[release-branch.r58] cgo: handle new Apple LLVM-based gcc from Xcode 4.2

««« CL 4607045 / 142f0bc0d6e7
cgo: handle new Apple LLVM-based gcc from Xcode 4.2

That gcc does not include enumerator names and values
in its DWARF debug output.  Create a data block from which
we can read the values instead.

Fixes #1881.

R=iant
CC=golang-dev
https://golang.org/cl/4607045
»»»

R=adg
CC=golang-dev
https://golang.org/cl/4708042

doc/devel/release.html

@@ -14,6 +14,86 @@ hg pull
 hg update release.r<i>NN</i>
 </pre>
 
+<h2 id="r58">r58 (released 2011/06/29)</h2>
+
+<p>
+The r58 release corresponds to 
+<code><a href="weekly.html#2011-06-09">weekly.2011-06-09</a></code>
+with additional bug fixes.
+This section highlights the most significant changes in this release.
+For a more detailed summary, see the
+<a href="weekly.html#2011-06-09">weekly release notes</a>.
+For complete information, see the
+<a href="http://code.google.com/p/go/source/list?r=release-branch.r58">Mercurial change list</a>.
+</p>
+
+<h3 id="r58.lang">Language</h3>
+
+<p>
+This release fixes a <a href="http://code.google.com/p/go/source/detail?r=b720749486e1">use of uninitialized memory in programs that misuse <code>goto</code></a>.
+</p>
+
+<h3 id="r58.pkg">Packages</h3>
+
+<p>
+As usual, <a href="/cmd/gofix/">gofix</a> will handle the bulk of the rewrites
+necessary for these changes to package APIs.
+</p>
+
+<p>
+<a href="/pkg/http/">Package http</a> drops the <code>finalURL</code> return
+value from the <a href="/pkg/http/#Client.Get">Client.Get</a> method. The value
+is now available via the new <code>Request</code> field on <a
+href="/pkg/http/#Response">http.Response</a>.
+Most instances of the type map[string][]string in have been
+replaced with the new <a href="/pkg/http/#Values">Values</a> type.
+</p>
+
+<p>
+<a href="/pkg/exec/">Package exec</a> has been redesigned with a more
+convenient and succinct API.
+</p>
+
+<p>
+<a href="/pkg/strconv/">Package strconv</a>'s <a href="/pkg/strconv/#Quote">Quote</a>
+function now escapes only those Unicode code points not classified as printable
+by <a href="/pkg/unicode/#IsPrint">unicode.IsPrint</a>.
+Previously Quote would escape all non-ASCII characters.
+This also affects the <a href="/pkg/fmt/">fmt</a> package's <code>"%q"</code>
+formatting directive. The previous quoting behavior is still available via
+strconv's new <a href="/pkg/strconv/#QuoteToASCII">QuoteToASCII</a> function.   
+</p>
+
+<p>
+<a href="/pkg/os/signal/">Package os/signal</a>'s
+<a href="/pkg/os/#Signal">Signal</a> and 
+<a href="/pkg/os/#UnixSignal">UnixSignal</a> types have been moved to the
+<a href="/pkg/os/">os</a> package.
+</p>
+
+<p>
+<a href="/pkg/image/draw/">Package image/draw</a> is the new name for
+<code>exp/draw</code>. The GUI-related code from <code>exp/draw</code> is now
+located in the <a href="/pkg/exp/gui/">exp/gui</a> package.
+</p>
+
+<h3 id="r58.cmd">Tools</h3>
+
+<p>
+<a href="/cmd/goinstall/">Goinstall</a> now observes the GOPATH environment
+variable to build and install your own code and external libraries outside of
+the Go tree (and avoid writing Makefiles).
+</p>
+
+
+<h3 id="r58.minor">Minor revisions</h3>
+
+<p>r58.1 adds 
+<a href="http://code.google.com/p/go/source/detail?r=293c25943586">build</a> and
+<a href="http://code.google.com/p/go/source/detail?r=bf17e96b6582">runtime</a>
+changes to make Go run on OS X 10.7 Lion.
+</p>
+
 <h2 id="r57">r57 (released 2011/05/03)</h2>
 
 <p>
@@ -162,6 +242,7 @@ For other uses, see the <a href="/pkg/runtime/pprof/">runtime/pprof</a> document
 <h3 id="r57.minor">Minor revisions</h3>
 
 <p>r57.1 fixes a <a href="http://code.google.com/p/go/source/detail?r=ff2bc62726e7145eb2ecc1e0f076998e4a8f86f0">nil pointer dereference in http.FormFile</a>.</p>
+<p>r57.2 fixes a <a href="http://code.google.com/p/go/source/detail?r=063b0ff67d8277df03c956208abc068076818dae">use of uninitialized memory in programs that misuse <code>goto</code></a>.</p>
 
 <h2 id="r56">r56 (released 2011/03/16)</h2>
 

doc/devel/weekly.html

@@ -14,7 +14,7 @@ hg pull
 hg update weekly.<i>YYYY-MM-DD</i>
 </pre>
 
-<h2 id="2011-06-09">2011-06-09</h2>
+<h2 id="2011-06-09">2011-06-09 (<a href="release.html#r58">base for r58</a>)</h2>
 
 <pre>
 This release includes changes to the strconv, http, and exp/draw packages.

doc/docs.html

@@ -23,6 +23,17 @@ concepts: syntax, types, allocation, constants, I/O, sorting, printing,
 goroutines, and channels.
 </p>
 
+<h3 id="course_notes">Course Notes</h3>
+<p>
+Slides from a 3-day course about the Go programming language.
+A more thorough introduction than the tutorial.
+</p>
+<ul>
+<li><a href="GoCourseDay1.pdf">Day 1: Basics</a> <small>[270KB PDF]</small>
+<li><a href="GoCourseDay2.pdf">Day 2: Types, Methods, Interfaces</a> <small>[270KB PDF]</small>
+<li><a href="GoCourseDay3.pdf">Day 3: Concurrency and Communication</a> <small>[180KB PDF]</small>
+</ul>
+
 <h3 id="effective_go"><a href="effective_go.html">Effective Go</a></h3>
 <p>
 A document that gives tips for writing clear, idiomatic Go code.
@@ -209,7 +220,7 @@ from Hoare’s 1978 paper to Go provides insight into how and why Go works as it
 does.
 </i></p>
 
-<h3 id="emerging_go"><a href="talks/gofrontend-gcc-summit-2010.pdf">The Go frontend for GCC</a></h3>
+<h3 id="go_frontend_gcc"><a href="talks/gofrontend-gcc-summit-2010.pdf">The Go frontend for GCC</a></h3>
 <p>
 A description of the Go language frontend for gcc.
 Ian Lance Taylor's paper delivered at the GCC Summit 2010.

doc/go_faq.html

@@ -183,16 +183,6 @@ easier to understand what happens when things combine.
 
