Introduce a FunctionSyntaxKind for FieldInitializer since special rules
(around `arguments`) apply. At the same time we can move the flag from the
scope to the JSScript. This is similar to how derived constructors are handled
and makes the initWithEnclosingScope code closer to initWithEnclosingContext.
This version is a bit more complex than Mozilla's due to different storage of
bit flags on JSScript.
Based-on: m-c 1636800
This state still has the initializers scoped on .initializers local variable, which will be changed later.
Based-on: m-c 1499448, 1530084, 1530832, 1529448 (partial), 1532921, 1528039, 1528038, 1535166, 1550628,
1535166, 1550628, 1541641, 1547133, 1540787, 1535804/9
Because of the wacko way we handle toString offsets for class default constructors, those offsets cannot be recovered if we relazify the functions.
Luckily there's no reason to relazify them, their JSScripts are very small: either a single 'retrval' for non-derived, and still fairly small for derived.
This is accomplished in the following ways.
LazyScripts and JSScripts now have 4 offsets:
- Source begin and end for the actual source. This is used for lazy parsing.
- toString begin and end for toString. Some kinds of functions, like async, only have a different begin offset. Class constructors have different offsets for both begin and end.
For syntactically present (i.e. non-default) constructors, the class source span is remembered directly on the LazyScript or JSScript. The toString implementation then splices out the substring directly. For default constructors, a new SRC_CLASS SrcNote type is added. It's binary and has as its arguments the begin and end offsets of the class expression or statement. MakeDefaultConstructor reads the note and overrides the cloned self-hosted function's source object. This is probably the least intrusive way to accomplish this.
Bug 1342553 - Part 0.1: Use try-catch for IteratorClose in for-of
Bug 1343072 - Update HasLiveStackValueAtDepth to follow the change in
JSTRY_FOR_OF
Bug 1344753 - Update for-of stack depth in
ControlFlowGenerator::processWhileOrForInLoop
Issue #74
