[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
// Copyright 2021 The Go Authors. All rights reserved.
|
|
|
|
|
// Use of this source code is governed by a BSD-style
|
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
|
|
|
|
|
|
package noder
|
|
|
|
|
|
|
|
|
|
import (
|
|
|
|
|
"cmd/compile/internal/base"
|
|
|
|
|
"cmd/compile/internal/ir"
|
|
|
|
|
"cmd/compile/internal/syntax"
|
|
|
|
|
"cmd/compile/internal/typecheck"
|
|
|
|
|
"cmd/compile/internal/types"
|
|
|
|
|
"cmd/compile/internal/types2"
|
|
|
|
|
"cmd/internal/src"
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
func (g *irgen) def(name *syntax.Name) (*ir.Name, types2.Object) {
|
|
|
|
|
obj, ok := g.info.Defs[name]
|
|
|
|
|
if !ok {
|
|
|
|
|
base.FatalfAt(g.pos(name), "unknown name %v", name)
|
|
|
|
|
}
|
|
|
|
|
return g.obj(obj), obj
|
|
|
|
|
}
|
|
|
|
|
|
2021-03-17 17:54:41 -07:00
|
|
|
// use returns the Name node associated with the use of name. The returned node
|
|
|
|
|
// will have the correct type and be marked as typechecked.
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
func (g *irgen) use(name *syntax.Name) *ir.Name {
|
2021-03-17 17:54:41 -07:00
|
|
|
obj2, ok := g.info.Uses[name]
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
if !ok {
|
|
|
|
|
base.FatalfAt(g.pos(name), "unknown name %v", name)
|
|
|
|
|
}
|
2021-03-17 17:54:41 -07:00
|
|
|
obj := ir.CaptureName(g.pos(obj2), ir.CurFunc, g.obj(obj2))
|
|
|
|
|
if obj.Defn != nil && obj.Defn.Op() == ir.ONAME {
|
|
|
|
|
// If CaptureName created a closure variable, then transfer the
|
|
|
|
|
// type of the captured name to the new closure variable.
|
|
|
|
|
obj.SetTypecheck(1)
|
|
|
|
|
obj.SetType(obj.Defn.Type())
|
|
|
|
|
}
|
|
|
|
|
return obj
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
}
|
|
|
|
|
|
2021-03-17 17:54:41 -07:00
|
|
|
// obj returns the Name that represents the given object. If no such Name exists
|
|
|
|
|
// yet, it will be implicitly created. The returned node will have the correct
|
|
|
|
|
// type and be marked as typechecked.
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
//
|
|
|
|
|
// For objects declared at function scope, ir.CurFunc must already be
|
|
|
|
|
// set to the respective function when the Name is created.
|
|
|
|
|
func (g *irgen) obj(obj types2.Object) *ir.Name {
|
|
|
|
|
// For imported objects, we use iimport directly instead of mapping
|
|
|
|
|
// the types2 representation.
|
|
|
|
|
if obj.Pkg() != g.self {
|
|
|
|
|
sym := g.sym(obj)
|
|
|
|
|
if sym.Def != nil {
|
|
|
|
|
return sym.Def.(*ir.Name)
|
|
|
|
|
}
|
|
|
|
|
n := typecheck.Resolve(ir.NewIdent(src.NoXPos, sym))
|
|
|
|
|
if n, ok := n.(*ir.Name); ok {
|
2021-03-17 17:54:41 -07:00
|
|
|
n.SetTypecheck(1)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
return n
|
|
|
|
|
}
|
|
|
|
|
base.FatalfAt(g.pos(obj), "failed to resolve %v", obj)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
if name, ok := g.objs[obj]; ok {
|
|
|
|
|
return name // previously mapped
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
var name *ir.Name
|
|
|
|
|
pos := g.pos(obj)
|
|
|
|
|
|
|
|
|
|
class := typecheck.DeclContext
|
|
|
|
|
if obj.Parent() == g.self.Scope() {
|
|
|
|
|
class = ir.PEXTERN // forward reference to package-block declaration
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// "You are in a maze of twisting little passages, all different."
|
|
|
|
|
switch obj := obj.(type) {
|
|
|
|
|
case *types2.Const:
|
|
|
|
|
name = g.objCommon(pos, ir.OLITERAL, g.sym(obj), class, g.typ(obj.Type()))
|
|
|
|
|
|
|
|
|
|
case *types2.Func:
|
|
|
|
|
sig := obj.Type().(*types2.Signature)
|
|
|
|
|
var sym *types.Sym
|
|
|
|
|
var typ *types.Type
|
|
|
|
|
if recv := sig.Recv(); recv == nil {
|
|
|
|
|
if obj.Name() == "init" {
|
|
|
|
|
sym = renameinit()
|
|
|
|
|
} else {
|
|
|
|
|
sym = g.sym(obj)
|
|
|
|
|
}
|
|
|
|
|
typ = g.typ(sig)
|
|
|
|
|
} else {
|
2021-01-24 23:31:25 -08:00
|
|
|
sym = g.selector(obj)
|
|
|
|
|
if !sym.IsBlank() {
|
|
|
|
|
sym = ir.MethodSym(g.typ(recv.Type()), sym)
|
|
|
|
|
}
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
typ = g.signature(g.param(recv), sig)
|
|
|
|
|
}
|
|
|
|
|
name = g.objCommon(pos, ir.ONAME, sym, ir.PFUNC, typ)
|
|
|
|
|
|
|
|
|
|
case *types2.TypeName:
|
|
|
|
|
if obj.IsAlias() {
|
|
|
|
|
name = g.objCommon(pos, ir.OTYPE, g.sym(obj), class, g.typ(obj.Type()))
|
|
|
|
|
} else {
|
|
|
|
|
name = ir.NewDeclNameAt(pos, ir.OTYPE, g.sym(obj))
|
|
|
|
|
g.objFinish(name, class, types.NewNamed(name))
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
case *types2.Var:
|
|
|
|
|
var sym *types.Sym
|
|
|
|
|
if class == ir.PPARAMOUT {
|
|
|
|
|
// Backend needs names for result parameters,
|
|
|
|
|
// even if they're anonymous or blank.