 <h2 id="Usage">Usage</h2>
 
-<h3 id="Who_should_use_the_language">
-Who should use the language?</h3>
-
-<p>
-Go is an experiment. We hope adventurous users will give it a try and see
-if they enjoy it. Not every programmer
-will, but we hope enough will find satisfaction in the approach it
-offers to justify further development.
-</p>
-
 <h3 id="Is_Google_using_go_internally"> Is Google using Go internally?</h3>
 
 <p>
@@ -598,6 +588,24 @@ the interface idea. Sometimes, though, they're necessary to resolve ambiguities
 among similar interfaces.
 </p>
 
+<h3 id="convert_slice_of_interface">
+Can I convert a []T to an []interface{}?</h3>
+
+<p>
+Not directly because they do not have the same representation in memory.
+It is necessary to copy the elements individually to the destination
+slice. This example converts a slice of <code>int</code> to a slice of
+<code>interface{}</code>:
+</p>
+
+<pre>
+t := []int{1, 2, 3, 4}
+s := make([]interface{}, len(t))
+for i, v := range t {
+	s[i] = v
+}
+</pre>
+
 <h2 id="values">Values</h2>
 
 <h3 id="conversions">

doc/go_tutorial.html

@@ -10,8 +10,7 @@ After you've read this tutorial, you should look at
 which digs deeper into how the language is used and
 talks about the style and idioms of programming in Go.
 Also, slides from a 3-day course about Go are available.
-Although they're badly out of date, they provide some
-background and a lot of examples:
+They provide some background and a lot of examples:
 <a href='/doc/GoCourseDay1.pdf'>Day 1</a>,
 <a href='/doc/GoCourseDay2.pdf'>Day 2</a>,
 <a href='/doc/GoCourseDay3.pdf'>Day 3</a>.

doc/go_tutorial.txt

@@ -11,8 +11,7 @@ After you've read this tutorial, you should look at
 which digs deeper into how the language is used and
 talks about the style and idioms of programming in Go.
 Also, slides from a 3-day course about Go are available.
-Although they're badly out of date, they provide some
-background and a lot of examples:
+They provide some background and a lot of examples:
 <a href='/doc/GoCourseDay1.pdf'>Day 1</a>,
 <a href='/doc/GoCourseDay2.pdf'>Day 2</a>,
 <a href='/doc/GoCourseDay3.pdf'>Day 3</a>.

src/cmd/5g/gsubr.c

@@ -124,6 +124,64 @@ newplist(void)
 	return pl;
 }
 
+void
+clearstk(void)
+{
+	Plist *pl;
+	Prog *p, *p1, *p2, *p3;
+	Node dst, end, zero, con;
+
+	if(plast->firstpc->to.offset <= 0)
+		return;
+
+	// reestablish context for inserting code
+	// at beginning of function.
+	pl = plast;
+	p1 = pl->firstpc;
+	p2 = p1->link;
+	pc = mal(sizeof(*pc));
+	clearp(pc);
+	p1->link = pc;
+	
+	// zero stack frame
+
+	// MOVW $4(SP), R1
+	nodreg(&dst, types[tptr], 1);
+	p = gins(AMOVW, N, &dst);
+	p->from.type = D_CONST;
+	p->from.reg = REGSP;
+	p->from.offset = 4;
+
+	// MOVW $n(R1), R2
+	nodreg(&end, types[tptr], 2);
+	p = gins(AMOVW, N, &end);
+	p->from.type = D_CONST;
+	p->from.reg = 1;
+	p->from.offset = p1->to.offset;
+	
+	// MOVW $0, R3
+	nodreg(&zero, types[TUINT32], 3);
+	nodconst(&con, types[TUINT32], 0);
+	gmove(&con, &zero);
+
+	// L:
+	//	MOVW.P R3, 0(R1) +4
+	//	CMP R1, R2
+	//	BNE L
+	p = gins(AMOVW, &zero, &dst);
+	p->to.type = D_OREG;
+	p->to.offset = 4;
+	p->scond |= C_PBIT;
+	p3 = p;
+	p = gins(ACMP, &dst, N);
+	raddr(&end, p);
+	patch(gbranch(ABNE, T), p3);
+
+	// continue with original code.
+	gins(ANOP, N, N)->link = p2;
+	pc = P;
+}	
+
 void
 gused(Node *n)
 {

src/cmd/6g/gsubr.c

@@ -120,6 +120,44 @@ newplist(void)
 	return pl;
 }
 
+void
+clearstk(void)
+{
+	Plist *pl;
+	Prog *p1, *p2;
+	Node sp, di, cx, con, ax;
+
+	if((uint32)plast->firstpc->to.offset <= 0)
+		return;
+
+	// reestablish context for inserting code
+	// at beginning of function.
+	pl = plast;
+	p1 = pl->firstpc;
+	p2 = p1->link;
+	pc = mal(sizeof(*pc));
+	clearp(pc);
+	p1->link = pc;
+	
+	// zero stack frame
+	nodreg(&sp, types[tptr], D_SP);
+	nodreg(&di, types[tptr], D_DI);
+	nodreg(&cx, types[TUINT64], D_CX);
+	nodconst(&con, types[TUINT64], (uint32)p1->to.offset / widthptr);
+	gins(ACLD, N, N);
+	gins(AMOVQ, &sp, &di);
+	gins(AMOVQ, &con, &cx);
+	nodconst(&con, types[TUINT64], 0);
+	nodreg(&ax, types[TUINT64], D_AX);
+	gins(AMOVQ, &con, &ax);
+	gins(AREP, N, N);
+	gins(ASTOSQ, N, N);
+
+	// continue with original code.
+	gins(ANOP, N, N)->link = p2;
+	pc = P;
+}	
+
 void
 gused(Node *n)
 {

src/cmd/8g/gsubr.c

@@ -122,6 +122,44 @@ newplist(void)
 	return pl;
 }
 
+void
+clearstk(void)
+{
+	Plist *pl;
+	Prog *p1, *p2;
+	Node sp, di, cx, con, ax;
+
+	if(plast->firstpc->to.offset <= 0)
+		return;
+
+	// reestablish context for inserting code
+	// at beginning of function.
+	pl = plast;
+	p1 = pl->firstpc;
+	p2 = p1->link;
+	pc = mal(sizeof(*pc));
+	clearp(pc);
+	p1->link = pc;
+	
+	// zero stack frame
+	nodreg(&sp, types[tptr], D_SP);
+	nodreg(&di, types[tptr], D_DI);
+	nodreg(&cx, types[TUINT32], D_CX);
+	nodconst(&con, types[TUINT32], p1->to.offset / widthptr);
+	gins(ACLD, N, N);
+	gins(AMOVL, &sp, &di);
+	gins(AMOVL, &con, &cx);
+	nodconst(&con, types[TUINT32], 0);
+	nodreg(&ax, types[TUINT32], D_AX);
+	gins(AMOVL, &con, &ax);
+	gins(AREP, N, N);
+	gins(ASTOSL, N, N);
+
+	// continue with original code.
+	gins(ANOP, N, N)->link = p2;
+	pc = P;
+}	
+
 void
 gused(Node *n)
 {

src/cmd/cgo/gcc.go

@@ -13,6 +13,7 @@ import (
 	"debug/elf"
 	"debug/macho"
 	"debug/pe"
+	"encoding/binary"
 	"flag"
 	"fmt"
 	"go/ast"
@@ -477,7 +478,27 @@ func (p *Package) loadDWARF(f *File, names []*Name) {
 			fmt.Fprintf(&b, "enum { __cgo_enum__%d = %s };\n", i, n.C)
 		}
 	}
-	d := p.gccDebug(b.Bytes())
+
+	// Apple's LLVM-based gcc does not include the enumeration
+	// names and values in its DWARF debug output.  In case we're
+	// using such a gcc, create a data block initialized with the values.
+	// We can read them out of the object file.
+	fmt.Fprintf(&b, "long long __cgodebug_data[] = {\n")
+	for _, n := range names {
+		if n.Kind == "const" {
+			fmt.Fprintf(&b, "\t%s,\n", n.C)
+		} else {
+			fmt.Fprintf(&b, "\t0,\n")
+		}
+	}
+	fmt.Fprintf(&b, "\t0\n")
+	fmt.Fprintf(&b, "};\n")
+
+	d, bo, debugData := p.gccDebug(b.Bytes())
+	enumVal := make([]int64, len(debugData)/8)
+	for i := range enumVal {
+		enumVal[i] = int64(bo.Uint64(debugData[i*8:]))
+	}
 
