOldGenerationAllocationLimit(PromotedSpaceSizeOfObjects(), 0);
     old_gen_exhausted_ = false;
     old_generation_size_configured_ = true;
   } else {
@@ -1194,8 +1195,8 @@ bool Heap::PerformGarbageCollection(
     // Register the amount of external allocated memory.
     amount_of_external_allocated_memory_at_last_global_gc_ =
         amount_of_external_allocated_memory_;
-    old_generation_allocation_limit_ = OldGenerationAllocationLimit(
-        PromotedSpaceSizeOfObjects(), freed_global_handles);
+    SetOldGenerationAllocationLimit(PromotedSpaceSizeOfObjects(),
+                                    freed_global_handles);
     // We finished a marking cycle. We can uncommit the marking deque until
     // we start marking again.
     mark_compact_collector_.UncommitMarkingDeque();
@@ -4558,7 +4559,7 @@ bool Heap::TryFinalizeIdleIncrementalMarking(
 
 bool Heap::WorthActivatingIncrementalMarking() {
   return incremental_marking()->IsStopped() &&
-         incremental_marking()->WorthActivating() && NextGCIsLikelyToBeFull();
+         incremental_marking()->ShouldActivate();
 }
 
 
@@ -4583,6 +4584,7 @@ bool Heap::IdleNotification(double deadline_in_seconds) {
       static_cast<double>(base::Time::kMillisecondsPerSecond);
   HistogramTimerScope idle_notification_scope(
       isolate_->counters()->gc_idle_notification());
+  double idle_time_in_ms = deadline_in_ms - MonotonicallyIncreasingTimeInMs();
 
   GCIdleTimeHandler::HeapState heap_state;
   heap_state.contexts_disposed = contexts_disposed_;
@@ -4591,8 +4593,15 @@ bool Heap::IdleNotification(double deadline_in_seconds) {
   heap_state.size_of_objects = static_cast<size_t>(SizeOfObjects());
   heap_state.incremental_marking_stopped = incremental_marking()->IsStopped();
   // TODO(ulan): Start incremental marking only for large heaps.
+  intptr_t limit = old_generation_allocation_limit_;
+  if (static_cast<size_t>(idle_time_in_ms) >
+      GCIdleTimeHandler::kMinIdleTimeToStartIncrementalMarking) {
+    limit = idle_old_generation_allocation_limit_;
+  }
+
   heap_state.can_start_incremental_marking =
-      incremental_marking()->ShouldActivate() && FLAG_incremental_marking;
+      incremental_marking()->WorthActivating() &&
+      NextGCIsLikelyToBeFull(limit) && FLAG_incremental_marking;
   heap_state.sweeping_in_progress =
       mark_compact_collector()->sweeping_in_progress();
   heap_state.mark_compact_speed_in_bytes_per_ms =
@@ -4610,7 +4619,6 @@ bool Heap::IdleNotification(double deadline_in_seconds) {
       static_cast<size_t>(
           tracer()->NewSpaceAllocationThroughputInBytesPerMillisecond());
 
-  double idle_time_in_ms = deadline_in_ms - MonotonicallyIncreasingTimeInMs();
   GCIdleTimeAction action =
       gc_idle_time_handler_.Compute(idle_time_in_ms, heap_state);
   isolate()->counters()->gc_idle_time_allotted_in_ms()->AddSample(
@@ -5358,21 +5366,37 @@ int64_t Heap::PromotedExternalMemorySize() {
 }
 
