#136: Start working on database persistence

vendor/modernc.org/cc/v3/cc.go

@@ -0,0 +1,960 @@
+// Copyright 2019 The CC Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+//TODO https://todo.sr.ht/~mcf/cc-issues/34
+//TODO http://mcpp.sourceforge.net/ "Provides a validation suite to test C/C++ preprocessor's conformance and quality comprehensively."
+
+//go:generate rm -f lexer.go
+//go:generate golex -o lexer.go lexer.l
+
+//go:generate rm -f ast.go
+//go:generate yy -o /dev/null -position -astImport "\"fmt\"\n\n\"modernc.org/token\"" -prettyString PrettyString -kind Case -noListKind -noPrivateHelpers -forceOptPos parser.yy
+
+//go:generate stringer -output stringer.go -linecomment -type=Kind,Linkage
+
+//go:generate sh -c "go test -run ^Example |fe"
+
+// Package cc is a C99 compiler front end (Work in progress).
+//
+// Installation
+//
+// To install/update cc/v3 invoke:
+//
+//     $ go get [-u] modernc.org/cc/v3
+//
+// Online documentation
+//
+// See https://godoc.org/modernc.org/cc/v3.
+//
+// Status
+//
+// Most of the functionality is now working.
+//
+// Supported platforms
+//
+// The code is known to work on Darwin, Linux and Windows, but the supported
+// features may vary.
+//
+// Links
+//
+// Referenced from elsewhere:
+//
+//  [0]: http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf
+//  [1]: https://www.spinellis.gr/blog/20060626/cpp.algo.pdf
+//  [2]: http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1570.pdf
+//  [3]: http://gallium.inria.fr/~fpottier/publis/jourdan-fpottier-2016.pdf
+//  [4]: https://gcc.gnu.org/onlinedocs/gcc-8.3.0/gcc/Attribute-Syntax.html#Attribute-Syntax
+package cc // import "modernc.org/cc/v3"
+
+import (
+	"fmt"
+	goscanner "go/scanner"
+	gotoken "go/token"
+	"hash/maphash"
+	"io"
+	"math"
+	"os"
+	"os/exec"
+	"reflect"
+	"regexp"
+	"runtime"
+	"sort"
+	"strconv"
+	"strings"
+	"sync"
+	"sync/atomic"
+
+	"modernc.org/strutil"
+	"modernc.org/token"
+)
+
+const (
+	scopeParent StringID = -iota - 1
+	scopeSkip
+)
+
+var (
+	_ Pragma = (*pragma)(nil)
+
+	cache       = newPPCache()
+	dict        = newDictionary()
+	dictStrings [math.MaxUint8 + 1]string
+	noPos       token.Position
+
+	debugIncludePaths bool
+	debugWorkingDir   bool
+	isTesting         bool
+	isTestingMingw    bool
+
+	idPtrdiffT = dict.sid("ptrdiff_t")
+	idSizeT    = dict.sid("size_t")
+	idWCharT   = dict.sid("wchar_t")
+
+	token4Pool = sync.Pool{New: func() interface{} { r := make([]token4, 0); return &r }} //DONE benchmrk tuned capacity
+	tokenPool  = sync.Pool{New: func() interface{} { r := make([]Token, 0); return &r }}  //DONE benchmrk tuned capacity
+
+	printHooks = strutil.PrettyPrintHooks{
+		reflect.TypeOf(Token{}): func(f strutil.Formatter, v interface{}, prefix, suffix string) {
+			t := v.(Token)
+			if (t == Token{}) {
+				return
+			}
+
+			f.Format(prefix)
+			r := t.Rune
+			if p := t.Position(); p.IsValid() {
+				f.Format("%v: ", p)
+			}
+			s := tokName(r)
+			if x := s[0]; x >= '0' && x <= '9' {
+				s = strconv.QuoteRune(r)
+			}
+			f.Format("%s", s)
+			if s := t.Value.String(); len(s) != 0 {
+				f.Format(" %q", s)
+			}
+			f.Format(suffix)
+		},
+		reflect.TypeOf((*operand)(nil)): func(f strutil.Formatter, v interface{}, prefix, suffix string) {
+			op := v.(*operand)
+			f.Format(prefix)
+			f.Format("[%v %T(%[2]v)]", op.Type(), op.Value())
+			f.Format(suffix)
+		},
+	}
+)
+
+func todo(s string, args ...interface{}) string { //TODO-
+	switch {
+	case s == "":
+		s = fmt.Sprintf(strings.Repeat("%v ", len(args)), args...)
