pkg/ast: support char constants

Frequently it's useful to do something like:

int8['a':'z']
punctuation = ',', '-', ':'

9 files changed, 65 insertions, 22 deletions

diff --git a/pkg/ast/ast.go b/pkg/ast/ast.go
index 723d6d9c4..610d22f30 100644
--- a/pkg/ast/ast.go
+++ b/pkg/ast/ast.go
@@ -169,11 +169,19 @@ func (n *String) Info() (Pos, string, string) {
 	return n.Pos, tok2str[tokString], ""
 }
 
+type intFmt int
+
+const (
+	intFmtDec intFmt = iota
+	intFmtHex
+	intFmtChar
+)
+
 type Int struct {
 	Pos Pos
 	// Only one of Value, Ident, CExpr is filled.
 	Value    uint64
-	ValueHex bool // says if value was in hex (for formatting)
+	valueFmt intFmt
 	Ident    string
 	CExpr    string
 }
@@ -186,7 +194,7 @@ type Type struct {
 	Pos Pos
 	// Only one of Value, Ident, String is filled.
 	Value     uint64
-	ValueHex  bool
+	valueFmt  intFmt
 	Ident     string
 	String    string
 	HasString bool
@@ -194,7 +202,7 @@ type Type struct {
 	HasColon  bool
 	Pos2      Pos
 	Value2    uint64
-	Value2Hex bool
+	value2Fmt intFmt
 	Ident2    string
 	Args      []*Type
 }
diff --git a/pkg/ast/clone.go b/pkg/ast/clone.go
index 4c938b807..9ccc3ea76 100644
--- a/pkg/ast/clone.go
+++ b/pkg/ast/clone.go
@@ -165,7 +165,7 @@ func (n *Int) Clone() Node {
 	return &Int{
 		Pos:      n.Pos,
 		Value:    n.Value,
-		ValueHex: n.ValueHex,
+		valueFmt: n.valueFmt,
 		Ident:    n.Ident,
 		CExpr:    n.CExpr,
 	}
@@ -179,14 +179,14 @@ func (n *Type) Clone() Node {
 	return &Type{
 		Pos:       n.Pos,
 		Value:     n.Value,
-		ValueHex:  n.ValueHex,
+		valueFmt:  n.valueFmt,
 		Ident:     n.Ident,
 		String:    n.String,
 		HasString: n.HasString,
 		HasColon:  n.HasColon,
 		Pos2:      n.Pos2,
 		Value2:    n.Value2,
-		Value2Hex: n.Value2Hex,
+		value2Fmt: n.value2Fmt,
 		Ident2:    n.Ident2,
 		Args:      args,
 	}
diff --git a/pkg/ast/format.go b/pkg/ast/format.go
index cf7edf975..5022d5da1 100644
--- a/pkg/ast/format.go
+++ b/pkg/ast/format.go
@@ -159,14 +159,14 @@ func fmtType(t *Type) string {
 	case t.HasString:
 		v = fmt.Sprintf("\"%v\"", t.String)
 	default:
-		v = fmtIntValue(t.Value, t.ValueHex)
+		v = fmtIntValue(t.Value, t.valueFmt)
 	}
 	if t.HasColon {
 		switch {
 		case t.Ident2 != "":
 			v += fmt.Sprintf(":%v", t.Ident2)
 		default:
-			v += fmt.Sprintf(":%v", fmtIntValue(t.Value2, t.Value2Hex))
+			v += fmt.Sprintf(":%v", fmtIntValue(t.Value2, t.value2Fmt))
 		}
 	}
 	v += fmtTypeList(t.Args)
@@ -206,15 +206,21 @@ func fmtInt(i *Int) string {
 	case i.CExpr != "":
 		return fmt.Sprintf("%v", i.CExpr)
 	default:
-		return fmtIntValue(i.Value, i.ValueHex)
+		return fmtIntValue(i.Value, i.valueFmt)
 	}
 }
 