 
-intptr_t Heap::OldGenerationAllocationLimit(intptr_t old_gen_size,
-                                            int freed_global_handles) {
+intptr_t Heap::CalculateOldGenerationAllocationLimit(double factor,
+                                                     intptr_t old_gen_size) {
+  CHECK(factor > 1.0);
+  CHECK(old_gen_size > 0);
+  intptr_t limit = static_cast<intptr_t>(old_gen_size * factor);
+  limit = Max(limit, kMinimumOldGenerationAllocationLimit);
+  limit += new_space_.Capacity();
+  intptr_t halfway_to_the_max = (old_gen_size + max_old_generation_size_) / 2;
+  return Min(limit, halfway_to_the_max);
+}
+
+
+void Heap::SetOldGenerationAllocationLimit(intptr_t old_gen_size,
+                                           int freed_global_handles) {
   const int kMaxHandles = 1000;
   const int kMinHandles = 100;
-  double min_factor = 1.1;
+  const double min_factor = 1.1;
   double max_factor = 4;
+  const double idle_max_factor = 1.5;
   // We set the old generation growing factor to 2 to grow the heap slower on
   // memory-constrained devices.
   if (max_old_generation_size_ <= kMaxOldSpaceSizeMediumMemoryDevice) {
     max_factor = 2;
   }
+
   // If there are many freed global handles, then the next full GC will
   // likely collect a lot of garbage. Choose the heap growing factor
   // depending on freed global handles.
   // TODO(ulan, hpayer): Take into account mutator utilization.
+  // TODO(hpayer): The idle factor could make the handles heuristic obsolete.
+  // Look into that.
   double factor;
   if (freed_global_handles <= kMinHandles) {
     factor = max_factor;
@@ -5391,11 +5415,10 @@ intptr_t Heap::OldGenerationAllocationLimit(intptr_t old_gen_size,
     factor = min_factor;
   }
 
-  intptr_t limit = static_cast<intptr_t>(old_gen_size * factor);
-  limit = Max(limit, kMinimumOldGenerationAllocationLimit);
-  limit += new_space_.Capacity();
-  intptr_t halfway_to_the_max = (old_gen_size + max_old_generation_size_) / 2;
-  return Min(limit, halfway_to_the_max);
+  old_generation_allocation_limit_ =
+      CalculateOldGenerationAllocationLimit(factor, old_gen_size);
+  idle_old_generation_allocation_limit_ = CalculateOldGenerationAllocationLimit(
+      Min(factor, idle_max_factor), old_gen_size);
 }
 
 

diff --git a/deps/v8/src/heap/heap.h b/deps/v8/src/heap/heap.h
@@ -628,6 +628,10 @@ class Heap {
   // Returns of size of all objects residing in the heap.
   intptr_t SizeOfObjects();
 
+  intptr_t old_generation_allocation_limit() const {
+    return old_generation_allocation_limit_;
+  }
+
   // Return the starting address and a mask for the new space.  And-masking an
   // address with the mask will result in the start address of the new space
   // for all addresses in either semispace.
@@ -1112,8 +1116,14 @@ class Heap {
   static const int kMaxExecutableSizeHugeMemoryDevice =
       256 * kPointerMultiplier;
 
-  intptr_t OldGenerationAllocationLimit(intptr_t old_gen_size,
-                                        int freed_global_handles);
+  // Calculates the allocation limit based on a given growing factor and a
+  // given old generation size.
+  intptr_t CalculateOldGenerationAllocationLimit(double factor,
+                                                 intptr_t old_gen_size);
+
+  // Sets the allocation limit to trigger the next full garbage collection.
+  void SetOldGenerationAllocationLimit(intptr_t old_gen_size,
+                                       int freed_global_handles);
 
   // Indicates whether inline bump-pointer allocation has been disabled.
   bool inline_allocation_disabled() { return inline_allocation_disabled_; }
@@ -1219,13 +1229,12 @@ class Heap {
     survived_since_last_expansion_ += survived;
   }
 
-  inline bool NextGCIsLikelyToBeFull() {
+  inline bool NextGCIsLikelyToBeFull(intptr_t limit) {
     if (FLAG_gc_global) return true;
 
     if (FLAG_stress_compaction && (gc_count_ & 1) != 0) return true;
 
-    intptr_t adjusted_allocation_limit =
-        old_generation_allocation_limit_ - new_space_.Capacity();
+    intptr_t adjusted_allocation_limit = limit - new_space_.Capacity();
 
     if (PromotedTotalSize() >= adjusted_allocation_limit) return true;
 
