560 lines
11 KiB
Go
560 lines
11 KiB
Go
|
// Copyright 2022 The Gc 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 gc // modernc.org/gc/v3
|
||
|
|
||
|
import (
|
||
|
"fmt"
|
||
|
"go/token"
|
||
|
"math"
|
||
|
"os"
|
||
|
"path/filepath"
|
||
|
"runtime"
|
||
|
"sort"
|
||
|
"strconv"
|
||
|
"strings"
|
||
|
"sync"
|
||
|
|
||
|
"github.com/dustin/go-humanize"
|
||
|
)
|
||
|
|
||
|
// The list of tokens.
|
||
|
const (
|
||
|
// Special tokens
|
||
|
ILLEGAL = token.ILLEGAL
|
||
|
EOF = token.EOF
|
||
|
COMMENT = token.COMMENT
|
||
|
|
||
|
// Identifiers and basic type literals
|
||
|
// (these tokens stand for classes of literals)
|
||
|
IDENT = token.IDENT // main
|
||
|
INT = token.INT // 12345
|
||
|
FLOAT = token.FLOAT // 123.45
|
||
|
IMAG = token.IMAG // 123.45i
|
||
|
CHAR = token.CHAR // 'a'
|
||
|
STRING = token.STRING // "abc"
|
||
|
|
||
|
// Operators and delimiters
|
||
|
ADD = token.ADD // +
|
||
|
SUB = token.SUB // -
|
||
|
MUL = token.MUL // *
|
||
|
QUO = token.QUO // /
|
||
|
REM = token.REM // %
|
||
|
|
||
|
AND = token.AND // &
|
||
|
OR = token.OR // |
|
||
|
XOR = token.XOR // ^
|
||
|
SHL = token.SHL // <<
|
||
|
SHR = token.SHR // >>
|
||
|
AND_NOT = token.AND_NOT // &^
|
||
|
|
||
|
ADD_ASSIGN = token.ADD_ASSIGN // +=
|
||
|
SUB_ASSIGN = token.SUB_ASSIGN // -=
|
||
|
MUL_ASSIGN = token.MUL_ASSIGN // *=
|
||
|
QUO_ASSIGN = token.QUO_ASSIGN // /=
|
||
|
REM_ASSIGN = token.REM_ASSIGN // %=
|
||
|
|
||
|
AND_ASSIGN = token.AND_ASSIGN // &=
|
||
|
OR_ASSIGN = token.OR_ASSIGN // |=
|
||
|
XOR_ASSIGN = token.XOR_ASSIGN // ^=
|
||
|
SHL_ASSIGN = token.SHL_ASSIGN // <<=
|
||
|
SHR_ASSIGN = token.SHR_ASSIGN // >>=
|
||
|
AND_NOT_ASSIGN = token.AND_NOT_ASSIGN // &^=
|
||
|
|
||
|
LAND = token.LAND // &&
|
||
|
LOR = token.LOR // ||
|
||
|
ARROW = token.ARROW // <-
|
||
|
INC = token.INC // ++
|
||
|
DEC = token.DEC // --
|
||
|
|
||
|
EQL = token.EQL // ==
|
||
|
LSS = token.LSS // <
|
||
|
GTR = token.GTR // >
|
||
|
ASSIGN = token.ASSIGN // =
|
||
|
NOT = token.NOT // !
|
||
|
|
||
|
NEQ = token.NEQ // !=
|
||
|
LEQ = token.LEQ // <=
|
||
|
GEQ = token.GEQ // >=
|
||
|
DEFINE = token.DEFINE // :=
|
||
|
ELLIPSIS = token.ELLIPSIS // ...
|
||
|
|
||
|
LPAREN = token.LPAREN // (
|
||
|
LBRACK = token.LBRACK // [
|
||
|
LBRACE = token.LBRACE // {
|
||
|
COMMA = token.COMMA // ,
|
||
|
PERIOD = token.PERIOD // .