 	// Scan DWARF info for top-level TagVariable entries with AttrName __cgo__i.
 	types := make([]dwarf.Type, len(names))
@@ -569,9 +590,12 @@ func (p *Package) loadDWARF(f *File, names []*Name) {
 				// Remove injected enum to ensure the value will deep-compare
 				// equally in future loads of the same constant.
 				n.Type.EnumValues[k] = 0, false
+			} else if n.Kind == "const" && i < len(enumVal) {
+				n.Const = strconv.Itoa64(enumVal[i])
 			}
 		}
 	}
+
 }
 
 // rewriteRef rewrites all the C.xxx references in f.AST to refer to the
@@ -593,6 +617,9 @@ func (p *Package) rewriteRef(f *File) {
 	// are trying to do a ,err call.  Also check that
 	// functions are only used in calls.
 	for _, r := range f.Ref {
+		if r.Name.Kind == "const" && r.Name.Const == "" {
+			error(r.Pos(), "unable to find value of constant C.%s", r.Name.Go)
+		}
 		var expr ast.Expr = ast.NewIdent(r.Name.Mangle) // default
 		switch r.Context {
 		case "call", "call2":
@@ -692,29 +719,57 @@ func (p *Package) gccCmd() []string {
 }
 
 // gccDebug runs gcc -gdwarf-2 over the C program stdin and
-// returns the corresponding DWARF data and any messages
-// printed to standard error.
-func (p *Package) gccDebug(stdin []byte) *dwarf.Data {
+// returns the corresponding DWARF data and, if present, debug data block.
+func (p *Package) gccDebug(stdin []byte) (*dwarf.Data, binary.ByteOrder, []byte) {
 	runGcc(stdin, p.gccCmd())
 
-	// Try to parse f as ELF and Mach-O and hope one works.
-	var f interface {
-		DWARF() (*dwarf.Data, os.Error)
-	}
-	var err os.Error
-	if f, err = elf.Open(gccTmp); err != nil {
-		if f, err = macho.Open(gccTmp); err != nil {
-			if f, err = pe.Open(gccTmp); err != nil {
-				fatalf("cannot parse gcc output %s as ELF or Mach-O or PE object", gccTmp)
+	if f, err := macho.Open(gccTmp); err == nil {
+		d, err := f.DWARF()
+		if err != nil {
+			fatalf("cannot load DWARF output from %s: %v", gccTmp, err)
+		}
+		var data []byte
+		if f.Symtab != nil {
+			for i := range f.Symtab.Syms {
+				s := &f.Symtab.Syms[i]
+				// Mach-O still uses a leading _ to denote non-assembly symbols.
+				if s.Name == "_"+"__cgodebug_data" {
+					// Found it.  Now find data section.
+					if i := int(s.Sect) - 1; 0 <= i && i < len(f.Sections) {
+						sect := f.Sections[i]
+						if sect.Addr <= s.Value && s.Value < sect.Addr+sect.Size {
+							if sdat, err := sect.Data(); err == nil {
+								data = sdat[s.Value-sect.Addr:]
+							}
+						}
+					}
+				}
 			}
 		}
+		return d, f.ByteOrder, data
 	}
 
-	d, err := f.DWARF()
-	if err != nil {
-		fatalf("cannot load DWARF debug information from %s: %s", gccTmp, err)
+	// Can skip debug data block in ELF and PE for now.
+	// The DWARF information is complete.
+
+	if f, err := elf.Open(gccTmp); err == nil {
+		d, err := f.DWARF()
+		if err != nil {
+			fatalf("cannot load DWARF output from %s: %v", gccTmp, err)
+		}
+		return d, f.ByteOrder, nil
 	}
-	return d
+
+	if f, err := pe.Open(gccTmp); err == nil {
+		d, err := f.DWARF()
+		if err != nil {
+			fatalf("cannot load DWARF output from %s: %v", gccTmp, err)
+		}
+		return d, binary.LittleEndian, nil
+	}
+
+	fatalf("cannot parse gcc output %s as ELF, Mach-O, PE object", gccTmp)
+	panic("not reached")
 }
 
 // gccDefines runs gcc -E -dM -xc - over the C program stdin

src/cmd/gc/dcl.c

@@ -1241,9 +1241,14 @@ funccompile(Node *n, int isclosure)
 	stksize = 0;
 	dclcontext = PAUTO;
 	funcdepth = n->funcdepth + 1;
+	hasgoto = 0;
 	compile(n);
+	if(hasgoto)
+		clearstk();
 	curfn = nil;
 	funcdepth = 0;
 	dclcontext = PEXTERN;
 }
 
+
+

src/cmd/gc/gen.c

@@ -209,6 +209,7 @@ gen(Node *n)
 		break;
 
 	case OGOTO:
+		hasgoto = 1;
 		newlab(OGOTO, n, N);
 		gjmp(P);
 		break;

src/cmd/gc/go.h

@@ -1255,3 +1255,6 @@ void	zhist(Biobuf *b, int line, vlong offset);
 void	zname(Biobuf *b, Sym *s, int t);
 void	data(void);
 void	text(void);
+
+EXTERN	int	hasgoto;
+void	clearstk(void);

src/cmd/gofix/Makefile

@@ -14,6 +14,7 @@ GOFILES=\
 	httpserver.go\
 	procattr.go\
 	reflect.go\
+	signal.go\
 	typecheck.go\
 
 include ../../Make.cmd

src/cmd/gofix/fix.go

@@ -10,6 +10,7 @@ import (
 	"go/token"
 	"os"
 	"strconv"
+	"strings"
 )
 
 type fix struct {
@@ -258,13 +259,28 @@ func walkBeforeAfter(x interface{}, before, after func(interface{})) {
 
 // imports returns true if f imports path.
 func imports(f *ast.File, path string) bool {
+	return importSpec(f, path) != nil
+}
+
+// importSpec returns the import spec if f imports path,
+// or nil otherwise.
+func importSpec(f *ast.File, path string) *ast.ImportSpec {
 	for _, s := range f.Imports {
-		t, err := strconv.Unquote(s.Path.Value)
-		if err == nil && t == path {
-			return true
+		if importPath(s) == path {
+			return s
 		}
 	}
-	return false
+	return nil
+}
+
+// importPath returns the unquoted import path of s,
+// or "" if the path is not properly quoted.
+func importPath(s *ast.ImportSpec) string {
+	t, err := strconv.Unquote(s.Path.Value)
+	if err == nil {
+		return t
+	}
+	return ""
 }
 
