unix: replace "mksyscall_solaris.pl" script with a Go program

Port mksyscall_solaris.pl Perl script to mksyscall_solaris.go.
mkall.sh script is modified to run mksyscall_solaris.go. Running
mkall.sh does not generate any git diff besides the command
name in comments of generated files.

Updates golang/go#27779

Change-Id: I773c9732a882780400a06f578c00e455e133231b
Reviewed-on: https://go-review.googlesource.com/c/162277
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Tobias Klauser <tobias.klauser@gmail.com>

unix/mkall.sh

@@ -170,7 +170,7 @@ openbsd_arm)
 	mktypes="GOARCH=$GOARCH go tool cgo -godefs -- -fsigned-char"
 	;;
 solaris_amd64)
-	mksyscall="./mksyscall_solaris.pl"
+	mksyscall="go run mksyscall_solaris.go"
 	mkerrors="$mkerrors -m64"
 	mksysnum=
 	mktypes="GOARCH=$GOARCH go tool cgo -godefs"

unix/mksyscall_solaris.go

@@ -0,0 +1,335 @@
+// Copyright 2019 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.
+
+// +build ignore
+
+/*
+ This program reads a file containing function prototypes
+ (like syscall_solaris.go) and generates system call bodies.
+ The prototypes are marked by lines beginning with "//sys"
+ and read like func declarations if //sys is replaced by func, but:
+	* The parameter lists must give a name for each argument.
+	  This includes return parameters.
+	* The parameter lists must give a type for each argument:
+	  the (x, y, z int) shorthand is not allowed.
+	* If the return parameter is an error number, it must be named err.
+	* If go func name needs to be different than its libc name,
+	* or the function is not in libc, name could be specified
+	* at the end, after "=" sign, like
+	  //sys getsockopt(s int, level int, name int, val uintptr, vallen *_Socklen) (err error) = libsocket.getsockopt
+*/
+
+package main
+
+import (
+	"bufio"
+	"flag"
+	"fmt"
+	"os"
+	"regexp"
+	"strings"
+)
+
+var (
+	b32  = flag.Bool("b32", false, "32bit big-endian")
+	l32  = flag.Bool("l32", false, "32bit little-endian")
+	tags = flag.String("tags", "", "build tags")
+)
+
+// cmdLine returns this programs's commandline arguments
+func cmdLine() string {
+	return "go run mksyscall_solaris.go " + strings.Join(os.Args[1:], " ")
+}
+
+// buildTags returns build tags
+func buildTags() string {
+	return *tags
+}
+
+// Param is function parameter
+type Param struct {
+	Name string
+	Type string
+}
+
+// usage prints the program usage
+func usage() {
+	fmt.Fprintf(os.Stderr, "usage: go run mksyscall_solaris.go [-b32 | -l32] [-tags x,y] [file ...]\n")
+	os.Exit(1)
+}
+
+// parseParamList parses parameter list and returns a slice of parameters
+func parseParamList(list string) []string {
+	list = strings.TrimSpace(list)
+	if list == "" {
+		return []string{}
+	}
+	return regexp.MustCompile(`\s*,\s*`).Split(list, -1)
+}
+
+// parseParam splits a parameter into name and type
+func parseParam(p string) Param {
+	ps := regexp.MustCompile(`^(\S*) (\S*)$`).FindStringSubmatch(p)
+	if ps == nil {
+		fmt.Fprintf(os.Stderr, "malformed parameter: %s\n", p)
+		os.Exit(1)
+	}
+	return Param{ps[1], ps[2]}
+}
+
+func main() {
+	flag.Usage = usage
+	flag.Parse()
+	if len(flag.Args()) <= 0 {
+		fmt.Fprintf(os.Stderr, "no files to parse provided\n")
+		usage()
+	}
+
+	endianness := ""
+	if *b32 {
+		endianness = "big-endian"
+	} else if *l32 {
+		endianness = "little-endian"
+	}
+
+	pack := ""
+	text := ""
+	dynimports := ""
+	linknames := ""
+	var vars []string
+	for _, path := range flag.Args() {
+		file, err := os.Open(path)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, err.Error())
+			os.Exit(1)
+		}
+		s := bufio.NewScanner(file)