@@ -1604,6 +1613,10 @@ class Heap {
   // generation and on every allocation in large object space.
   intptr_t old_generation_allocation_limit_;
 
+  // The allocation limit when there is > kMinIdleTimeToStartIncrementalMarking
+  // idle time in the idle time handler.
+  intptr_t idle_old_generation_allocation_limit_;
+
   // Indicates that an allocation has failed in the old generation since the
   // last GC.
   bool old_gen_exhausted_;

diff --git a/deps/v8/src/heap/incremental-marking.cc b/deps/v8/src/heap/incremental-marking.cc
@@ -422,7 +422,9 @@ void IncrementalMarking::ActivateIncrementalWriteBarrier() {
 
 
 bool IncrementalMarking::ShouldActivate() {
-  return WorthActivating() && heap_->NextGCIsLikelyToBeFull();
+  return WorthActivating() &&
+         heap_->NextGCIsLikelyToBeFull(
+             heap_->old_generation_allocation_limit());
 }
 
 

diff --git a/deps/v8/src/hydrogen.cc b/deps/v8/src/hydrogen.cc
@@ -2391,6 +2391,8 @@ HInstruction* HGraphBuilder::BuildUncheckedMonomorphicElementAccess(
     PropertyAccessType access_type,
     LoadKeyedHoleMode load_mode,
     KeyedAccessStoreMode store_mode) {
+  DCHECK(top_info()->IsStub() || checked_object->IsCompareMap() ||
+         checked_object->IsCheckMaps());
   DCHECK((!IsExternalArrayElementsKind(elements_kind) &&
               !IsFixedTypedArrayElementsKind(elements_kind)) ||
          !is_js_array);
@@ -8401,11 +8403,10 @@ bool HOptimizedGraphBuilder::TryInlineBuiltinMethodCall(
         new_size = AddUncasted<HAdd>(length, graph()->GetConstant1());
 
         bool is_array = receiver_map->instance_type() == JS_ARRAY_TYPE;
-        BuildUncheckedMonomorphicElementAccess(array, length,
-                                               value_to_push, is_array,
-                                               elements_kind, STORE,
-                                               NEVER_RETURN_HOLE,
-                                               STORE_AND_GROW_NO_TRANSITION);
+        HValue* checked_array = Add<HCheckMaps>(array, receiver_map);
+        BuildUncheckedMonomorphicElementAccess(
+            checked_array, length, value_to_push, is_array, elements_kind,
+            STORE, NEVER_RETURN_HOLE, STORE_AND_GROW_NO_TRANSITION);
 
         if (!ast_context()->IsEffect()) Push(new_size);
         Add<HSimulate>(expr->id(), REMOVABLE_SIMULATE);
@@ -8763,18 +8764,9 @@ void HOptimizedGraphBuilder::HandleIndirectCall(Call* expr, HValue* function,
   int args_count_no_receiver = arguments_count - 1;
   if (function->IsConstant() &&
       HConstant::cast(function)->handle(isolate())->IsJSFunction()) {
-    HValue* receiver = environment()->ExpressionStackAt(args_count_no_receiver);
-    Handle<Map> receiver_map;
-    if (receiver->IsConstant() &&
-        HConstant::cast(receiver)->handle(isolate())->IsHeapObject()) {
-      receiver_map =
-          handle(Handle<HeapObject>::cast(
-                     HConstant::cast(receiver)->handle(isolate()))->map());
-    }
-
     known_function =
         Handle<JSFunction>::cast(HConstant::cast(function)->handle(isolate()));
-    if (TryInlineBuiltinMethodCall(expr, known_function, receiver_map,
+    if (TryInlineBuiltinMethodCall(expr, known_function, Handle<Map>(),
                                    args_count_no_receiver)) {
       if (FLAG_trace_inlining) {
         PrintF("Inlining builtin ");