+	default:
+		s = fmt.Sprintf(s, args...)
+	}
+	pc, fn, fl, _ := runtime.Caller(1)
+	f := runtime.FuncForPC(pc)
+	var fns string
+	if f != nil {
+		fns = f.Name()
+		if x := strings.LastIndex(fns, "."); x > 0 {
+			fns = fns[x+1:]
+		}
+	}
+	r := fmt.Sprintf("%s:%d:%s: TODOTODO %s", fn, fl, fns, s) //TODOOK
+	fmt.Fprintf(os.Stdout, "%s\n", r)
+	os.Stdout.Sync()
+	return r
+}
+
+func trc(s string, args ...interface{}) string { //TODO-
+	switch {
+	case s == "":
+		s = fmt.Sprintf(strings.Repeat("%v ", len(args)), args...)
+	default:
+		s = fmt.Sprintf(s, args...)
+	}
+	_, fn, fl, _ := runtime.Caller(1)
+	r := fmt.Sprintf("%s:%d: TRC %s", fn, fl, s)
+	fmt.Fprintf(os.Stdout, "%s\n", r)
+	os.Stdout.Sync()
+	return r
+}
+
+func origin(skip int) string {
+	pc, fn, fl, _ := runtime.Caller(skip)
+	f := runtime.FuncForPC(pc)
+	var fns string
+	if f != nil {
+		fns = f.Name()
+		if x := strings.LastIndex(fns, "."); x > 0 {
+			fns = fns[x+1:]
+		}
+	}
+	return fmt.Sprintf("%s:%d:%s", fn, fl, fns)
+}
+
+// String returns a StringID for a given value.
+func String(s string) StringID {
+	return dict.sid(s)
+}
+
+// Linkage represents identifier linkage.
+//
+// [0]6.2.2: An identifier declared in different scopes or in the same scope
+// more than once can be made to refer to the same object or function by a
+// process called linkage. There are three kinds of linkage: External,
+// Internal, and None.
+type Linkage int
+
+// StorageClass determines storage duration.
+//
+// [0]6.2.4: An object has a storage duration that determines its lifetime.
+// There are three storage durations: Static, Automatic, and Allocated.
+type StorageClass int
+
+// Pragma defines behavior of the object passed to Config.PragmaHandler.
+type Pragma interface {
+	Error(msg string, args ...interface{}) // Report error.
+	MaxAligment() int                      // Returns the current maximum alignment. May return zero.
+	MaxInitialAligment() int               // Support #pragma pack(). Returns the maximum alignment in effect at start. May return zero.
+	PopMacro(string)
+	PushMacro(string)
+	SetAlignment(n int) // Support #pragma pack(n)
+}
+
+type pragma struct {
+	tok cppToken
+	c   *cpp
+}
+
+func (p *pragma) Error(msg string, args ...interface{}) { p.c.err(p.tok, msg, args...) }
+
+func (p *pragma) MaxAligment() int { return p.c.ctx.maxAlign }
+
+func (p *pragma) MaxInitialAligment() int { return p.c.ctx.maxAlign0 }
+
+func (p *pragma) SetAlignment(n int) {
+	if n <= 0 {
+		p.Error("%T.SetAlignment(%d): invalid argument", p, n)
+		return
+	}
+
+	p.c.ctx.maxAlign = n
+}
+
+func (p *pragma) PushMacro(nm string) {
+	id := dict.sid(nm)
+	if p.c.macroStack == nil {
+		p.c.macroStack = map[StringID][]*Macro{}
+	}
+	if m := p.c.macros[id]; m != nil {
+		p.c.macroStack[id] = append(p.c.macroStack[id], p.c.macros[id])
+	}
+}
+
+func (p *pragma) PopMacro(nm string) {
+	id := dict.sid(nm)
+	a := p.c.macroStack[id]
+	if n := len(a); n != 0 {
+		p.c.macros[id] = a[n-1]
+		p.c.macroStack[id] = a[:n-1]
+	}
+}
+
+// PrettyString returns a formatted representation of things produced by this package.