|
|
|
|
|
switch obj.Name() {
|
|
|
|
|
case "":
|
|
|
|
|
sym = typecheck.LookupNum("~r", len(ir.CurFunc.Dcl)) // 'r' for "result"
|
|
|
|
|
case "_":
|
|
|
|
|
sym = typecheck.LookupNum("~b", len(ir.CurFunc.Dcl)) // 'b' for "blank"
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if sym == nil {
|
|
|
|
|
sym = g.sym(obj)
|
|
|
|
|
}
|
|
|
|
|
name = g.objCommon(pos, ir.ONAME, sym, class, g.typ(obj.Type()))
|
|
|
|
|
|
|
|
|
|
default:
|
|
|
|
|
g.unhandled("object", obj)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
g.objs[obj] = name
|
2021-03-17 17:54:41 -07:00
|
|
|
name.SetTypecheck(1)
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
return name
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (g *irgen) objCommon(pos src.XPos, op ir.Op, sym *types.Sym, class ir.Class, typ *types.Type) *ir.Name {
|
|
|
|
|
name := ir.NewDeclNameAt(pos, op, sym)
|
|
|
|
|
g.objFinish(name, class, typ)
|
|
|
|
|
return name
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
func (g *irgen) objFinish(name *ir.Name, class ir.Class, typ *types.Type) {
|
|
|
|
|
sym := name.Sym()
|
|
|
|
|
|
|
|
|
|
name.SetType(typ)
|
|
|
|
|
name.Class = class
|
|
|
|
|
if name.Class == ir.PFUNC {
|
|
|
|
|
sym.SetFunc(true)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
name.SetTypecheck(1)
|
|
|
|
|
name.SetWalkdef(1)
|
|
|
|
|
|
|
|
|
|
if ir.IsBlank(name) {
|
|
|
|
|
return
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
switch class {
|
|
|
|
|
case ir.PEXTERN:
|
|
|
|
|
g.target.Externs = append(g.target.Externs, name)
|
|
|
|
|
fallthrough
|
|
|
|
|
case ir.PFUNC:
|
|
|
|
|
sym.Def = name
|
|
|
|
|
if name.Class == ir.PFUNC && name.Type().Recv() != nil {
|
|
|
|
|
break // methods are exported with their receiver type
|
|
|
|
|
}
|
|
|
|
|
if types.IsExported(sym.Name) {
|
2021-02-08 10:23:05 -08:00
|
|
|
if name.Class == ir.PFUNC && name.Type().NumTParams() > 0 {
|
|
|
|
|
base.FatalfAt(name.Pos(), "Cannot export a generic function (yet): %v", name)
|
|
|
|
|
}
|
[dev.typeparams] cmd/compile: add types2-based noder
This CL adds "irgen", a new noding implementation that utilizes types2
to guide IR construction. Notably, it completely skips dealing with
constant and type expressions (aside from using ir.TypeNode to
interoperate with the types1 typechecker), because types2 already
handled those. It also omits any syntax checking, trusting that types2
already rejected any errors.
It currently still utilizes the types1 typechecker for the desugaring
operations it handles (e.g., turning OAS2 into OAS2FUNC/etc, inserting
implicit conversions, rewriting f(g()) functions, and so on). However,
the IR is constructed in a fully incremental fashion, so it should be
easy to now piecemeal replace those dependencies as needed.
Nearly all of "go test std cmd" passes with -G=3 enabled by
default. The main remaining blocker is the number of test/run.go
failures. There also appear to be cases where types2 does not provide
us with position information. These will be iterated upon.
Portions and ideas from Dan Scales's CL 276653.
Change-Id: Ic99e8f2d0267b0312d30c10d5d043f5817a59c9d
Reviewed-on: https://go-review.googlesource.com/c/go/+/281932
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Go Bot <gobot@golang.org>
Reviewed-by: Dan Scales <danscales@google.com>
Reviewed-by: Robert Griesemer <gri@golang.org>
Trust: Matthew Dempsky <mdempsky@google.com>
Trust: Robert Griesemer <gri@golang.org>
2021-01-09 00:57:55 -08:00
|
|
|
typecheck.Export(name)
|
|
|
|
|
}
|
|
|
|
|
if base.Flag.AsmHdr != "" && !name.Sym().Asm() {
|
|
|
|
|
name.Sym().SetAsm(true)
|
|
|
|
|
g.target.Asms = append(g.target.Asms, name)
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
default:
|
|
|
|
|
// Function-scoped declaration.
|
|
|
|
|
name.Curfn = ir.CurFunc
|
|
|
|
|
if name.Op() == ir.ONAME {
|
|
|
|
|
ir.CurFunc.Dcl = append(ir.CurFunc.Dcl, name)
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|