 // isPkgDot returns true if t is the expression "pkg.name"
@@ -420,3 +436,138 @@ func newPkgDot(pos token.Pos, pkg, name string) ast.Expr {
 		},
 	}
 }
+
+// addImport adds the import path to the file f, if absent.
+func addImport(f *ast.File, path string) {
+	if imports(f, path) {
+		return
+	}
+
+	newImport := &ast.ImportSpec{
+		Path: &ast.BasicLit{
+			Kind:  token.STRING,
+			Value: strconv.Quote(path),
+		},
+	}
+
+	var impdecl *ast.GenDecl
+
+	// Find an import decl to add to.
+	for _, decl := range f.Decls {
+		gen, ok := decl.(*ast.GenDecl)
+
+		if ok && gen.Tok == token.IMPORT {
+			impdecl = gen
+			break
+		}
+	}
+
+	// No import decl found.  Add one.
+	if impdecl == nil {
+		impdecl = &ast.GenDecl{
+			Tok: token.IMPORT,
+		}
+		f.Decls = append(f.Decls, nil)
+		copy(f.Decls[1:], f.Decls)
+		f.Decls[0] = impdecl
+	}
+
+	// Ensure the import decl has parentheses, if needed.
+	if len(impdecl.Specs) > 0 && !impdecl.Lparen.IsValid() {
+		impdecl.Lparen = impdecl.Pos()
+	}
+
+	// Assume the import paths are alphabetically ordered.
+	// If they are not, the result is ugly, but legal.
+	insertAt := len(impdecl.Specs) // default to end of specs
+	for i, spec := range impdecl.Specs {
+		impspec := spec.(*ast.ImportSpec)
+		if importPath(impspec) > path {
+			insertAt = i
+			break
+		}
+	}
+
+	impdecl.Specs = append(impdecl.Specs, nil)
+	copy(impdecl.Specs[insertAt+1:], impdecl.Specs[insertAt:])
+	impdecl.Specs[insertAt] = newImport
+
+	f.Imports = append(f.Imports, newImport)
+}
+
+// deleteImport deletes the import path from the file f, if present.
+func deleteImport(f *ast.File, path string) {
+	oldImport := importSpec(f, path)
+
+	// Find the import node that imports path, if any.
+	for i, decl := range f.Decls {
+		gen, ok := decl.(*ast.GenDecl)
+		if !ok || gen.Tok != token.IMPORT {
+			continue
+		}
+		for j, spec := range gen.Specs {
+			impspec := spec.(*ast.ImportSpec)
+
+			if oldImport != impspec {
+				continue
+			}
+
+			// We found an import spec that imports path.
+			// Delete it.
+			copy(gen.Specs[j:], gen.Specs[j+1:])
+			gen.Specs = gen.Specs[:len(gen.Specs)-1]
+
+			// If this was the last import spec in this decl,
+			// delete the decl, too.
+			if len(gen.Specs) == 0 {
+				copy(f.Decls[i:], f.Decls[i+1:])
+				f.Decls = f.Decls[:len(f.Decls)-1]
+			} else if len(gen.Specs) == 1 {
+				gen.Lparen = token.NoPos // drop parens
+			}
+
+			break
+		}
+	}
+
+	// Delete it from f.Imports.
+	for i, imp := range f.Imports {
+		if imp == oldImport {
+			copy(f.Imports[i:], f.Imports[i+1:])
+			f.Imports = f.Imports[:len(f.Imports)-1]
+			break
+		}
+	}
+}
+
+func usesImport(f *ast.File, path string) (used bool) {
+	spec := importSpec(f, path)
+	if spec == nil {
+		return
+	}
+
+	name := spec.Name.String()
+	switch name {
+	case "<nil>":
+		// If the package name is not explicitly specified,
+		// make an educated guess. This is not guaranteed to be correct.
+		lastSlash := strings.LastIndex(path, "/")
+		if lastSlash == -1 {
+			name = path
+		} else {
+			name = path[lastSlash+1:]
+		}
+	case "_", ".":
+		// Not sure if this import is used - err on the side of caution.
+		return true
+	}
+
+	walk(f, func(n interface{}) {
+		sel, ok := n.(*ast.SelectorExpr)
+		if ok && isTopName(sel.X, name) {
+			used = true
+		}
+	})
+
+	return
+}

src/cmd/gofix/signal.go

@@ -0,0 +1,49 @@
+// Copyright 2011 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 main
+
+import (
+	"go/ast"
+	"strings"
+)
+
+func init() {
+	register(fix{
+		"signal",
+		signal,
+		`Adapt code to types moved from os/signal to signal.
+
+http://codereview.appspot.com/4437091
+`,
+	})
+}
+
+func signal(f *ast.File) (fixed bool) {
+	if !imports(f, "os/signal") {
+		return
+	}
+
+	walk(f, func(n interface{}) {
+		s, ok := n.(*ast.SelectorExpr)
+
+		if !ok || !isTopName(s.X, "signal") {
+			return
+		}
+
+		sel := s.Sel.String()
+		if sel == "Signal" || sel == "UnixSignal" || strings.HasPrefix(sel, "SIG") {
+			s.X = &ast.Ident{Name: "os"}
+			fixed = true
+		}
+	})
+
+	if fixed {
+		addImport(f, "os")
+		if !usesImport(f, "os/signal") {
+			deleteImport(f, "os/signal")
+		}
+	}
+	return
+}

src/cmd/gofix/signal_test.go

@@ -0,0 +1,96 @@
+// Copyright 2011 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 main
+
+func init() {
+	addTestCases(signalTests)
+}
+
+var signalTests = []testCase{
+	{
+		Name: "signal.0",
+		In: `package main
+
+import (
+	_ "a"
+	"os/signal"
+	_ "z"
+)
+
+type T1 signal.UnixSignal
+type T2 signal.Signal
+
+func f() {
+	_ = signal.SIGHUP
+	_ = signal.Incoming
+}
+`,
+		Out: `package main
+
+import (
+	_ "a"
+	"os"
+	"os/signal"
+	_ "z"
+)
+
+type T1 os.UnixSignal
+type T2 os.Signal
+
+func f() {
+	_ = os.SIGHUP
+	_ = signal.Incoming
+}
+`,
+	},
+	{
+		Name: "signal.1",
+		In: `package main
+
+import (
+	"os"
+	"os/signal"
+)
+
+func f() {
+	var _ os.Error
+	_ = signal.SIGHUP
+}
+`,
+		Out: `package main
+
+import "os"
+
+
+func f() {
+	var _ os.Error
+	_ = os.SIGHUP
+}
+`,
+	},
+	{
+		Name: "signal.2",
+		In: `package main
+
+import "os"
+import "os/signal"
+
+func f() {
+	var _ os.Error
+	_ = signal.SIGHUP
+}
+`,
+		Out: `package main
+
+import "os"
+
+
+func f() {
+	var _ os.Error
+	_ = os.SIGHUP
+}
+`,
+	},
+}

src/cmd/ld/dwarf.c

@@ -1804,7 +1804,7 @@ mkvarname(char* name, int da)
 
 // flush previous compilation unit.
 static void
-flushunit(DWDie *dwinfo, vlong pc, vlong unitstart)
+flushunit(DWDie *dwinfo, vlong pc, vlong unitstart, int32 header_length)
 {
 	vlong here;
 
@@ -1820,7 +1820,9 @@ flushunit(DWDie *dwinfo, vlong pc, vlong unitstart)
 
 		here = cpos();
 		seek(cout, unitstart, 0);
-		LPUT(here - unitstart - sizeof(int32));
+		LPUT(here - unitstart - sizeof(int32));	 // unit_length
+		WPUT(3);  // dwarf version
+		LPUT(header_length); // header lenght starting here
 		cflush();
 		seek(cout, here, 0);
 	}
@@ -1832,7 +1834,7 @@ writelines(void)
 	Prog *q;
 	Sym *s;
 	Auto *a;
-	vlong unitstart, offs;
+	vlong unitstart, headerend, offs;
 	vlong pc, epc, lc, llc, lline;
 	int currfile;
 	int i, lang, da, dt;
@@ -1842,6 +1844,7 @@ writelines(void)
 	char *n, *nn;
 
 	unitstart = -1;
+	headerend = -1;
 	pc = 0;
 	epc = 0;
 	lc = 1;
@@ -1859,7 +1862,7 @@ writelines(void)
 		// we're entering a new compilation unit
 
 		if (inithist(s->autom)) {
-			flushunit(dwinfo, epc, unitstart);
+			flushunit(dwinfo, epc, unitstart, headerend - unitstart - 10);
 			unitstart = cpos();
 
 			if(debug['v'] > 1) {
@@ -1880,10 +1883,10 @@ writelines(void)
 