+func PrettyString(v interface{}) string {
+	return strutil.PrettyString(v, "", "", printHooks)
+}
+
+// StringID is a process-unique string numeric identifier. Its zero value
+// represents an empty string.
+type StringID int32
+
+// String implements fmt.Stringer.
+func (n StringID) String() (r string) {
+	if n < 256 {
+		return dictStrings[byte(n)]
+	}
+
+	dict.mu.RLock()
+	r = dict.strings[n]
+	dict.mu.RUnlock()
+	return r
+}
+
+// Node is implemented by Token and all AST nodes.
+type Node interface {
+	Position() token.Position
+}
+
+type noder struct{}
+
+func (noder) Position() token.Position { panic(internalError()) }
+
+// Scope maps identifiers to definitions.
+type Scope map[StringID][]Node
+
+func (s *Scope) new() (r Scope) {
+	if *s == nil {
+		*s = Scope{}
+	}
+	r = Scope{scopeParent: []Node{struct {
+		noder
+		Scope
+	}{Scope: *s}}}
+	return r
+}
+
+func (s *Scope) declare(nm StringID, n Node) {
+	sc := *s
+	if sc == nil {
+		*s = map[StringID][]Node{nm: {n}}
+		// t := ""
+		// if x, ok := n.(*Declarator); ok && x.IsTypedefName {
+		// 	t = ", typedefname"
+		// }
+		// dbg("declared %s%s at %v in scope %p", nm, t, n.Position(), *s)
+		return
+	}
+
+	switch x := n.(type) {
+	case *Declarator, *StructDeclarator, *LabeledStatement, *BlockItem:
+		// nop
+	case *StructOrUnionSpecifier, *EnumSpecifier, *Enumerator:
+		for {
+			if _, ok := sc[scopeSkip]; !ok {
+				break
+			}
+
+			sc = sc.Parent()
+		}
+	default:
+		panic(todo("%T", x))
+	}
+
+	sc[nm] = append(sc[nm], n)
+	// t := ""
+	// if x, ok := n.(*Declarator); ok && x.IsTypedefName {
+	// 	t = ", typedefname"
+	// }
+	// dbg("declared %s%s at %v in scope %p", nm, t, n.Position(), sc)
+}
+
+// Parent returns s's outer scope, if any.
+func (s Scope) Parent() Scope {
+	if s == nil {
+		return nil
+	}
+
+	if x, ok := s[scopeParent]; ok {
+		return x[0].(struct {
+			noder
+			Scope
+		}).Scope
+	}
+
+	return nil
+}
+
+func (s *Scope) typedef(nm StringID, tok Token) *Declarator {
+	seq := tok.seq
+	for s := *s; s != nil; s = s.Parent() {
+		for _, v := range s[nm] {
+			switch x := v.(type) {
+			case *Declarator:
+				if !x.isVisible(seq) {
+					continue
+				}
+
+				if x.IsTypedefName {
+					return x
+				}
+
+				return nil
+			case *Enumerator:
+				return nil
+			case *EnumSpecifier, *StructOrUnionSpecifier, *StructDeclarator:
+				// nop
+			default:
+				panic(internalError())
+			}
+		}
+	}
+	return nil
+}
+
+func (s *Scope) declarator(nm StringID, tok Token) *Declarator {
+	seq := tok.seq
+	for s := *s; s != nil; s = s.Parent() {
+		defs := s[nm]
+		for _, v := range defs {
+			switch x := v.(type) {
+			case *Declarator:
+				if !x.isVisible(seq) {
+					continue
+				}
+
+				for _, v := range defs {
+					if x, ok := v.(*Declarator); ok {
+						t := x.Type()
+						if t != nil && t.Kind() == Function {
+							if x.fnDef {
+								return x
+							}
+
+							continue
+						}
+
+						if t != nil && !x.Type().IsIncomplete() {
+							return x
+						}
+					}
+
+				}
+				return x
+			case *Enumerator:
+				return nil
+			case *EnumSpecifier, *StructOrUnionSpecifier, *StructDeclarator:
+				// nop
+			default:
+				panic(internalError())
+			}
+		}
+	}
+	return nil
+}
+
+func (s *Scope) enumerator(nm StringID, tok Token) *Enumerator {
+	seq := tok.seq
+	for s := *s; s != nil; s = s.Parent() {
+		for _, v := range s[nm] {
+			switch x := v.(type) {
+			case *Declarator:
+				if !x.isVisible(seq) {
+					continue
+				}
+
+				return nil
+			case *Enumerator:
+				return x
+			case *EnumSpecifier, *StructOrUnionSpecifier, *StructDeclarator:
+				// nop
+			default:
+				panic(internalError())
+			}
+		}
+	}
+	return nil
+}
+
+// Config3 amends behavior of translation phases 1 to 3.