|
||
|
|
||
|
RPAREN = token.RPAREN // )
|
||
|
RBRACK = token.RBRACK // ]
|
||
|
RBRACE = token.RBRACE // }
|
||
|
SEMICOLON = token.SEMICOLON // ;
|
||
|
COLON = token.COLON // :
|
||
|
|
||
|
// Keywords
|
||
|
BREAK = token.BREAK
|
||
|
CASE = token.CASE
|
||
|
CHAN = token.CHAN
|
||
|
CONST = token.CONST
|
||
|
CONTINUE = token.CONTINUE
|
||
|
|
||
|
DEFAULT = token.DEFAULT
|
||
|
DEFER = token.DEFER
|
||
|
ELSE = token.ELSE
|
||
|
FALLTHROUGH = token.FALLTHROUGH
|
||
|
FOR = token.FOR
|
||
|
|
||
|
FUNC = token.FUNC
|
||
|
GO = token.GO
|
||
|
GOTO = token.GOTO
|
||
|
IF = token.IF
|
||
|
IMPORT = token.IMPORT
|
||
|
|
||
|
INTERFACE = token.INTERFACE
|
||
|
MAP = token.MAP
|
||
|
PACKAGE = token.PACKAGE
|
||
|
RANGE = token.RANGE
|
||
|
RETURN = token.RETURN
|
||
|
|
||
|
SELECT = token.SELECT
|
||
|
STRUCT = token.STRUCT
|
||
|
SWITCH = token.SWITCH
|
||
|
TYPE = token.TYPE
|
||
|
VAR = token.VAR
|
||
|
|
||
|
// additional tokens, handled in an ad-hoc manner
|
||
|
TILDE = token.TILDE
|
||
|
)
|
||
|
|
||
|
var (
|
||
|
trcTODOs bool
|
||
|
extendedErrors bool
|
||
|
)
|
||
|
|
||
|
// origin returns caller's short position, skipping skip frames.
|
||
|
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:]
|
||
|
}
|
||
|
if strings.HasPrefix(fns, "func") {
|
||
|
num := true
|
||
|
for _, c := range fns[len("func"):] {
|
||
|
if c < '0' || c > '9' {
|
||
|
num = false
|
||
|
break
|
||
|
}
|
||
|
}
|
||
|
if num {
|
||
|
return origin(skip + 2)
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
return fmt.Sprintf("%s:%d:%s", filepath.Base(fn), fl, fns)
|
||
|
}
|
||
|
|
||
|
// todo prints and returns caller's position and an optional message tagged with TODO. Output goes to stderr.
|
||
|
//
|
||
|
//lint:ignore U1000 whatever
|
||
|
func todo(s string, args ...interface{}) string {
|
||
|
switch {
|
||
|
case s == "":
|
||
|
s = fmt.Sprintf(strings.Repeat("%v ", len(args)), args...)
|
||
|
default:
|
||
|
s = fmt.Sprintf(s, args...)
|
||
|
}
|
||
|
r := fmt.Sprintf("%s\n\tTODO (%s)", origin(2), s)
|
||
|
// fmt.Fprintf(os.Stderr, "%s\n", r)
|
||
|
// os.Stdout.Sync()
|
||
|
return r
|
||
|
}
|
||
|
|
||
|
// trc prints and returns caller's position and an optional message tagged with TRC. Output goes to stderr.
|
||
|
//
|
||
|
//lint:ignore U1000 whatever
|
||
|
func trc(s string, args ...interface{}) string {
|
||
|
switch {
|
||
|
case s == "":
|
||
|
s = fmt.Sprintf(strings.Repeat("%v ", len(args)), args...)
|
||
|
default:
|
||
|
s = fmt.Sprintf(s, args...)
|
||
|
}
|
||
|
r := fmt.Sprintf("%s: TRC (%s)", origin(2), s)
|
||
|
fmt.Fprintf(os.Stderr, "%s\n", r)
|
||
|
os.Stderr.Sync()
|
||
|
return r
|
||
|
}
|
||
|
|
||
|
func extractPos(s string) (p token.Position, ok bool) {
|
||
|
var prefix string
|
||
|
if len(s) > 1 && s[1] == ':' { // c:\foo
|
||
|
prefix = s[:2]
|
||
|
s = s[2:]
|
||
|
}
|
||
|
// "testdata/parser/bug/001.c:1193: ..."
|
||
|
a := strings.Split(s, ":")
|
||
|
// ["testdata/parser/bug/001.c" "1193" "..."]
|
||
|
if len(a) < 2 {
|
||
|
return p, false
|
||
|
}
|
||
|
|
||
|
line, err := strconv.Atoi(a[1])
|
||
|
if err != nil {
|
||
|
return p, false
|
||
|
}
|
||
|
|
||
|
col, err := strconv.Atoi(a[2])
|
||
|
if err != nil {
|
||
|
col = 1
|
||
|
}
|
||
|
|
||
|
return token.Position{Filename: prefix + a[0], Line: line, Column: col}, true
|
||
|
}
|
||
|
|
||
|
// errorf constructs an error value. If extendedErrors is true, the error will
|
||
|
// contain its origin.
