| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
| |
Local and anonymous classes need to have an EnclosingMethod attribute
denoting the enclosing class and method. In fact, the enclosing class
must always be defined for local and anonymous classes, but the
enclosing method may be null (for local / anonymous classes defined
in field initializers or local blocks within a class body).
The new test here ensures that classes declared within a lazy val
initializer block indeed have the enclosing method set to null.
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Until now, concrete methods in traits were encoded with
"trait implementation classes".
- Such a trait would compile to two class files
- the trait interface, a Java interface, and
- the implementation class, containing "trait implementation methods"
- trait implementation methods are static methods has an explicit self
parameter.
- some methods don't require addition of an interface method, such as
private methods. Calls to these directly call the implementation method
- classes that mixin a trait install "trait forwarders", which implement
the abstract method in the interface by forwarding to the trait
implementation method.
The new encoding:
- no longer emits trait implementation classes or trait implementation
methods.
- instead, concrete methods are simply retained in the interface, as JVM 8
default interface methods (the JVM spec changes in
[JSR-335](http://download.oracle.com/otndocs/jcp/lambda-0_9_3-fr-eval-spec/index.html)
pave the way)
- use `invokespecial` to call private or particular super implementations
of a method (rather `invokestatic`)
- in cases when we `invokespecial` to a method in an indirect ancestor, we add
that ancestor redundantly as a direct parent. We are investigating alternatives
approaches here.
- we still emit trait fowrarders, although we are
[investigating](https://github.com/scala/scala-dev/issues/98) ways to only do
this when the JVM would be unable to resolve the correct method using its rules
for default method resolution.
Here's an example:
```
trait T {
println("T")
def m1 = m2
private def m2 = "m2"
}
trait U extends T {
println("T")
override def m1 = super[T].m1
}
class C extends U {
println("C")
def test = m1
}
```
The old and new encodings are displayed and diffed here: https://gist.github.com/retronym/f174d23f859f0e053580
Some notes in the implementation:
- No need to filter members from class decls at all in AddInterfaces
(although we do have to trigger side effecting info transformers)
- We can now emit an EnclosingMethod attribute for classes nested
in private trait methods
- Created a factory method for an AST shape that is used in
a number of places to symbolically bind to a particular
super method without needed to specify the qualifier of
the `Super` tree (which is too limiting, as it only allows
you to refer to direct parents.)
- I also found a similar tree shape created in Delambdafy,
that is better expressed with an existing tree creation
factory method, mkSuperInit.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The InnerClass attribute needs to contain an entry for every nested
class that is defined or referenced in a class. Details are in a
doc comment in BTypes.scala.
Instead of collecting ClassBTypes of nested classes into a hash map
during code generation, traverse the class before writing it out to
disk. The previous approach was incorrect as soon as the generated
bytecode was modified by the optimzier (DCE, inlining).
Fixes https://github.com/scala/scala-dev/issues/21.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Move run/t8960 to pending
It tests the serialVersionUID field on closure classes. The field
doesn't exist for indyLambda closures.
See https://issues.scala-lang.org/browse/SI-9373
Move some reify tests to pending
They fail at runtime in GenBCode since scala is built with indyLambda
enabled:
java.lang.AssertionError: assertion failed: Bad superClass for trait JFunction1: class Any
at scala.tools.nsc.Global.assert(Global.scala:261)
at scala.tools.nsc.backend.jvm.BTypesFromSymbols.setClassInfo(BTypesFromSymbols.scala:228)
Noted in https://issues.scala-lang.org/browse/SI-9374
force t6546 to GenASM - no closure elimination in GenBCode yet
Noted in https://issues.scala-lang.org/browse/SI-9364.
Fix or disable some tests that fail because of the old optimizer
The old inliner fails more often when the library is built with
indylambda.
Noted in https://issues.scala-lang.org/browse/SI-9374.
Example: List.foreach
➜ sandbox git:(jfun) ✗ qs -Ybackend:GenASM -optimize -Yinline-warnings
Welcome to Scala version 2.12.0-20150630-220939-1cb032d806 (Java HotSpot(TM) 64-Bit Server VM, Java 1.8.0_45).
Type in expressions to have them evaluated.
Type :help for more information.
scala> List(1,2,3).foreach(x => x + 1)
<console>:11: warning: Could not inline required method foreach because bytecode unavailable.