+type Config3 struct {
+	// If IgnoreInclude is not nil, its MatchString method will be called by the
+	// preprocessor with the argument any include directive expands to. If the call
+	// evaluates to is true the include directive will be ignored completely.
+	IgnoreInclude *regexp.Regexp
+
+	// Name of a macro to use instead of FD_ZERO.
+	//
+	// Note: Temporary solution will be removed/replaced
+	ReplaceMacroFdZero string
+	// Name of a macro to use instead of TCL_DEFAULT_DOUBLE_ROUNDING.
+	//
+	// Note: Temporary solution will be removed/replaced
+	ReplaceMacroTclDefaultDoubleRounding string // Name of a macro to use instead of TCL_DEFAULT_DOUBLE_ROUNDING. Note: Temporrary solution will be removed/replaced
+	// Name of a macro to use instead of TCL_IEEE_DOUBLE_ROUNDING.
+	//
+	// Note: Temporary solution will be removed/replaced
+	ReplaceMacroTclIeeeDoubleRounding string
+
+	WorkingDir string     // Overrides os.Getwd if non empty.
+	Filesystem Filesystem // Overrides filesystem access if not empty.
+
+	MaxSourceLine int // Zero: Scanner will use default buffer. Non zero: Scanner will use max(default buffer size, MaxSourceLine).
+
+	// DisableBuiltinResolution disables resolution of undefined identifiers such
+	// that eg. abort, becomes the same as __builtin_abort, prototype of which is
+	// expected to be provided by one of the sources passed to Parse, Preprocess or
+	// Translate.
+	DisableBuiltinResolution bool
+
+	DisableTrigraphs bool // GCC ignores them unless -trigraphs is used: https://gcc.gnu.org/onlinedocs/cpp/Initial-processing.html
+	GCCStructs       bool // Assume __attribute__(gcc_struct) applied to structs by default.
+	//TODO MSStructs                               bool // Assume __attribute__(ms_struct) applied to structs by default.
+	NoFieldAndBitfieldOverlap               bool // Only bitfields can be grouped together.
+	PreserveOnlyLastNonBlankSeparator       bool // If PreserveWhiteSpace is true, keep only the last white space, do not combine
+	PreserveWhiteSpace                      bool // Including also comments.
+	RejectElseExtraTokens                   bool // Pedantic: do not silently accept "#else foo".
+	RejectEndifExtraTokens                  bool // Pedantic: do not silently accept "#endif foo".
+	RejectFinalBackslash                    bool // Pedantic: do not silently accept "foo\\\n".
+	RejectFunctionMacroEmptyReplacementList bool // Pedantic: do not silently accept "#define foo(bar)\n".
+	RejectIfdefExtraTokens                  bool // Pedantic: do not silently accept "#ifdef foo bar".
+	RejectIfndefExtraTokens                 bool // Pedantic: do not silently accept "#ifndef foo bar".
+	RejectIncludeNext                       bool // Pedantic: do not silently accept "#include_next".
+	RejectInvalidVariadicMacros             bool // Pedantic: do not silently accept "#define foo(bar...)". Standard allows only #define foo(bar, ...)
+	RejectLineExtraTokens                   bool // Pedantic: do not silently accept "#line 1234 \"foo.c\" bar".
+	RejectMissingFinalNewline               bool // Pedantic: do not silently accept "foo\nbar".
+	RejectUndefExtraTokens                  bool // Pedantic: do not silently accept "#undef foo bar".
+	UnsignedEnums                           bool // GCC compatibility: enums with no negative values will have unsigned type.