+		for s.Scan() {
+			t := s.Text()
+			t = strings.TrimSpace(t)
+			t = regexp.MustCompile(`\s+`).ReplaceAllString(t, ` `)
+			if p := regexp.MustCompile(`^package (\S+)$`).FindStringSubmatch(t); p != nil && pack == "" {
+				pack = p[1]
+			}
+			nonblock := regexp.MustCompile(`^\/\/sysnb `).FindStringSubmatch(t)
+			if regexp.MustCompile(`^\/\/sys `).FindStringSubmatch(t) == nil && nonblock == nil {
+				continue
+			}
+
+			// Line must be of the form
+			//	func Open(path string, mode int, perm int) (fd int, err error)
+			// Split into name, in params, out params.
+			f := regexp.MustCompile(`^\/\/sys(nb)? (\w+)\(([^()]*)\)\s*(?:\(([^()]+)\))?\s*(?:=\s*(?:(\w*)\.)?(\w*))?$`).FindStringSubmatch(t)
+			if f == nil {
+				fmt.Fprintf(os.Stderr, "%s:%s\nmalformed //sys declaration\n", path, t)
+				os.Exit(1)
+			}
+			funct, inps, outps, modname, sysname := f[2], f[3], f[4], f[5], f[6]
+
+			// Split argument lists on comma.
+			in := parseParamList(inps)
+			out := parseParamList(outps)
+
+			inps = strings.Join(in, ", ")
+			outps = strings.Join(out, ", ")
+
+			// Try in vain to keep people from editing this file.
+			// The theory is that they jump into the middle of the file
+			// without reading the header.
+			text += "// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT\n\n"
+
+			// So file name.
+			if modname == "" {
+				modname = "libc"
+			}
+
+			// System call name.
+			if sysname == "" {
+				sysname = funct
+			}
+
+			// System call pointer variable name.
+			sysvarname := fmt.Sprintf("proc%s", sysname)
+
+			strconvfunc := "BytePtrFromString"
+			strconvtype := "*byte"
+
+			sysname = strings.ToLower(sysname) // All libc functions are lowercase.
+
+			// Runtime import of function to allow cross-platform builds.
+			dynimports += fmt.Sprintf("//go:cgo_import_dynamic libc_%s %s \"%s.so\"\n", sysname, sysname, modname)
+			// Link symbol to proc address variable.
+			linknames += fmt.Sprintf("//go:linkname %s libc_%s\n", sysvarname, sysname)
+			// Library proc address variable.
+			vars = append(vars, sysvarname)
+
+			// Go function header.
+			outlist := strings.Join(out, ", ")
+			if outlist != "" {
+				outlist = fmt.Sprintf(" (%s)", outlist)
+			}
+			if text != "" {
+				text += "\n"
+			}
+			text += fmt.Sprintf("func %s(%s)%s {\n", funct, strings.Join(in, ", "), outlist)
+
+			// Check if err return available
+			errvar := ""
+			for _, param := range out {
+				p := parseParam(param)
+				if p.Type == "error" {
+					errvar = p.Name
+					continue
+				}
+			}
+
+			// Prepare arguments to Syscall.
+			var args []string
+			n := 0
+			for _, param := range in {
+				p := parseParam(param)
+				if regexp.MustCompile(`^\*`).FindStringSubmatch(p.Type) != nil {
+					args = append(args, "uintptr(unsafe.Pointer("+p.Name+"))")
+				} else if p.Type == "string" && errvar != "" {
+					text += fmt.Sprintf("\tvar _p%d %s\n", n, strconvtype)
+					text += fmt.Sprintf("\t_p%d, %s = %s(%s)\n", n, errvar, strconvfunc, p.Name)
+					text += fmt.Sprintf("\tif %s != nil {\n\t\treturn\n\t}\n", errvar)
+					args = append(args, fmt.Sprintf("uintptr(unsafe.Pointer(_p%d))", n))
+					n++
+				} else if p.Type == "string" {
+					fmt.Fprintf(os.Stderr, path+":"+funct+" uses string arguments, but has no error return\n")
+					text += fmt.Sprintf("\tvar _p%d %s\n", n, strconvtype)
+					text += fmt.Sprintf("\t_p%d, _ = %s(%s)\n", n, strconvfunc, p.Name)
+					args = append(args, fmt.Sprintf("uintptr(unsafe.Pointer(_p%d))", n))
+					n++
+				} else if s := regexp.MustCompile(`^\[\](.*)`).FindStringSubmatch(p.Type); s != nil {