 			// Write .debug_line Line Number Program Header (sec 6.2.4)
 			// Fields marked with (*) must be changed for 64-bit dwarf
-			LPUT(0);   // unit_length (*), will be filled in later.
+			LPUT(0);   // unit_length (*), will be filled in by flushunit.
 			WPUT(3);   // dwarf version (appendix F)
-			LPUT(11);  // header_length (*), starting here.
-
+			LPUT(0);   // header_length (*), filled in by flushunit.
+			// cpos == unitstart + 4 + 2 + 4
 			cput(1);   // minimum_instruction_length
 			cput(1);   // default_is_stmt
 			cput(LINE_BASE);     // line_base
@@ -1894,17 +1897,15 @@ writelines(void)
 			cput(1);   // standard_opcode_lengths[3]
 			cput(1);   // standard_opcode_lengths[4]
 			cput(0);   // include_directories  (empty)
-			cput(0);   // file_names (empty) (emitted by DW_LNE's below)
-			// header_length ends here.
 
 			for (i=1; i < histfilesize; i++) {
-				cput(0);  // start extended opcode
-				uleb128put(1 + strlen(histfile[i]) + 4);
-				cput(DW_LNE_define_file);
 				strnput(histfile[i], strlen(histfile[i]) + 4);
 				// 4 zeros: the string termination + 3 fields.
 			}
 
+			cput(0);   // terminate file_names.
+			headerend = cpos();
+
 			pc = s->text->pc;
 			epc = pc;
 			currfile = 1;
@@ -2009,7 +2010,7 @@ writelines(void)
 		dwfunc->hash = nil;
 	}
 
-	flushunit(dwinfo, epc, unitstart);
+	flushunit(dwinfo, epc, unitstart, headerend - unitstart - 10);
 	linesize = cpos() - lineo;
 }
 

src/cmd/prof/gopprof

@@ -150,7 +150,8 @@ pprof [options] <profile>
    The /<service> can be $HEAP_PAGE, $PROFILE_PAGE, /pprof/pmuprofile,
                          $GROWTH_PAGE, $CONTENTION_PAGE, /pprof/wall,
                          or /pprof/filteredprofile.
-   For instance: "pprof http://myserver.com:80$HEAP_PAGE".
+   For instance:
+     pprof http://myserver.com:80$HEAP_PAGE
    If /<service> is omitted, the service defaults to $PROFILE_PAGE (cpu profiling).
 pprof --symbols <program>
    Maps addresses to symbol names.  In this mode, stdin should be a
@@ -532,7 +533,7 @@ sub Init() {
     ConfigureObjTools($main::prog)
   }
 
-  # Break the opt_list_prefix into the prefix_list array
+  # Break the opt_lib_prefix into the prefix_list array
   @prefix_list = split (',', $main::opt_lib_prefix);
 
   # Remove trailing / from the prefixes, in the list to prevent
@@ -626,7 +627,7 @@ sub Main() {
     if ($main::opt_disasm) {
       PrintDisassembly($libs, $flat, $cumulative, $main::opt_disasm, $total);
     } elsif ($main::opt_list) {
-      PrintListing($libs, $flat, $cumulative, $main::opt_list);
+      PrintListing($total, $libs, $flat, $cumulative, $main::opt_list, 0);
     } elsif ($main::opt_text) {
       # Make sure the output is empty when have nothing to report
       # (only matters when --heapcheck is given but we must be
@@ -814,7 +815,7 @@ sub InteractiveCommand {
     my $ignore;
     ($routine, $ignore) = ParseInteractiveArgs($3);
 
-    my $profile = ProcessProfile($orig_profile, $symbols, "", $ignore);
+    my $profile = ProcessProfile($total, $orig_profile, $symbols, "", $ignore);
     my $reduced = ReduceProfile($symbols, $profile);
 
     # Get derived profiles
@@ -841,21 +842,22 @@ sub InteractiveCommand {
 
     return 1;
   }
-  if (m/^\s*list\s*(.+)/) {
+  if (m/^\s*(web)?list\s*(.+)/) {
+    my $html = (defined($1) && ($1 eq "web"));
     $main::opt_list = 1;
 
     my $routine;
     my $ignore;
-    ($routine, $ignore) = ParseInteractiveArgs($1);
+    ($routine, $ignore) = ParseInteractiveArgs($2);
 
-    my $profile = ProcessProfile($orig_profile, $symbols, "", $ignore);
+    my $profile = ProcessProfile($total, $orig_profile, $symbols, "", $ignore);
     my $reduced = ReduceProfile($symbols, $profile);
 
     # Get derived profiles
     my $flat = FlatProfile($reduced);
     my $cumulative = CumulativeProfile($reduced);
 
-    PrintListing($libs, $flat, $cumulative, $routine);
+    PrintListing($total, $libs, $flat, $cumulative, $routine, $html);
     return 1;
   }
   if (m/^\s*disasm\s*(.+)/) {
@@ -866,7 +868,7 @@ sub InteractiveCommand {
     ($routine, $ignore) = ParseInteractiveArgs($1);
 
     # Process current profile to account for various settings
-    my $profile = ProcessProfile($orig_profile, $symbols, "", $ignore);
+    my $profile = ProcessProfile($total, $orig_profile, $symbols, "", $ignore);
     my $reduced = ReduceProfile($symbols, $profile);
 
     # Get derived profiles
@@ -890,7 +892,7 @@ sub InteractiveCommand {
     ($focus, $ignore) = ParseInteractiveArgs($2);
 
     # Process current profile to account for various settings
-    my $profile = ProcessProfile($orig_profile, $symbols, $focus, $ignore);
+    my $profile = ProcessProfile($total, $orig_profile, $symbols, $focus, $ignore);
     my $reduced = ReduceProfile($symbols, $profile);
 
     # Get derived profiles
@@ -916,6 +918,7 @@ sub InteractiveCommand {
 
 
 sub ProcessProfile {
+  my $total_count = shift;
   my $orig_profile = shift;
   my $symbols = shift;
   my $focus = shift;
@@ -923,7 +926,6 @@ sub ProcessProfile {
 
   # Process current profile to account for various settings
   my $profile = $orig_profile;
-  my $total_count = TotalProfile($profile);
   printf("Total: %s %s\n", Unparse($total_count), Units());
   if ($focus ne '') {
     $profile = FocusProfile($symbols, $profile, $focus);
@@ -970,6 +972,11 @@ Commands:
   list [routine_regexp] [-ignore1] [-ignore2]
       Show source listing of routines whose names match "routine_regexp"
 
+  weblist [routine_regexp] [-ignore1] [-ignore2]
+      Displays a source listing of routines whose names match "routine_regexp"
+      in a web browser.  You can click on source lines to view the
+      corresponding disassembly.
+
   top [--cum] [-ignore1] [-ignore2]
   top20 [--cum] [-ignore1] [-ignore2]
   top37 [--cum] [-ignore1] [-ignore2]
@@ -1144,7 +1151,7 @@ sub PrintText {
              $sym);
     }
     $lines++;
-    last if ($line_limit >= 0 && $lines > $line_limit);
+    last if ($line_limit >= 0 && $lines >= $line_limit);
   }
 }
 
@@ -1291,11 +1298,32 @@ sub ByName {
 
 # Print source-listing for all all routines that match $main::opt_list
 sub PrintListing {
+  my $total = shift;
   my $libs = shift;
   my $flat = shift;
   my $cumulative = shift;
   my $list_opts = shift;
+  my $html = shift;
+  
+  my $output = \*STDOUT;
+  my $fname = "";
+  
 