+}
+
+type SharedFunctionDefinitions struct {
+	M    map[*FunctionDefinition]struct{}
+	m    map[sharedFunctionDefinitionKey]*FunctionDefinition //TODO
+	hash maphash.Hash
+}
+
+type sharedFunctionDefinitionKey struct {
+	pos  StringID
+	nm   StringID
+	hash uint64
+}
+
+// Config amends behavior of translation phase 4 and above. Instances of Config
+// are not mutated by this package and it's safe to share/reuse them.
+//
+// The *Config passed to Parse or Translate should not be mutated afterwards.
+type Config struct {
+	Config3
+	ABI ABI
+
+	PragmaHandler func(Pragma, []Token) // Called on pragmas, other than #pragma STDC ..., if non nil
+
+	// SharedFunctionDefinitions collects function definitions having the
+	// same position and definition. This can happen, for example, when a
+	// function is defined in a header file included multiple times. Either
+	// within a single translation unit or across translation units. In the
+	// later case just supply the same SharedFunctionDefinitions in Config
+	// when translating/parsing each translation unit.
+	SharedFunctionDefinitions *SharedFunctionDefinitions
+
+	MaxErrors int // 0: default (10), < 0: unlimited, n: n.
+
+	CheckExternInlineFnBodies              bool // Translate will consider extern inline function bodies.
+	DebugIncludePaths                      bool // Output to stderr.
+	DebugWorkingDir                        bool // Output to stderr.
+	DoNotTypecheckAsm                      bool
+	EnableAssignmentCompatibilityChecking  bool // No such checks performed up to v3.31.0. Currently only partially implemented.
+	InjectTracingCode                      bool // Output to stderr.
+	LongDoubleIsDouble                     bool
+	PreprocessOnly                         bool
+	RejectAnonymousFields                  bool // Pedantic: do not silently accept "struct{int;}".
+	RejectCaseRange                        bool // Pedantic: do not silently accept "case 'a'...'z':".
+	RejectEmptyCompositeLiterals           bool // Pedantic: do not silently accept "foo = (T){}".
+	RejectEmptyDeclarations                bool // Pedantic: do not silently accept "int foo(){};".
+	RejectEmptyFields                      bool // Pedantic: do not silently accept "struct {int a;;} foo;".
+	RejectEmptyInitializerList             bool // Pedantic: do not silently accept "foo f = {};".
+	RejectEmptyStructDeclaration           bool // Pedantic: do not silently accept "struct{; int i}".
+	RejectEmptyStructs                     bool // Pedantic: do not silently accept "struct foo {};".
+	RejectIncompatibleMacroRedef           bool // Pedantic: do not silently accept "#define MIN(A,B) ...\n#define MIN(a,b) ...\n" etc.
+	RejectLabelValues                      bool // Pedantic: do not silently accept "foo: bar(); void *ptr = &&foo;" or "goto *ptr".
+	RejectLateBinding                      bool // Pedantic: do not silently accept void f() { g(); } void g() {}
+	RejectMissingConditionalExpr           bool // Pedantic: do not silently accept "foo = bar ? : baz;".
+	RejectMissingDeclarationSpecifiers     bool // Pedantic: do not silently accept "main() {}".
+	RejectMissingFinalStructFieldSemicolon bool // Pedantic: do not silently accept "struct{int i; int j}".
+	RejectNestedFunctionDefinitions        bool // Pedantic: do not silently accept nested function definitons.
+	RejectParamSemicolon                   bool // Pedantic: do not silently accept "int f(int a; int b)".
+	RejectStatementExpressions             bool // Pedantic: do not silently accept "i = ({foo();})".
+	RejectTypeof                           bool // Pedantic: do not silently accept "typeof foo" or "typeof(bar*)".
+	RejectUninitializedDeclarators         bool // Reject int f() { int j; return j; }
+	TrackAssignments                       bool // Collect a list of LHS declarators a declarator is used in RHS or as an function argument.
+	doNotSanityCheckComplexTypes           bool // Testing only
+	fakeIncludes                           bool // Testing only.
+	ignoreErrors                           bool // Testing only.
+	ignoreIncludes                         bool // Testing only.
+	ignoreUndefinedIdentifiers             bool // Testing only.
