1195972 - js::ctypes::FieldInfo is only valid on the heap, but is allocated on the stack

Status: RESOLVED WONTFIX

(Core :: JavaScript: GC, defect)

Description

[Nika Layzell [:nika] (ni? for response)]

js/src/ctypes/CTypes.cpp:5565:7: error: variable of type 'js::ctypes::FieldInfo' only valid on the heap
      FieldInfo info;
      ^
js/src/ctypes/CTypes.cpp:5565:7: note: value incorrectly allocated in an automatic variable
js/src/ctypes/CTypes.h:236:23: note: 'js::ctypes::FieldInfo' is a heap type because member 'mType' is a heap type 'JS::Heap<JSObject *>'
  JS::Heap<JSObject*> mType;    // CType of the field
                      ^

Comment 1

[Nika Layzell [:nika] (ni? for response)]

Related:

../../dist/include/js/HashTable.h:1696:9: warning: variable of type 'typename HashTableEntry<HashMapEntry<JSFlatString *, FieldInfo> >::NonConstT' (aka 'js::HashMapEntry<JSFlatString *, js::ctypes::FieldInfo>') only valid on the heap
        typename HashTableEntry<T>::NonConstT t(mozilla::Move(*p));
        ^
../../dist/include/js/HashTable.h:1696:9: note: value incorrectly allocated in an automatic variable
../../dist/include/js/HashTable.h:650:11: note: 'typename HashTableEntry<HashMapEntry<JSFlatString *, FieldInfo> >::NonConstT' (aka 'js::HashMapEntry<JSFlatString *, js::ctypes::FieldInfo>') is a heap type because member 'value_' is a heap type 'js::ctypes::FieldInfo'
    Value value_;
          ^
js/src/ctypes/CTypes.h:236:23: note: 'js::ctypes::FieldInfo' is a heap type because member 'mType' is a heap type 'JS::Heap<JSObject *>'
  JS::Heap<JSObject*> mType;    // CType of the field
                      ^

Comment 2

[Jon Coppeard (:jonco)]

Part of the problem is that HashTable::rekeyWithoutRehash() constructs a hash table entry on the stack.  It might be possible to rework it to avoid doing that, or maybe we can get rid of rekeying when entirely Terrence's work on giving objects stable hash codes lands.

Comment 3

[Nika Layzell [:nika] (ni? for response)]

Has there been any progress on getting rid of this problem?

Comment 4

[Jon Coppeard (:jonco)]

(In reply to Michael Layzell [:mystor] from comment #3)
Work on this is progressing, but we're not there yet.  Needinfo'ing myself to remind me to check back later.

Comment 5

[Jon Coppeard (:jonco)]

Closing as per bug 1195568 comment 6.