|
||
|
func errorf(s string, args ...interface{}) error {
|
||
|
switch {
|
||
|
case s == "":
|
||
|
s = fmt.Sprintf(strings.Repeat("%v ", len(args)), args...)
|
||
|
default:
|
||
|
s = fmt.Sprintf(s, args...)
|
||
|
}
|
||
|
if trcTODOs && strings.HasPrefix(s, "TODO") {
|
||
|
fmt.Fprintf(os.Stderr, "%s (%v)\n", s, origin(2))
|
||
|
os.Stderr.Sync()
|
||
|
}
|
||
|
switch {
|
||
|
case extendedErrors:
|
||
|
return fmt.Errorf("%s (%v: %v: %v)", s, origin(4), origin(3), origin(2))
|
||
|
default:
|
||
|
return fmt.Errorf("%s", s)
|
||
|
}
|
||
|
}
|
||
|
|
||
|
func tokSource(t token.Token) string {
|
||
|
switch t {
|
||
|
case ILLEGAL:
|
||
|
return "ILLEGAL"
|
||
|
case EOF:
|
||
|
return "EOF"
|
||
|
case COMMENT:
|
||
|
return "COMMENT"
|
||
|
case IDENT:
|
||
|
return "IDENT"
|
||
|
case INT:
|
||
|
return "INT"
|
||
|
case FLOAT:
|
||
|
return "FLOAT"
|
||
|
case IMAG:
|
||
|
return "IMAG"
|
||
|
case CHAR:
|
||
|
return "CHAR"
|
||
|
case STRING:
|
||
|
return "STRING"
|
||
|
case ADD:
|
||
|
return "ADD"
|
||
|
case SUB:
|
||
|
return "SUB"
|
||
|
case MUL:
|
||
|
return "MUL"
|
||
|
case QUO:
|
||
|
return "QUO"
|
||
|
case REM:
|
||
|
return "REM"
|
||
|
case AND:
|
||
|
return "AND"
|
||
|
case OR:
|
||
|
return "OR"
|
||
|
case XOR:
|
||
|
return "XOR"
|
||
|
case SHL:
|
||
|
return "SHL"
|
||
|
case SHR:
|
||
|
return "SHR"
|
||
|
case AND_NOT:
|
||
|
return "AND_NOT"
|
||
|
case ADD_ASSIGN:
|
||
|
return "ADD_ASSIGN"
|
||
|
case SUB_ASSIGN:
|
||
|
return "SUB_ASSIGN"
|
||
|
case MUL_ASSIGN:
|
||
|
return "MUL_ASSIGN"
|
||
|
case QUO_ASSIGN:
|
||
|
return "QUO_ASSIGN"
|
||
|
case REM_ASSIGN:
|
||
|
return "REM_ASSIGN"
|
||
|
case AND_ASSIGN:
|
||
|
return "AND_ASSIGN"
|
||
|
case OR_ASSIGN:
|
||
|
return "OR_ASSIGN"
|
||
|
case XOR_ASSIGN:
|
||
|
return "XOR_ASSIGN"
|
||
|
case SHL_ASSIGN:
|
||
|
return "SHL_ASSIGN"
|
||
|
case SHR_ASSIGN:
|
||
|
return "SHR_ASSIGN"
|
||
|
case AND_NOT_ASSIGN:
|
||
|
return "AND_NOT_ASSIGN"
|
||
|
case LAND:
|
||
|
return "LAND"
|
||
|
case LOR:
|
||
|
return "LOR"
|
||
|
case ARROW:
|
||
|
return "ARROW"
|
||
|
case INC:
|
||
|
return "INC"
|
||
|
case DEC:
|
||
|
return "DEC"
|
||
|
case EQL:
|
||
|
return "EQL"
|
||
|
case LSS:
|
||
|
return "LSS"
|
||
|
case GTR:
|
||
|
return "GTR"
|
||
|
case ASSIGN:
|
||
|
return "ASSIGN"
|
||
|
case NOT:
|
||
|
return "NOT"
|
||
|
case NEQ:
|
||
|
return "NEQ"
|
||
|
case LEQ:
|
||
|
return "LEQ"
|
||
|
case GEQ:
|
||
|
return "GEQ"
|
||
|
case DEFINE:
|
||
|
return "DEFINE"
|
||
|
case ELLIPSIS:
|
||
|
return "ELLIPSIS"
|
||
|
case LPAREN:
|
||
|
return "LPAREN"
|
||
|
case LBRACK:
|
||
|
return "LBRACK"
|
||
|