+}
+
+type context struct {
+	ast         *AST
+	breakCtx    Node
+	breaks      int
+	casePromote Type
+	cases       []*LabeledStatement // switch
+	cfg         *Config
+	checkFn     *FunctionDefinition
+	closure     map[StringID]struct{}
+	continues   int
+	enums       map[StringID]Operand //TODO putting this in alphabetical order within the struct causes crashes in VirtualBox/386 ???
+	goscanner.ErrorList
+	includePaths    []string
+	intBits         int
+	intMaxWidth     int64 // Set if the preprocessor saw __INTMAX_WIDTH__.
+	keywords        map[StringID]rune
+	maxAlign        int // If non zero: maximum alignment of members of structures (other than zero-width bitfields).
+	maxAlign0       int
+	maxErrors       int
+	mode            mode
+	modes           []mode
+	mu              sync.Mutex
+	ptrdiffT        Type
+	readDelta       int
+	sizeT           Type
+	structTypes     map[StringID]Type
+	structs         map[StructInfo]struct{}
+	switches        int
+	sysIncludePaths []string
+	tuSize0         int64 // Sum of sizes of processed inputs
+	tuSources0      int32 // Number of processed inputs
+	wcharT          Type
+
+	capture        bool
+	evalIdentError bool
+}
+
+func newContext(cfg *Config) *context {
+	maxErrors := cfg.MaxErrors
+	if maxErrors == 0 {
+		maxErrors = 10
+	}
+	return &context{
+		cfg:         cfg,
+		enums:       map[StringID]Operand{},
+		keywords:    keywords,
+		maxErrors:   maxErrors,
+		structTypes: map[StringID]Type{},
+		structs:     map[StructInfo]struct{}{},
+	}
+}
+
+func (c *context) tuSizeAdd(n int64)    { atomic.AddInt64(&c.tuSize0, n) }
+func (c *context) tuSize() int64        { return atomic.LoadInt64(&c.tuSize0) }
+func (c *context) tuSourcesAdd(n int32) { atomic.AddInt32(&c.tuSources0, n) }
+func (c *context) tuSources() int       { return int(atomic.LoadInt32(&c.tuSources0)) }
+
+func (c *context) stddef(nm StringID, s Scope, tok Token) Type {
+	if d := s.typedef(nm, tok); d != nil {
+		if t := d.Type(); t != nil && t.Kind() != Invalid {
+			return t
+		}
+	}
+
+	c.errNode(&tok, "front-end: undefined: %s", nm)
+	return noType
+}
+
+func (c *context) assignmentCompatibilityErrorCond(n Node, a, b Type) (stop bool) {
+	if !c.cfg.EnableAssignmentCompatibilityChecking {
+		return
+	}
+
+	return c.errNode(n, "invalid type combination of conditional operator: %v and %v", a, b)
+}
+
+func (c *context) assignmentCompatibilityError(n Node, lhs, rhs Type) (stop bool) {
+	if !c.cfg.EnableAssignmentCompatibilityChecking {
+		return
+	}
+
+	return c.errNode(n, "cannot use %v as type %v in assignment", rhs, lhs)
+}
+
+func (c *context) errNode(n Node, msg string, args ...interface{}) (stop bool) {
+	return c.err(n.Position(), msg, args...)
+}
+
+func (c *context) err(pos token.Position, msg string, args ...interface{}) (stop bool) {
+	// dbg("FAIL "+msg, args...)
+	//fmt.Printf("FAIL "+msg+"\n", args...)
+	if c.cfg.ignoreErrors {
+		return false
+	}
+
+	s := fmt.Sprintf(msg, args...)
+	c.mu.Lock()
+	max := c.maxErrors
+	switch {
+	case max < 0 || max > len(c.ErrorList):
+		c.ErrorList.Add(gotoken.Position(pos), s)
+	default:
+		stop = true
+	}
+	c.mu.Unlock()
+	return stop
+}
+
+func (c *context) errs(list goscanner.ErrorList) (stop bool) {
+	c.mu.Lock()
+
+	defer c.mu.Unlock()
+
+	max := c.maxErrors
+	for _, v := range list {
+		switch {
+		case max < 0 || max > len(c.ErrorList):
+			c.ErrorList = append(c.ErrorList, v)
+		default:
+			return true
+		}
+	}
+	return false
+}
+
+func (c *context) Err() error {
+	c.mu.Lock()
+	switch x := c.ErrorList.Err().(type) {
+	case goscanner.ErrorList:
+		x = append(goscanner.ErrorList(nil), x...)