+					// Convert slice into pointer, length.
+					// Have to be careful not to take address of &a[0] if len == 0:
+					// pass nil in that case.
+					text += fmt.Sprintf("\tvar _p%d *%s\n", n, s[1])
+					text += fmt.Sprintf("\tif len(%s) > 0 {\n\t\t_p%d = &%s[0]\n\t}\n", p.Name, n, p.Name)
+					args = append(args, fmt.Sprintf("uintptr(unsafe.Pointer(_p%d))", n), fmt.Sprintf("uintptr(len(%s))", p.Name))
+					n++
+				} else if p.Type == "int64" && endianness != "" {
+					if endianness == "big-endian" {
+						args = append(args, fmt.Sprintf("uintptr(%s>>32)", p.Name), fmt.Sprintf("uintptr(%s)", p.Name))
+					} else {
+						args = append(args, fmt.Sprintf("uintptr(%s)", p.Name), fmt.Sprintf("uintptr(%s>>32)", p.Name))
+					}
+				} else if p.Type == "bool" {
+					text += fmt.Sprintf("\tvar _p%d uint32\n", n)
+					text += fmt.Sprintf("\tif %s {\n\t\t_p%d = 1\n\t} else {\n\t\t_p%d = 0\n\t}\n", p.Name, n, n)
+					args = append(args, fmt.Sprintf("uintptr(_p%d)", n))
+					n++
+				} else {
+					args = append(args, fmt.Sprintf("uintptr(%s)", p.Name))
+				}
+			}
+			nargs := len(args)
+
+			// Determine which form to use; pad args with zeros.
+			asm := "sysvicall6"
+			if nonblock != nil {
+				asm = "rawSysvicall6"
+			}
+			if len(args) <= 6 {
+				for len(args) < 6 {
+					args = append(args, "0")
+				}
+			} else {
+				fmt.Fprintf(os.Stderr, "%s: too many arguments to system call\n", path)
+				os.Exit(1)
+			}
+
+			// Actual call.
+			arglist := strings.Join(args, ", ")
+			call := fmt.Sprintf("%s(uintptr(unsafe.Pointer(&%s)), %d, %s)", asm, sysvarname, nargs, arglist)
+
+			// Assign return values.
+			body := ""
+			ret := []string{"_", "_", "_"}
+			doErrno := false
+			for i := 0; i < len(out); i++ {
+				p := parseParam(out[i])
+				reg := ""
+				if p.Name == "err" {
+					reg = "e1"
+					ret[2] = reg
+					doErrno = true
+				} else {
+					reg = fmt.Sprintf("r%d", i)
+					ret[i] = reg
+				}
+				if p.Type == "bool" {
+					reg = fmt.Sprintf("%d != 0", reg)
+				}
+				if p.Type == "int64" && endianness != "" {
+					// 64-bit number in r1:r0 or r0:r1.
+					if i+2 > len(out) {
+						fmt.Fprintf(os.Stderr, "%s: not enough registers for int64 return\n", path)
+						os.Exit(1)
+					}
+					if endianness == "big-endian" {
+						reg = fmt.Sprintf("int64(r%d)<<32 | int64(r%d)", i, i+1)
+					} else {
+						reg = fmt.Sprintf("int64(r%d)<<32 | int64(r%d)", i+1, i)
+					}
+					ret[i] = fmt.Sprintf("r%d", i)
+					ret[i+1] = fmt.Sprintf("r%d", i+1)
+				}
+				if reg != "e1" {
+					body += fmt.Sprintf("\t%s = %s(%s)\n", p.Name, p.Type, reg)
+				}
+			}
+			if ret[0] == "_" && ret[1] == "_" && ret[2] == "_" {
+				text += fmt.Sprintf("\t%s\n", call)
+			} else {
+				text += fmt.Sprintf("\t%s, %s, %s := %s\n", ret[0], ret[1], ret[2], call)
+			}
+			text += body
+
+			if doErrno {
+				text += "\tif e1 != 0 {\n"
+				text += "\t\terr = e1\n"
+				text += "\t}\n"
+			}
+			text += "\treturn\n"
+			text += "}\n"
+		}
+		if err := s.Err(); err != nil {
+			fmt.Fprintf(os.Stderr, err.Error())
+			os.Exit(1)
+		}
+		file.Close()
+	}
+	imp := ""
+	if pack != "unix" {
+		imp = "import \"golang.org/x/sys/unix\"\n"
+
+	}
+	vardecls := "\t" + strings.Join(vars, ",\n\t")
+	vardecls += " syscallFunc"
+	fmt.Printf(srcTemplate, cmdLine(), buildTags(), pack, imp, dynimports, linknames, vardecls, text)
+}
+
+const srcTemplate = `// %s
+// Code generated by the command above; see README.md. DO NOT EDIT.
+
+// +build %s
+
+package %s
+
+import (
+	"syscall"
+	"unsafe"
+)
+%s
+%s
+%s
+var (
+%s	
+)
+
+%s
+`