+  if ($html) {
+    # Arrange to write the output to a temporary file
+    $fname = TempName($main::next_tmpfile, "html");
+    $main::next_tmpfile++;
+    if (!open(TEMP, ">$fname")) {
+      print STDERR "$fname: $!\n";
+      return;
+    }
+    $output = \*TEMP;
+    print $output HtmlListingHeader();
+    printf $output ("<div class=\"legend\">%s<br>Total: %s %s</div>\n",
+                    $main::prog, Unparse($total), Units());
+  }
+ 
+  my $listed = 0;
   foreach my $lib (@{$libs}) {
     my $symbol_table = GetProcedureBoundaries($lib->[0], $list_opts);
     my $offset = AddressSub($lib->[1], $lib->[3]);
@@ -1307,15 +1335,98 @@ sub PrintListing {
       my $addr = AddressAdd($start_addr, $offset);
       for (my $i = 0; $i < $length; $i++) {
         if (defined($cumulative->{$addr})) {
-          PrintSource($lib->[0], $offset,
-                      $routine, $flat, $cumulative,
-                      $start_addr, $end_addr);
+          $listed += PrintSource(
+            $lib->[0], $offset,
+            $routine, $flat, $cumulative,
+            $start_addr, $end_addr,
+            $html,
+            $output);
           last;
         }
         $addr = AddressInc($addr);
       }
     }
   }
+
+  if ($html) {
+    if ($listed > 0) {
+      print $output HtmlListingFooter();
+      close($output);
+      RunWeb($fname);
+    } else {
+      close($output);
+      unlink($fname);
+    }
+  }
+}
+
+sub HtmlListingHeader {
+  return <<'EOF';
+<DOCTYPE html>
+<html>
+<head>
+<title>Pprof listing</title>
+<style type="text/css">
+body {
+  font-family: sans-serif;
+}
+h1 {
+  font-size: 1.5em;
+  margin-bottom: 4px;
+}
+.legend {
+  font-size: 1.25em;
+}
+.line {
+  color: #aaaaaa;
+}
+.livesrc {
+  color: #0000ff;
+  cursor: pointer;
+}
+.livesrc:hover {
+  background-color: #cccccc;
+}
+.asm {
+  color: #888888;
+  display: none;
+}
+</style>
+<script type="text/javascript">
+function pprof_toggle_asm(e) {
+  var target;
+  if (!e) e = window.event;
+  if (e.target) target = e.target;
+  else if (e.srcElement) target = e.srcElement;
+
+  if (target && target.className == "livesrc") {
+    var asm = target.nextSibling;
+    if (asm && asm.className == "asm") {
+      asm.style.display = (asm.style.display == "block" ? "none" : "block");
+      e.preventDefault();
+      return false;
+    }
+  }
+}
+</script>
+</head>
+<body>
+EOF
+}
+
+sub HtmlListingFooter {
+  return <<'EOF';
+</body>
+</html>
+EOF
+}
+
+sub HtmlEscape {
+  my $text = shift;
+  $text =~ s/&/&amp;/g;
+  $text =~ s/</&lt;/g;
+  $text =~ s/>/&gt;/g;
+  return $text;
 }
 
 # Returns the indentation of the line, if it has any non-whitespace
@@ -1338,6 +1449,8 @@ sub PrintSource {
   my $cumulative = shift;
   my $start_addr = shift;
   my $end_addr = shift;
+  my $html = shift;
+  my $output = shift;
 
   # Disassemble all instructions (just to get line numbers)
   my @instructions = Disassemble($prog, $offset, $start_addr, $end_addr);
@@ -1353,7 +1466,7 @@ sub PrintSource {
   }
   if (!defined($filename)) {
     print STDERR "no filename found in $routine\n";
-    return;
+    return 0;
   }
 
   # Hack 2: assume that the largest line number from $filename is the
@@ -1386,7 +1499,7 @@ sub PrintSource {
   {
     if (!open(FILE, "<$filename")) {
       print STDERR "$filename: $!\n";
-      return;
+      return 0;
     }
     my $l = 0;
     my $first_indentation = -1;
@@ -1414,12 +1527,21 @@ sub PrintSource {
   # Assign all samples to the range $firstline,$lastline,
   # Hack 4: If an instruction does not occur in the range, its samples
   # are moved to the next instruction that occurs in the range.
-  my $samples1 = {};
-  my $samples2 = {};
-  my $running1 = 0;     # Unassigned flat counts
-  my $running2 = 0;     # Unassigned cumulative counts
-  my $total1 = 0;       # Total flat counts
-  my $total2 = 0;       # Total cumulative counts
+  my $samples1 = {};        # Map from line number to flat count
+  my $samples2 = {};        # Map from line number to cumulative count
+  my $running1 = 0;         # Unassigned flat counts
+  my $running2 = 0;         # Unassigned cumulative counts
+  my $total1 = 0;           # Total flat counts
+  my $total2 = 0;           # Total cumulative counts
+  my %disasm = ();          # Map from line number to disassembly
+  my $running_disasm = "";  # Unassigned disassembly
+  my $skip_marker = "---\n";
+  if ($html) {
+    $skip_marker = "";
+    for (my $l = $firstline; $l <= $lastline; $l++) {
+      $disasm{$l} = "";
+    }
+  }
   foreach my $e (@instructions) {
     # Add up counts for all address that fall inside this instruction
     my $c1 = 0;
@@ -1428,6 +1550,15 @@ sub PrintSource {
       $c1 += GetEntry($flat, $a);
       $c2 += GetEntry($cumulative, $a);
     }
+
+    if ($html) {
+      $running_disasm .= sprintf("      %6s %6s \t\t%8s: %s\n",
+                                 HtmlPrintNumber($c1),
+                                 HtmlPrintNumber($c2),
+                                 $e->[0],
+                                 CleanDisassembly($e->[3]));
+    }
+
     $running1 += $c1;
     $running2 += $c2;
     $total1 += $c1;
@@ -1442,6 +1573,10 @@ sub PrintSource {
       AddEntry($samples2, $line, $running2);
       $running1 = 0;
       $running2 = 0;
+      if ($html) {
+        $disasm{$line} .= $running_disasm;
+        $running_disasm = '';
+      }
     }
   }
 
@@ -1449,16 +1584,28 @@ sub PrintSource {
   AddEntry($samples1, $lastline, $running1);
   AddEntry($samples2, $lastline, $running2);
 
-  printf("ROUTINE ====================== %s in %s\n" .
-         "%6s %6s Total %s (flat / cumulative)\n",
-         ShortFunctionName($routine),
-         $filename,
-         Units(),
-         Unparse($total1),
-         Unparse($total2));
+  if ($html) {
+    printf $output (
+      "<h1>%s</h1>%s\n<pre onClick=\"pprof_toggle_asm()\">\n" .
+      "Total:%6s %6s (flat / cumulative %s)\n",
+      HtmlEscape(ShortFunctionName($routine)),
+      HtmlEscape($filename),
+      Unparse($total1),
+      Unparse($total2),
+      Units());
+  } else {
+    printf $output (
+      "ROUTINE ====================== %s in %s\n" .
+      "%6s %6s Total %s (flat / cumulative)\n",
+      ShortFunctionName($routine),
+      $filename,
+      Unparse($total1),
+      Unparse($total2),
+      Units());
+  }
   if (!open(FILE, "<$filename")) {
     print STDERR "$filename: $!\n";
-    return;
+    return 0;
   }
   my $l = 0;
   while (<FILE>) {
@@ -1468,16 +1615,47 @@ sub PrintSource {
         (($l <= $oldlastline + 5) || ($l <= $lastline))) {
       chop;
       my $text = $_;
-      if ($l == $firstline) { printf("---\n"); }
-      printf("%6s %6s %4d: %s\n",
-             UnparseAlt(GetEntry($samples1, $l)),
-             UnparseAlt(GetEntry($samples2, $l)),
-             $l,
-             $text);
-      if ($l == $lastline)  { printf("---\n"); }
+      if ($l == $firstline) { print $output $skip_marker; }
+      my $n1 = GetEntry($samples1, $l);
+      my $n2 = GetEntry($samples2, $l);
+      if ($html) {
+        my $dis = $disasm{$l};
+        if (!defined($dis) || $n1 + $n2 == 0) {
+          # No samples/disassembly for this source line
+          printf $output (
+            "<span class=\"line\">%5d</span> " .
+            "<span class=\"deadsrc\">%6s %6s %s</span>\n",
+            $l,
+            HtmlPrintNumber($n1),
+            HtmlPrintNumber($n2),
+            HtmlEscape($text));
+        } else {
+          printf $output (
+            "<span class=\"line\">%5d</span> " .
+            "<span class=\"livesrc\">%6s %6s %s</span>" .
+            "<span class=\"asm\">%s</span>\n",
+            $l,
+            HtmlPrintNumber($n1),
+            HtmlPrintNumber($n2),
+            HtmlEscape($text),
+            HtmlEscape($dis));
+        }
+      } else {
+        printf $output(
+          "%6s %6s %4d: %s\n",
+          UnparseAlt($n1),
+          UnparseAlt($n2),
+          $l,
+          $text);
+      }
+      if ($l == $lastline)  { print $output $skip_marker; }
     };
   }
   close(FILE);
+  if ($html) {
+    print $output "</pre>\n";
+  }
+  return 1;
 }
 