+		c.mu.Unlock()
+		var lpos gotoken.Position
+		w := 0
+		for _, v := range x {
+			if lpos.Filename != "" {
+				if v.Pos.Filename == lpos.Filename && v.Pos.Line == lpos.Line {
+					continue
+				}
+			}
+
+			x[w] = v
+			w++
+			lpos = v.Pos
+		}
+		x = x[:w]
+		sort.Slice(x, func(i, j int) bool {
+			a := x[i]
+			b := x[j]
+			if !a.Pos.IsValid() && b.Pos.IsValid() {
+				return true
+			}
+
+			if a.Pos.IsValid() && !b.Pos.IsValid() {
+				return false
+			}
+
+			if a.Pos.Filename < b.Pos.Filename {
+				return true
+			}
+
+			if a.Pos.Filename > b.Pos.Filename {
+				return false
+			}
+
+			if a.Pos.Line < b.Pos.Line {
+				return true
+			}
+
+			if a.Pos.Line > b.Pos.Line {
+				return false
+			}
+
+			return a.Pos.Column < b.Pos.Column
+		})
+		a := make([]string, 0, len(x))
+		for _, v := range x {
+			a = append(a, v.Error())
+		}
+		return fmt.Errorf("%s", strings.Join(a, "\n"))
+	default:
+		c.mu.Unlock()
+		return x
+	}
+}
+
+func (c *context) not(n Node, mode mode) {
+	if c.mode&mode != 0 {
+		switch mode {
+		case mIntConstExpr:
+			c.errNode(n, "invalid integer constant expression")
+		default:
+			panic(internalError())
+		}
+	}
+}
+
+func (c *context) push(mode mode) {
+	c.modes = append(c.modes, c.mode)
+	c.mode = mode
+}
+
+func (c *context) pop() {
+	n := len(c.modes)
+	c.mode = c.modes[n-1]
+	c.modes = c.modes[:n-1]
+}
+
+func (c *context) statFile(name string, sys bool) (os.FileInfo, error) {
+	fs := c.cfg.Config3.Filesystem
+	if fs == nil {
+		fs = LocalFS()
+	}
+	return fs.Stat(name, sys)
+}
+
+func (c *context) openFile(name string, sys bool) (io.ReadCloser, error) {
+	fs := c.cfg.Config3.Filesystem
+	if fs == nil {
+		fs = LocalFS()
+	}
+	return fs.Open(name, sys)
+}
+
+// HostConfig returns the system C preprocessor/compiler configuration, or an
+// error, if any.  The configuration is obtained by running the command named
+// by the cpp argumnent or "cpp" when it's empty.  For the predefined macros
+// list the '-dM' options is added. For the include paths lists, the option
+// '-v' is added and the output is parsed to extract the "..." include and
+// <...> include paths. To add any other options to cpp, list them in opts.
+//
+// The function relies on a POSIX/GCC compatible C preprocessor installed.
+// Execution of HostConfig is not free, so caching of the results is
+// recommended.
+func HostConfig(cpp string, opts ...string) (predefined string, includePaths, sysIncludePaths []string, err error) {
+	if cpp == "" {
+		cpp = "cpp"
+	}
+	args := append(append([]string{"-dM"}, opts...), os.DevNull)
+	pre, err := exec.Command(cpp, args...).Output()
+	if err != nil {
+		return "", nil, nil, err
+	}
+
+	args = append(append([]string{"-v"}, opts...), os.DevNull)
+	out, err := exec.Command(cpp, args...).CombinedOutput()
+	if err != nil {
+		return "", nil, nil, err
+	}
+
+	sep := "\n"
+	if env("GOOS", runtime.GOOS) == "windows" {
+		sep = "\r\n"
+	}
+
+	a := strings.Split(string(out), sep)
+	for i := 0; i < len(a); {
+		switch a[i] {
+		case "#include \"...\" search starts here:":
+		loop:
+			for i = i + 1; i < len(a); {
+				switch v := a[i]; {
+				case strings.HasPrefix(v, "#") || v == "End of search list.":
+					break loop
+				default:
+					includePaths = append(includePaths, strings.TrimSpace(v))
+					i++
+				}
+			}
+		case "#include <...> search starts here:":
+			for i = i + 1; i < len(a); {
+				switch v := a[i]; {
+				case strings.HasPrefix(v, "#") || v == "End of search list.":
+					return string(pre), includePaths, sysIncludePaths, nil
+				default:
+					sysIncludePaths = append(sysIncludePaths, strings.TrimSpace(v))
+					i++
+				}
+			}
+		default:
+			i++
+		}
+	}
+	return "", nil, nil, fmt.Errorf("failed parsing %s -v output", cpp)
+}
+
+func env(key, val string) string {
+	if s := os.Getenv(key); s != "" {
+		return s
+	}
+
+	return val
+}
+
+// Token is a grammar terminal.