unix/mksyscall_solaris.pl

@@ -1,294 +0,0 @@
-#!/usr/bin/env perl
-# Copyright 2009 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.
-
-# This program reads a file containing function prototypes
-# (like syscall_solaris.go) and generates system call bodies.
-# The prototypes are marked by lines beginning with "//sys"
-# and read like func declarations if //sys is replaced by func, but:
-#	* The parameter lists must give a name for each argument.
-#	  This includes return parameters.
-#	* The parameter lists must give a type for each argument:
-#	  the (x, y, z int) shorthand is not allowed.
-#	* If the return parameter is an error number, it must be named err.
-#	* If go func name needs to be different than its libc name,
-#	* or the function is not in libc, name could be specified
-#	* at the end, after "=" sign, like
-#	  //sys getsockopt(s int, level int, name int, val uintptr, vallen *_Socklen) (err error) = libsocket.getsockopt
-
-use strict;
-
-my $cmdline = "mksyscall_solaris.pl " . join(' ', @ARGV);
-my $errors = 0;
-my $_32bit = "";
-my $tags = "";  # build tags
-
-binmode STDOUT;
-
-if($ARGV[0] eq "-b32") {
-	$_32bit = "big-endian";
-	shift;
-} elsif($ARGV[0] eq "-l32") {
-	$_32bit = "little-endian";
-	shift;
-}
-if($ARGV[0] eq "-tags") {
-	shift;
-	$tags = $ARGV[0];
-	shift;
-}
-
-if($ARGV[0] =~ /^-/) {
-	print STDERR "usage: mksyscall_solaris.pl [-b32 | -l32] [-tags x,y] [file ...]\n";
-	exit 1;
-}
-
-sub parseparamlist($) {
-	my ($list) = @_;
-	$list =~ s/^\s*//;
-	$list =~ s/\s*$//;
-	if($list eq "") {
-		return ();
-	}
-	return split(/\s*,\s*/, $list);
-}
-
-sub parseparam($) {
-	my ($p) = @_;
-	if($p !~ /^(\S*) (\S*)$/) {
-		print STDERR "$ARGV:$.: malformed parameter: $p\n";
-		$errors = 1;
-		return ("xx", "int");
-	}
-	return ($1, $2);
-}
-
-my $package = "";
-my $text = "";
-my $dynimports = "";
-my $linknames = "";
-my @vars = ();
-while(<>) {
-	chomp;
-	s/\s+/ /g;
-	s/^\s+//;
-	s/\s+$//;
-	$package = $1 if !$package && /^package (\S+)$/;
-	my $nonblock = /^\/\/sysnb /;
-	next if !/^\/\/sys / && !$nonblock;
-
-	# Line must be of the form
-	#	func Open(path string, mode int, perm int) (fd int, err error)
-	# Split into name, in params, out params.
-	if(!/^\/\/sys(nb)? (\w+)\(([^()]*)\)\s*(?:\(([^()]+)\))?\s*(?:=\s*(?:(\w*)\.)?(\w*))?$/) {
-		print STDERR "$ARGV:$.: malformed //sys declaration\n";
-		$errors = 1;
-		next;
-	}
-	my ($nb, $func, $in, $out, $modname, $sysname) = ($1, $2, $3, $4, $5, $6);
-
-	# Split argument lists on comma.
-	my @in = parseparamlist($in);
-	my @out = parseparamlist($out);
-
-	# Try in vain to keep people from editing this file.