-func fmtIntValue(v uint64, hex bool) string {
-	if hex {
+func fmtIntValue(v uint64, format intFmt) string {
+	switch format {
+	case intFmtDec:
+		return fmt.Sprint(v)
+	case intFmtHex:
 		return fmt.Sprintf("0x%x", v)
+	case intFmtChar:
+		return fmt.Sprintf("'%c'", v)
+	default:
+		panic(fmt.Sprintf("unknown int format %v", format))
 	}
-	return fmt.Sprint(v)
 }
 
 func comma(i int, or string) string {
diff --git a/pkg/ast/parser.go b/pkg/ast/parser.go
index fb3327223..e4c5d8742 100644
--- a/pkg/ast/parser.go
+++ b/pkg/ast/parser.go
@@ -413,7 +413,7 @@ func (p *parser) parseType() *Type {
 	switch p.tok {
 	case tokInt:
 		allowColon = true
-		arg.Value, arg.ValueHex = p.parseIntValue()
+		arg.Value, arg.valueFmt = p.parseIntValue()
 	case tokIdent:
 		allowColon = true
 		arg.Ident = p.lit
@@ -429,7 +429,7 @@ func (p *parser) parseType() *Type {
 		arg.Pos2 = p.pos
 		switch p.tok {
 		case tokInt:
-			arg.Value2, arg.Value2Hex = p.parseIntValue()
+			arg.Value2, arg.value2Fmt = p.parseIntValue()
 		case tokIdent:
 			arg.Ident2 = p.lit
 		default:
@@ -479,7 +479,7 @@ func (p *parser) parseInt() *Int {
 	}
 	switch p.tok {
 	case tokInt:
-		i.Value, i.ValueHex = p.parseIntValue()
+		i.Value, i.valueFmt = p.parseIntValue()
 	case tokIdent:
 		i.Ident = p.lit
 	default:
@@ -489,13 +489,16 @@ func (p *parser) parseInt() *Int {
 	return i
 }
 
-func (p *parser) parseIntValue() (uint64, bool) {
+func (p *parser) parseIntValue() (uint64, intFmt) {
+	if p.lit[0] == '\'' {
+		return uint64(p.lit[1]), intFmtChar
+	}
 	if v, err := strconv.ParseUint(p.lit, 10, 64); err == nil {
-		return v, false
+		return v, intFmtDec
 	}
 	if len(p.lit) > 2 && p.lit[0] == '0' && p.lit[1] == 'x' {
 		if v, err := strconv.ParseUint(p.lit[2:], 16, 64); err == nil {
-			return v, true
+			return v, intFmtHex
 		}
 	}
 	panic(fmt.Sprintf("scanner returned bad integer %q", p.lit))
diff --git a/pkg/ast/scanner.go b/pkg/ast/scanner.go
index 5ad5e5808..343c3b9ab 100644
--- a/pkg/ast/scanner.go
+++ b/pkg/ast/scanner.go
@@ -200,6 +200,17 @@ func (s *scanner) Scan() (tok token, lit string, pos Pos) {
 			s.Error(pos, fmt.Sprintf("bad integer %q", lit))
 			lit = "0"
 		}
+	case s.ch == '\'':
+		tok = tokInt
+		lit = "0"
+		s.next()
+		s.next()
+		if s.ch != '\'' {
+			s.Error(pos, "char literal is not terminated")
+			return
+		}
+		s.next()
+		lit = string(s.data[pos.Off : pos.Off+3])
 	case s.ch == '_' || s.ch >= 'a' && s.ch <= 'z' || s.ch >= 'A' && s.ch <= 'Z':
 		tok = tokIdent
 		for s.ch == '_' || s.ch == '$' ||
diff --git a/pkg/ast/testdata/all.txt b/pkg/ast/testdata/all.txt
index 7e1db7309..6c818eca6 100644
--- a/pkg/ast/testdata/all.txt
+++ b/pkg/ast/testdata/all.txt
@@ -9,6 +9,9 @@ int_flags0 = 0, 0x1, 0xab
 int_flags1 = 123ab0x			### bad integer "123ab0x"
 int_flags1 == 0, 1			### unexpected '=', expecting int, identifier, string
 int_flags = 0, "foo"			### unexpected string, expecting int, identifier
+int_flags2 = '				### char literal is not terminated
+int_flags3 = 'a				### char literal is not terminated
+int_flags3 = 'a, 1			### char literal is not terminated
 