+type Token struct {
+	Rune  rune     // ';' or IDENTIFIER etc.
+	Sep   StringID // If Config3.PreserveWhiteSpace is in effect: All preceding white space combined, including comments.
+	Value StringID // ";" or "foo" etc.
+	Src   StringID
+	file  *tokenFile
+	macro StringID
+	pos   int32
+	seq   int32
+}
+
+// Seq returns t's sequential number.
+//
+// Comparing positions as in 'before', 'after' is complicated as tokens in a
+// translation unit usually come from more than one source file. Macro
+// expansion further complicates that. The solution is sequentially numbering
+// the tokens as they are finally seen by the parser, so the usual arithmetic
+// '<', '>' operators can be used for that purpose.
+func (t Token) Seq() int { return int(t.seq) }
+
+// Macro returns the name of a macro that expanded to this token, if any.
+func (t *Token) Macro() StringID { return t.macro }
+
+// String implements fmt.Stringer.
+func (t Token) String() string { return t.Value.String() }
+
+// Position implements Node.
+func (t *Token) Position() (r token.Position) {
+	if t.pos != 0 && t.file != nil {
+		r = t.file.PositionFor(token.Pos(t.pos), true)
+	}
+	return r
+}
+
+func tokStr(toks interface{}, sep string) string {
+	var b strings.Builder
+	switch x := toks.(type) {
+	case []token3:
+		for i, v := range x {
+			if i != 0 {
+				b.WriteString(sep)
+			}
+			b.WriteString(v.String())
+		}
+	case []token4:
+		for i, v := range x {
+			if i != 0 {
+				b.WriteString(sep)
+			}
+			b.WriteString(v.String())
+		}
+	case []cppToken:
+		for i, v := range x {
+			if i != 0 {
+				b.WriteString(sep)
+			}
+			b.WriteString(v.String())
+		}
+	case []Token:
+		for i, v := range x {
+			if i != 0 {
+				b.WriteString(sep)
+			}
+			b.WriteString(v.String())
+		}
+	default:
+		panic(internalError())
+	}
+	return b.String()
+}
+
+func internalError() int {
+	panic(fmt.Errorf("%v: internal error", origin(2)))
+}
+
+func internalErrorf(s string, args ...interface{}) int {
+	s = fmt.Sprintf(s, args)
+	panic(fmt.Errorf("%v: %s", origin(2), s))
+}
+
+func detectMingw(s string) bool {
+	return strings.Contains(s, "#define __MINGW")
+}
+
+func nodeSource(n ...Node) (r string) {
+	if len(n) == 0 {
+		return ""
+	}
+
+	var a []*Token
+	for _, v := range n {
+		Inspect(v, func(n Node, _ bool) bool {
+			if x, ok := n.(*Token); ok && x.Seq() != 0 {
+				a = append(a, x)
+			}
+			return true
+		})
+	}
+	sort.Slice(a, func(i, j int) bool {
+		return a[i].Seq() < a[j].Seq()
+	})
+	w := 0
+	seq := -1
+	for _, v := range a {
+		if n := v.Seq(); n != seq {
+			seq = n
+			a[w] = v
+			w++
+		}
+	}
+	a = a[:w]
+	var b strings.Builder
+	for _, v := range a {
+		b.WriteString(v.Sep.String())
+		b.WriteString(v.Src.String())
+	}
+	return b.String()
+}