 # Return the source line for the specified file/linenumber.
@@ -1625,16 +1803,11 @@ sub PrintDisassembledFunction {
       $address =~ s/^0x//;
       $address =~ s/^0*//;
 
-      # Trim symbols
-      my $d = $e->[3];
-      while ($d =~ s/\([^()%]*\)(\s*const)?//g) { } # Argument types, not (%rax)
-      while ($d =~ s/(\w+)<[^<>]*>/$1/g)  { }       # Remove template arguments
-
       printf("%6s %6s    %8s: %6s\n",
              UnparseAlt($flat_count[$x]),
              UnparseAlt($cum_count[$x]),
              $address,
-             $d);
+             CleanDisassembly($e->[3]));
     }
   }
 }
@@ -2254,6 +2427,16 @@ sub UnparseAlt {
   }
 }
 
+# Alternate pretty-printed form: 0 maps to ""
+sub HtmlPrintNumber {
+  my $num = shift;
+  if ($num == 0) {
+    return "";
+  } else {
+    return Unparse($num);
+  }
+}
+
 # Return output units
 sub Units {
   if ($main::profile_type eq 'heap' || $main::profile_type eq 'growth') {
@@ -2415,6 +2598,8 @@ sub RemoveUninterestingFrames {
                       'copyin',
                       'gostring',
                       'gostringsize',
+                      'growslice1',
+                      'appendslice1',
                       'hash_init',
                       'hash_subtable_new',
                       'hash_conv',
@@ -2422,6 +2607,8 @@ sub RemoveUninterestingFrames {
                       'hash_insert_internal',
                       'hash_insert',
                       'mapassign',
+                      'runtime.mapassign',
+                      'runtime.appendslice',
                       'runtime.mapassign1',
                       'makechan',
                       'makemap',
@@ -2433,11 +2620,13 @@ sub RemoveUninterestingFrames {
                       'unsafe.New',
                       'runtime.mallocgc',
                       'runtime.catstring',
+                      'runtime.growslice',
                       'runtime.ifaceT2E',
                       'runtime.ifaceT2I',
                       'runtime.makechan',
                       'runtime.makechan_c',
                       'runtime.makemap',
+                      'runtime.makemap_c',
                       'runtime.makeslice',
                       'runtime.mal',
                       'runtime.slicebytetostring',
@@ -4302,6 +4491,14 @@ sub ShortFunctionName {
   return $function;
 }
 
+# Trim overly long symbols found in disassembler output
+sub CleanDisassembly {
+  my $d = shift;
+  while ($d =~ s/\([^()%]*\)(\s*const)?//g) { } # Argument types, not (%rax)
+  while ($d =~ s/(\w+)<[^<>]*>/$1/g)  { }       # Remove template arguments
+  return $d;
+}
+
 ##### Miscellaneous #####
 
 # Find the right versions of the above object tools to use.  The

src/pkg/debug/dwarf/type.go

@@ -352,8 +352,8 @@ func (d *Data) Type(off Offset) (Type, os.Error) {
 			}
 		}
 		if ndim == 0 {
-			err = DecodeError{"info", e.Offset, "missing dimension for array"}
-			goto Error
+			// LLVM generates this for x[].
+			t.Count = -1
 		}
 
 	case TagBaseType:

src/pkg/debug/elf/file.go

@@ -546,6 +546,12 @@ func (f *File) DWARF() (*dwarf.Data, os.Error) {
 	return dwarf.New(abbrev, nil, nil, info, nil, nil, nil, str)
 }
 
+// Symbols returns the symbol table for f.
+func (f *File) Symbols() ([]Symbol, os.Error) {
+	sym, _, err := f.getSymbols(SHT_SYMTAB)
+	return sym, err
+}
+
 type ImportedSymbol struct {
 	Name    string
 	Version string

src/pkg/deps.bash

@@ -15,7 +15,13 @@ fi
 
 # Get list of directories from Makefile
 dirs=$(gomake --no-print-directory echo-dirs)
-dirpat=$(echo $dirs C | sed 's/ /|/g; s/.*/^(&)$/')
+dirpat=$(echo $dirs C | awk '{
+	for(i=1;i<=NF;i++){ 
+		x=$i
+		gsub("/", "\\/", x)
+		printf("/^(%s)$/\n", x)
+	}
+}')
 
 for dir in $dirs; do (
 	cd $dir || exit 1
@@ -30,7 +36,7 @@ for dir in $dirs; do (
 	deps=$(
 		sed -n '/^import.*"/p; /^import[ \t]*(/,/^)/p' $sources /dev/null |
 		cut -d '"' -f2 |
-		egrep "$dirpat" |
+		awk "$dirpat" |
 		grep -v "^$dir\$" |
 		sed 's/$/.install/' |
 		sed 's;^C\.install;runtime/cgo.install;' |

src/pkg/runtime/cgo/darwin_386.c

@@ -8,11 +8,13 @@
 static void* threadentry(void*);
 static pthread_key_t k1, k2;
 
+#define magic1 (0x23581321U)
+
 static void
 inittls(void)
 {
 	uint32 x, y;
-	pthread_key_t tofree[16], k;
+	pthread_key_t tofree[128], k;
 	int i, ntofree;
 	int havek1, havek2;
 