-	# The theory is that they jump into the middle of the file
-	# without reading the header.
-	$text .= "// THIS FILE IS GENERATED BY THE COMMAND AT THE TOP; DO NOT EDIT\n\n";
-
-	# So file name.
-	if($modname eq "") {
-		$modname = "libc";
-	}
-
-	# System call name.
-	if($sysname eq "") {
-		$sysname = "$func";
-	}
-
-	# System call pointer variable name.
-	my $sysvarname = "proc$sysname";
-
-	my $strconvfunc = "BytePtrFromString";
-	my $strconvtype = "*byte";
-
-	$sysname =~ y/A-Z/a-z/; # All libc functions are lowercase.
-
-	# Runtime import of function to allow cross-platform builds.
-	$dynimports .= "//go:cgo_import_dynamic libc_${sysname} ${sysname} \"$modname.so\"\n";
-	# Link symbol to proc address variable.
-	$linknames .= "//go:linkname ${sysvarname} libc_${sysname}\n";
-	# Library proc address variable.
-	push @vars, $sysvarname;
-
-	# Go function header.
-	$out = join(', ', @out);
-	if($out ne "") {
-		$out = " ($out)";
-	}
-	if($text ne "") {
-		$text .= "\n"
-	}
-	$text .= sprintf "func %s(%s)%s {\n", $func, join(', ', @in), $out;
-
-	# Check if err return available
-	my $errvar = "";
-	foreach my $p (@out) {
-		my ($name, $type) = parseparam($p);
-		if($type eq "error") {
-			$errvar = $name;
-			last;
-		}
-	}
-
-	# Prepare arguments to Syscall.
-	my @args = ();
-	my $n = 0;
-	foreach my $p (@in) {
-		my ($name, $type) = parseparam($p);
-		if($type =~ /^\*/) {
-			push @args, "uintptr(unsafe.Pointer($name))";
-		} elsif($type eq "string" && $errvar ne "") {
-			$text .= "\tvar _p$n $strconvtype\n";
-			$text .= "\t_p$n, $errvar = $strconvfunc($name)\n";
-			$text .= "\tif $errvar != nil {\n\t\treturn\n\t}\n";
-			push @args, "uintptr(unsafe.Pointer(_p$n))";
-			$n++;
-		} elsif($type eq "string") {
-			print STDERR "$ARGV:$.: $func uses string arguments, but has no error return\n";
-			$text .= "\tvar _p$n $strconvtype\n";
-			$text .= "\t_p$n, _ = $strconvfunc($name)\n";
-			push @args, "uintptr(unsafe.Pointer(_p$n))";
-			$n++;
-		} elsif($type =~ /^\[\](.*)/) {
-			# Convert slice into pointer, length.
-			# Have to be careful not to take address of &a[0] if len == 0:
-			# pass nil in that case.
-			$text .= "\tvar _p$n *$1\n";
-			$text .= "\tif len($name) > 0 {\n\t\t_p$n = \&$name\[0]\n\t}\n";
-			push @args, "uintptr(unsafe.Pointer(_p$n))", "uintptr(len($name))";
-			$n++;
-		} elsif($type eq "int64" && $_32bit ne "") {
-			if($_32bit eq "big-endian") {
-				push @args, "uintptr($name >> 32)", "uintptr($name)";
-			} else {
-				push @args, "uintptr($name)", "uintptr($name >> 32)";
-			}
-		} elsif($type eq "bool") {
- 			$text .= "\tvar _p$n uint32\n";