case LBRACE:
|
||
|
return "LBRACE"
|
||
|
case COMMA:
|
||
|
return "COMMA"
|
||
|
case PERIOD:
|
||
|
return "PERIOD"
|
||
|
case RPAREN:
|
||
|
return "RPAREN"
|
||
|
case RBRACK:
|
||
|
return "RBRACK"
|
||
|
case RBRACE:
|
||
|
return "RBRACE"
|
||
|
case SEMICOLON:
|
||
|
return "SEMICOLON"
|
||
|
case COLON:
|
||
|
return "COLON"
|
||
|
case BREAK:
|
||
|
return "BREAK"
|
||
|
case CASE:
|
||
|
return "CASE"
|
||
|
case CHAN:
|
||
|
return "CHAN"
|
||
|
case CONST:
|
||
|
return "CONST"
|
||
|
case CONTINUE:
|
||
|
return "CONTINUE"
|
||
|
case DEFAULT:
|
||
|
return "DEFAULT"
|
||
|
case DEFER:
|
||
|
return "DEFER"
|
||
|
case ELSE:
|
||
|
return "ELSE"
|
||
|
case FALLTHROUGH:
|
||
|
return "FALLTHROUGH"
|
||
|
case FOR:
|
||
|
return "FOR"
|
||
|
case FUNC:
|
||
|
return "FUNC"
|
||
|
case GO:
|
||
|
return "GO"
|
||
|
case GOTO:
|
||
|
return "GOTO"
|
||
|
case IF:
|
||
|
return "IF"
|
||
|
case IMPORT:
|
||
|
return "IMPORT"
|
||
|
case INTERFACE:
|
||
|
return "INTERFACE"
|
||
|
case MAP:
|
||
|
return "MAP"
|
||
|
case PACKAGE:
|
||
|
return "PACKAGE"
|
||
|
case RANGE:
|
||
|
return "RANGE"
|
||
|
case RETURN:
|
||
|
return "RETURN"
|
||
|
case SELECT:
|
||
|
return "SELECT"
|
||
|
case STRUCT:
|
||
|
return "STRUCT"
|
||
|
case SWITCH:
|
||
|
return "SWITCH"
|
||
|
case TYPE:
|
||
|
return "TYPE"
|
||
|
case VAR:
|
||
|
return "VAR"
|
||
|
case TILDE:
|
||
|
return "TILDE"
|
||
|
default:
|
||
|
panic(todo("", int(t), t))
|
||
|
}
|
||
|
}
|
||
|
|
||
|
type data struct {
|
||
|
line int
|
||
|
cases int
|
||
|
cnt int
|
||
|
}
|
||
|
|
||
|
type analyzer struct {
|
||
|
sync.Mutex
|
||
|
m map[int]*data // line: data
|
||
|
}
|
||
|
|
||
|
func newAnalyzer() *analyzer {
|
||
|
return &analyzer{m: map[int]*data{}}
|
||
|
}
|
||
|
|
||
|
func (a *analyzer) record(line, cnt int) {
|
||
|
d := a.m[line]
|
||
|
if d == nil {
|
||
|
d = &data{line: line}
|
||
|
a.m[line] = d
|
||
|
}
|
||
|
d.cases++
|
||
|
d.cnt += cnt
|
||
|
}
|
||
|
|
||
|
func (a *analyzer) merge(b *analyzer) {
|
||
|
a.Lock()
|
||
|
defer a.Unlock()
|
||
|
|
||
|
for k, v := range b.m {
|
||
|
d := a.m[k]
|
||
|
if d == nil {
|
||
|
d = &data{line: k}
|
||
|
a.m[k] = d
|
||
|
}
|
||
|
d.cases += v.cases
|
||
|
d.cnt += v.cnt
|
||
|
}
|
||
|
}
|
||
|
|
||
|
func (a *analyzer) report() string {
|
||
|
var rows []*data
|
||
|
for _, v := range a.m {
|
||
|
rows = append(rows, v)
|
||
|
}
|
||
|
sort.Slice(rows, func(i, j int) bool {
|
||
|
a := rows[i]
|
||
|
b := rows[j]
|
||
|
if a.cases < b.cases {
|
||
|
return true
|
||
|
}
|
||
|
|
||
|
if a.cases > b.cases {
|
||
|
return false
|
||
|
}
|
||
|
|
||
|
// a.cases == b.cases
|