List(1,2,3).foreach(x => x + 1)
^
<console>:11: warning: At the end of the day, could not inline @inline-marked method foreach
List(1,2,3).foreach(x => x + 1)
^
Upate a number of tests for having indyLambda enabled
The delambdafyLambdaClassNames tests was removed, there's nothing to
tests with indyLambda.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Recently, in 029cce7, I changed uncurry to selectively fallback
to the old method of emitting lambdas when we detect that
`-Ydelambdafy:method`.
The change in classfile names breaks the expectations of
the test `innerClassAttribute`.
This commit changes that test to avoid using specialized
functions, so that under -Ydelambdafy:method all functions
are uniform. This changes a few fresh suffixes for anonymous
class names under both `-Ydelambdafy:{inline,method}`, so the
expectations have been duly updated.
Similarly, I have changed `javaReflection` in the same manner.
Its checkfiles remained unchanged.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Members of value classes are moved over to the companion object early.
This change ensures that closure classes nested in value classes
appear that way to Java reflection.
This commit also changes the EnclosingMethod attribute for classes
(and anonymous functions) nested in anonymous function bodies. Before,
the enclosing method was in some cases the function's apply method.
Not always though:
() => { class C ... val a = { class D ...} }
The class C used to be nested in the function's apply method, but not
D, because the value definition for a was lifted out of the apply.
After this commit, we uniformly set the enclosing method of classes
nested in function bodies to `null`. This is consistent with the
source-level view of the code.
Note that under delambdafy:method, closures never appear as enclosing
classes (this didn't change in this commit).
|
| |
|
|
|
|
|
|
|
|
| |
Trait implementation classes and specialized classes are always
considered top-level in terms of the InnerClass / EnclosingMethod
attributes. These attributes describe source-level properties, and
such classes are a compilation artifact.
Note that the same is true for delambdafy:method closure classes
(they are always top-level).
|
| |
|
|
|
|
|
|
|
|
| |
Private trait methods are not added to the generated interface, they
end up only in the implementation class. For classes nested in such
methods, the EnclosingMethod attribute was incorrect.
Since the EnclosingMethod attribute expresses a source-level property,
but the actual enclosing method does not exist in the bytecode, we
set the enclosing method to null.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change fixes both GenASM and GenBCode, except for the change
to renaming in LamdaLift mentioned below.
The reason for an inconsistent EnclosingMethod attribute was the
symbol owner chain. Initially, closure class symbols don't exist, they
are only created in UnCurry (delambdafy:inline). So walking the
originalOwner of a definition does not yield closure classes.
The commit also fixes uses of isAnonymousClass, isAnonymousFunction
and isDelambdafyFunction in two ways:
1. by phase-travelling to an early phase. after flatten, the name
includes the name of outer classes, so the properties may become
accidentally true (they check for a substring in the name)
2. by ensuring that the (destructive) renames during LambdaLift
don't make the above properties accidentally true. This was in
fact the cause for SI-8900.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit seems bigger than it is. Most of it is tests, and moving
some code around. The actual changes are small, but a bit subtle.
The InnerClass and EnclosingMethod attributes should now be close to
the JVM spec (which is summarized in BTypes.scala). New tests make
sure that changes to these attributes, and changes to the way Java
reflection sees Scala classfiles, don't go unnoticed.
A new file, BCodeAsmCommon, holds code that's shared between the two
backend (it could hold more, future work).
In general, the difficulty with emitting InnerClass / EnclosingMethod
is that we need to find out source-level properties. We need to make
sure to do enough phase-travelling, and work around destructive
changes to the ownerchain in lambdalift (we use originalOwner a lot).
The change to JavaMirrors is prompted by the change to the
EnclosingMethod attribute, which changes Java reflection's answer to
getEnclosingMethod and getEnclosingConstructor. Classes defined in
field initializers no longer have an enclosing method, just an
enclosing class, which broke an assumption in JavaMirrors.
There's one change in erasure. Before this change, when an object
declaration implements / overrides a method, and a bridge is required,
then the bridge method was actually a ModuleSymbol (it would get the
lateMETHOD flag and be emitted as a method anyway). This is confusing,
when iterating through the members of a class, you can find two
modules with the same name, and one of them doesn't have a module
class. Now, such bridge methods will be MethodSymbols.
Removed Symbol.originalEnclosingMethod, that is a backend thing and
doesn't need to live in the symbol API.
|
|
|
Some parts of the test assert (current) buggy behavior. This is marked
in the test file with TODO. It will be fixed in later work on the
backend.
|