 str_flags0 = "foo", "bar"
 str_flags1 = "non terminated		### string literal is not terminated
diff --git a/pkg/compiler/compiler_test.go b/pkg/compiler/compiler_test.go
index a5cc6a86a..c52d4d39d 100644
--- a/pkg/compiler/compiler_test.go
+++ b/pkg/compiler/compiler_test.go
@@ -5,6 +5,7 @@ package compiler
 
 import (
 	"bytes"
+	"flag"
 	"fmt"
 	"io/ioutil"
 	"path/filepath"
@@ -15,6 +16,8 @@ import (
 	"github.com/google/syzkaller/sys/targets"
 )
 
+var flagUpdate = flag.Bool("update", false, "reformat all.txt")
+
 func TestCompileAll(t *testing.T) {
 	for os, arches := range targets.List {
 		os, arches := os, arches
@@ -63,7 +66,8 @@ func TestNoErrors(t *testing.T) {
 				eh := func(pos ast.Pos, msg string) {
 					t.Logf("%v: %v", pos, msg)
 				}
-				data, err := ioutil.ReadFile(filepath.Join("testdata", name))
+				fileName := filepath.Join("testdata", name)
+				data, err := ioutil.ReadFile(fileName)
 				if err != nil {
 					t.Fatal(err)
 				}
@@ -71,6 +75,13 @@ func TestNoErrors(t *testing.T) {
 				if astDesc == nil {
 					t.Fatalf("parsing failed")
 				}
+				formatted := ast.Format(astDesc)
+				if !bytes.Equal(data, formatted) {
+					if *flagUpdate {
+						ioutil.WriteFile(fileName, formatted, 0644)
+					}
+					t.Fatalf("description is not formatted")
+				}
 				constInfo := ExtractConsts(astDesc, target, eh)
 				if constInfo == nil {
 					t.Fatalf("const extraction failed")
diff --git a/pkg/compiler/testdata/all.txt b/pkg/compiler/testdata/all.txt
index 081bb1aab..d451f5a60 100644
--- a/pkg/compiler/testdata/all.txt
+++ b/pkg/compiler/testdata/all.txt
@@ -6,6 +6,7 @@ foo$1(a int8[C1:C2])
 foo$2(a ptr[out, array[int32]])
 foo$3(a union_arg)
 foo$4() r0
+foo$5(a int8['a':'z'])
 
 resource r0[intptr]
 
@@ -139,8 +140,8 @@ type templ_struct1[C] {
 }
 
 union_with_templ_struct [
-	f1		templ_struct0[C1, type0]
-	f2		templ_struct0[C2, struct0]
+	f1	templ_struct0[C1, type0]
+	f2	templ_struct0[C2, struct0]
 ] [varlen]
 
 struct0 {
@@ -183,7 +184,6 @@ foo$templ4(a ptr[in, templ_struct1[3]])
 foo$templ5(a ptr[in, templ_struct1[3]])
 foo$templ6(a ptr[in, templ_struct4])
 
-
 # Structs.
 
 s0 {
diff --git a/pkg/compiler/testdata/errors2.txt b/pkg/compiler/testdata/errors2.txt
index 86148a563..69fa46e62 100644
--- a/pkg/compiler/testdata/errors2.txt
+++ b/pkg/compiler/testdata/errors2.txt
@@ -183,5 +183,6 @@ foo$505(a proc[20, 0])		### proc per-process values must not be 0
 foo$506(a ptr[in, array[int32, 0]])	### arrays of size 0 are not supported
 foo$507(a ptr[in, array[int32, 0:0]])	### arrays of size 0 are not supported
 foo$508(a ptr[in, string["foo", 3]])	### string value "foo\x00" exceeds buffer length 3
+foo$509(a int8['b':'a'])		### bad int range [98:97]
 
 type type500 proc[C1, 8, int8]	### values starting from 1 with step 8 overflow base type for 32 procs