Inliner

Very similar to TreeInfo.isPureExpr, but with the following inliner-only exceptions:

• synthetic case class apply methods, when the case class constructor is empty, are elideable but not pure. Elsewhere, accessing the apply method might cause the initialization of a containing object so they are merely idempotent.

For a retained inline method, another method that keeps track of the body that is kept at runtime. For instance, an inline method

inline override def f(x: T) = b

is complemented by the body retainer method

private def f$retainedBody(x: T) = f(x)

where the call f(x) is inline-expanded. This body is then transferred back to f at erasure, using method addRetainedInlineBodies.

The body to inline for method sym, or EmptyTree if none exists.

Replace Inlined node by a block that contains its bindings and expansion

sym is an inline method with a known body to inline.

Are we in an inline method body?

Try to inline a call to an inline method. Fail with error if the maximal inline depth is exceeded.

Leave only a call trace consisting of

• a reference to the top-level class from which the call was inlined,
• the call's position in the call field of an Inlined node. The trace has enough info to completely reconstruct positions. Note: For macros it returns a Select and for other inline methods it returns an Ident (this distinction is only temporary to be able to run YCheckPositions)

Try to inline a pattern with an inline unapply method. Fail with error if the maximal inline depth is exceeded.

Can a call to method meth be inlined?

Should call be inlined in this context?