-			$text .= "\tif $name {\n\t\t_p$n = 1\n\t} else {\n\t\t_p$n = 0\n\t}\n";
-			push @args, "uintptr(_p$n)";
-			$n++;
-		} else {
-			push @args, "uintptr($name)";
-		}
-	}
-	my $nargs = @args;
-
-	# Determine which form to use; pad args with zeros.
-	my $asm = "sysvicall6";
-	if ($nonblock) {
-		$asm = "rawSysvicall6";
-	}
-	if(@args <= 6) {
-		while(@args < 6) {
-			push @args, "0";
-		}
-	} else {
-		print STDERR "$ARGV:$.: too many arguments to system call\n";
-	}
-
-	# Actual call.
-	my $args = join(', ', @args);
-	my $call = "$asm(uintptr(unsafe.Pointer(&$sysvarname)), $nargs, $args)";
-
-	# Assign return values.
-	my $body = "";
-	my $failexpr = "";
-	my @ret = ("_", "_", "_");
-	my @pout= ();
-	my $do_errno = 0;
-	for(my $i=0; $i<@out; $i++) {
-		my $p = $out[$i];
-		my ($name, $type) = parseparam($p);
-		my $reg = "";
-		if($name eq "err") {
-			$reg = "e1";
-			$ret[2] = $reg;
-			$do_errno = 1;
-		} else {
-			$reg = sprintf("r%d", $i);
-			$ret[$i] = $reg;
-		}
-		if($type eq "bool") {
-			$reg = "$reg != 0";
-		}
-		if($type eq "int64" && $_32bit ne "") {
-			# 64-bit number in r1:r0 or r0:r1.
-			if($i+2 > @out) {
-				print STDERR "$ARGV:$.: not enough registers for int64 return\n";
-			}
-			if($_32bit eq "big-endian") {
-				$reg = sprintf("int64(r%d)<<32 | int64(r%d)", $i, $i+1);
-			} else {
-				$reg = sprintf("int64(r%d)<<32 | int64(r%d)", $i+1, $i);
-			}
-			$ret[$i] = sprintf("r%d", $i);
-			$ret[$i+1] = sprintf("r%d", $i+1);
-		}
-		if($reg ne "e1") {
-			$body .= "\t$name = $type($reg)\n";
-		}
-	}
-	if ($ret[0] eq "_" && $ret[1] eq "_" && $ret[2] eq "_") {
-		$text .= "\t$call\n";
-	} else {
-		$text .= "\t$ret[0], $ret[1], $ret[2] := $call\n";
-	}
-	$text .= $body;
-
-	if ($do_errno) {
-		$text .= "\tif e1 != 0 {\n";
-		$text .= "\t\terr = e1\n";
-		$text .= "\t}\n";
-	}
-	$text .= "\treturn\n";
-	$text .= "}\n";
-}
-
-if($errors) {
-	exit 1;
-}
-
-print <<EOF;
-// $cmdline
-// Code generated by the command above; see README.md. DO NOT EDIT.
-
-// +build $tags
-
-package $package
-
-import (
-	"syscall"
-	"unsafe"
-)
-EOF
-
-print "import \"golang.org/x/sys/unix\"\n" if $package ne "unix";
-
-my $vardecls = "\t" . join(",\n\t", @vars);
-$vardecls .= " syscallFunc";
-
-chomp($_=<<EOF);
-
-$dynimports
-$linknames
-var (
-$vardecls
-)
-
-$text
-EOF
-print $_;
-exit 0;

unix/zsyscall_solaris_amd64.go

@@ -1,4 +1,4 @@
-// mksyscall_solaris.pl -tags solaris,amd64 syscall_solaris.go syscall_solaris_amd64.go
+// go run mksyscall_solaris.go -tags solaris,amd64 syscall_solaris.go syscall_solaris_amd64.go
 // Code generated by the command above; see README.md. DO NOT EDIT.
 
 // +build solaris,amd64