||
|
if a.cnt < b.cnt {
|
||
|
return true
|
||
|
}
|
||
|
|
||
|
if a.cnt > b.cnt {
|
||
|
return false
|
||
|
}
|
||
|
|
||
|
// a.cnt == b.cnt
|
||
|
return a.line < b.line
|
||
|
})
|
||
|
var b strings.Builder
|
||
|
var cases, cnt int
|
||
|
for _, row := range rows {
|
||
|
cases += row.cases
|
||
|
cnt += row.cnt
|
||
|
avg := float64(row.cnt) / float64(row.cases)
|
||
|
fmt.Fprintf(&b, "parser.go:%d:\t%16s %16s %8.1f\n", row.line, h(row.cases), h(row.cnt), avg)
|
||
|
}
|
||
|
avg := float64(cnt) / float64(cases)
|
||
|
fmt.Fprintf(&b, "<total>\t\t%16s %16s %8.1f\n", h(cases), h(cnt), avg)
|
||
|
return b.String()
|
||
|
}
|
||
|
|
||
|
func h(v interface{}) string {
|
||
|
switch x := v.(type) {
|
||
|
case int:
|
||
|
return humanize.Comma(int64(x))
|
||
|
case int32:
|
||
|
return humanize.Comma(int64(x))
|
||
|
case int64:
|
||
|
return humanize.Comma(x)
|
||
|
case uint32:
|
||
|
return humanize.Comma(int64(x))
|
||
|
case uint64:
|
||
|
if x <= math.MaxInt64 {
|
||
|
return humanize.Comma(int64(x))
|
||
|
}
|
||
|
|
||
|
return "-" + humanize.Comma(-int64(x))
|
||
|
}
|
||
|
return fmt.Sprint(v)
|
||
|
}
|
||
|
|
||
|
type parallel struct {
|
||
|
limiter chan struct{}
|
||
|
}
|
||
|
|
||
|
func newParallel() *parallel {
|
||
|
return ¶llel{
|
||
|
limiter: make(chan struct{}, runtime.GOMAXPROCS(0)),
|
||
|
}
|
||
|
}
|
||
|
|
||
|
func (p *parallel) throttle(f func()) {
|
||
|
p.limiter <- struct{}{}
|
||
|
|
||
|
defer func() {
|
||
|
<-p.limiter
|
||
|
}()
|
||
|
|
||
|
f()
|
||
|
}
|
||
|
|
||
|
func extraTags(verMajor, verMinor int, goos, goarch string) (r []string) {
|
||
|
// https://github.com/golang/go/commit/eeb7899137cda1c2cd60dab65ff41f627436db5b
|
||
|
//
|
||
|
// In Go 1.17 we added register ABI on AMD64 on Linux/macOS/Windows
|
||
|
// as a GOEXPERIMENT, on by default. In Go 1.18, we commit to always
|
||
|
// enabling register ABI on AMD64.
|
||
|
//
|
||
|
// Now "go build" for AMD64 always have goexperiment.regabi* tags
|
||
|
// set. However, at bootstrapping cmd/dist does not set the tags
|
||
|
// when building go_bootstrap. For this to work, unfortunately, we
|
||
|
// need to hard-code AMD64 to use register ABI in runtime code.
|
||
|
if verMajor == 1 {
|
||
|
switch {
|
||
|
case verMinor == 17:
|
||
|
switch goos {
|
||
|
case "linux", "darwin", "windows":
|
||
|
if goarch == "amd64" {
|
||
|
r = append(r, "goexperiment.regabiargs", "goexperiment.regabiwrappers")
|
||
|
}
|
||
|
}
|
||
|
case verMinor >= 18:
|
||
|
if goarch == "amd64" {
|
||
|
r = append(r, "goexperiment.regabiargs", "goexperiment.regabiwrappers")
|
||
|
}
|
||
|
}
|
||
|
}
|
||
|
return r
|
||
|
}
|