@@ -35,9 +37,8 @@ inittls(void)
 	 * 0x48+4*0x108 = 0x468 and 0x48+4*0x109 = 0x46c.
 	 *
 	 * The linker and runtime hard-code these constant offsets
-	 * from %gs where we expect to find m and g.  The code
-	 * below verifies that the constants are correct once it has
-	 * obtained the keys.  Known to ../cmd/8l/obj.c:/468
+	 * from %gs where we expect to find m and g.
+	 * Known to ../cmd/8l/obj.c:/468
 	 * and to ../pkg/runtime/darwin/386/sys.s:/468
 	 *
 	 * This is truly disgusting and a bit fragile, but taking care
@@ -48,55 +49,54 @@ inittls(void)
 	 * require an extra instruction and memory reference in
 	 * every stack growth prolog and would also require
 	 * rewriting the code that 8c generates for extern registers.
+	 *
+	 * Things get more disgusting on OS X 10.7 Lion.
+	 * The 0x48 base mentioned above is the offset of the tsd
+	 * array within the per-thread structure on Leopard and Snow Leopard.
+	 * On Lion, the base moved a little, so while the math above
+	 * still applies, the base is different.  Thus, we cannot
+	 * look for specific key values if we want to build binaries
+	 * that run on both systems.  Instead, forget about the
+	 * specific key values and just allocate and initialize per-thread
+	 * storage until we find a key that writes to the memory location
+	 * we want.  Then keep that key.
 	 */
 	havek1 = 0;
 	havek2 = 0;
 	ntofree = 0;
 	while(!havek1 || !havek2) {
 		if(pthread_key_create(&k, nil) < 0) {
-			fprintf(stderr, "libcgo: pthread_key_create failed\n");
+			fprintf(stderr, "runtime/cgo: pthread_key_create failed\n");
 			abort();
 		}
-		if(k == 0x108) {
+		pthread_setspecific(k, (void*)magic1);
+		asm volatile("movl %%gs:0x468, %0" : "=r"(x));
+		asm volatile("movl %%gs:0x46c, %0" : "=r"(y));
+		if(x == magic1) {
 			havek1 = 1;
 			k1 = k;
-			continue;
-		}
-		if(k == 0x109) {
+		} else if(y == magic1) {
 			havek2 = 1;
 			k2 = k;
-			continue;
+		} else {
+			if(ntofree >= nelem(tofree)) {
+				fprintf(stderr, "runtime/cgo: could not obtain pthread_keys\n");
+				fprintf(stderr, "\ttried");
+				for(i=0; i<ntofree; i++)
+					fprintf(stderr, " %#x", (unsigned)tofree[i]);
+				fprintf(stderr, "\n");
+				abort();
+			}
+			tofree[ntofree++] = k;
 		}
-		if(ntofree >= nelem(tofree)) {
-			fprintf(stderr, "libcgo: could not obtain pthread_keys\n");
-			fprintf(stderr, "\twanted 0x108 and 0x109\n");
-			fprintf(stderr, "\tgot");
-			for(i=0; i<ntofree; i++)
-				fprintf(stderr, " %#lx", tofree[i]);
-			fprintf(stderr, "\n");
-			abort();
-		}
-		tofree[ntofree++] = k;
+		pthread_setspecific(k, 0);
 	}
 
-	for(i=0; i<ntofree; i++)
-		pthread_key_delete(tofree[i]);
-
 	/*
-	 * We got the keys we wanted.  Make sure that we observe
-	 * updates to k1 at 0x468, to verify that the TLS array
-	 * offset from %gs hasn't changed.
+	 * We got the keys we wanted.  Free the others.
 	 */
-	pthread_setspecific(k1, (void*)0x12345678);
-	asm volatile("movl %%gs:0x468, %0" : "=r"(x));
-
-	pthread_setspecific(k1, (void*)0x87654321);
-	asm volatile("movl %%gs:0x468, %0" : "=r"(y));
-
-	if(x != 0x12345678 || y != 0x87654321) {
-		printf("libcgo: thread-local storage %#lx not at %%gs:0x468 - x=%#x y=%#x\n", k1, x, y);
-		abort();
-	}
+	for(i=0; i<ntofree; i++)
+		pthread_key_delete(tofree[i]);
 }
 
 static void

src/pkg/runtime/cgo/darwin_amd64.c

@@ -8,24 +8,25 @@
 static void* threadentry(void*);
 static pthread_key_t k1, k2;
 
+#define magic1 (0x23581321345589ULL)
+
 static void
 inittls(void)
 {
 	uint64 x, y;
-	pthread_key_t tofree[16], k;
+	pthread_key_t tofree[128], k;
 	int i, ntofree;
 	int havek1, havek2;
 
 	/*
 	 * Same logic, code as darwin_386.c:/inittls, except that words
-	 * are 8 bytes long now, and the thread-local storage starts at 0x60.
-	 * So the offsets are
+	 * are 8 bytes long now, and the thread-local storage starts
+	 * at 0x60 on Leopard / Snow Leopard. So the offsets are
 	 * 0x60+8*0x108 = 0x8a0 and 0x60+8*0x109 = 0x8a8.
 	 *
 	 * The linker and runtime hard-code these constant offsets
-	 * from %gs where we expect to find m and g.  The code
-	 * below verifies that the constants are correct once it has
-	 * obtained the keys.  Known to ../cmd/6l/obj.c:/8a0
+	 * from %gs where we expect to find m and g.
+	 * Known to ../cmd/6l/obj.c:/8a0
 	 * and to ../pkg/runtime/darwin/amd64/sys.s:/8a0
 	 *
 	 * As disgusting as on the 386; same justification.
@@ -35,49 +36,37 @@ inittls(void)
 	ntofree = 0;
 	while(!havek1 || !havek2) {
 		if(pthread_key_create(&k, nil) < 0) {
-			fprintf(stderr, "libcgo: pthread_key_create failed\n");
+			fprintf(stderr, "runtime/cgo: pthread_key_create failed\n");
 			abort();
 		}
-		if(k == 0x108) {
+		pthread_setspecific(k, (void*)magic1);
+		asm volatile("movq %%gs:0x8a0, %0" : "=r"(x));
+		asm volatile("movq %%gs:0x8a8, %0" : "=r"(y));
+		if(x == magic1) {
 			havek1 = 1;
 			k1 = k;
-			continue;
-		}
-		if(k == 0x109) {
+		} else if(y == magic1) {
 			havek2 = 1;
 			k2 = k;
-			continue;
+		} else {
+			if(ntofree >= nelem(tofree)) {
+				fprintf(stderr, "runtime/cgo: could not obtain pthread_keys\n");
+				fprintf(stderr, "\ttried");
+				for(i=0; i<ntofree; i++)
+					fprintf(stderr, " %#x", (unsigned)tofree[i]);
+				fprintf(stderr, "\n");
+				abort();
+			}
+			tofree[ntofree++] = k;
 		}
-		if(ntofree >= nelem(tofree)) {
-			fprintf(stderr, "libcgo: could not obtain pthread_keys\n");
-			fprintf(stderr, "\twanted 0x108 and 0x109\n");
-			fprintf(stderr, "\tgot");
-			for(i=0; i<ntofree; i++)
-				fprintf(stderr, " %#x", (unsigned)tofree[i]);
-			fprintf(stderr, "\n");
-			abort();
-		}
-		tofree[ntofree++] = k;
+		pthread_setspecific(k, 0);
 	}
 
-	for(i=0; i<ntofree; i++)
-		pthread_key_delete(tofree[i]);
-
 	/*
-	 * We got the keys we wanted.  Make sure that we observe
-	 * updates to k1 at 0x8a0, to verify that the TLS array
-	 * offset from %gs hasn't changed.
+	 * We got the keys we wanted.  Free the others.
 	 */
-	pthread_setspecific(k1, (void*)0x123456789abcdef0ULL);
-	asm volatile("movq %%gs:0x8a0, %0" : "=r"(x));
-
-	pthread_setspecific(k2, (void*)0x0fedcba987654321);
-	asm volatile("movq %%gs:0x8a8, %0" : "=r"(y));
-
-	if(x != 0x123456789abcdef0ULL || y != 0x0fedcba987654321) {
-		printf("libcgo: thread-local storage %#x not at %%gs:0x8a0 - x=%#llx y=%#llx\n", (unsigned)k1, x, y);
-		abort();
-	}
+	for(i=0; i<ntofree; i++)
+		pthread_key_delete(tofree[i]);
 }
 
 void

src/pkg/runtime/cgo/nacl_386.c

@@ -1,19 +0,0 @@
-// Copyright 2010 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.
-
-#include "libcgo.h"
-
-static void
-xinitcgo(void)
-{
-}
-
-void (*initcgo)(void) = xinitcgo;
-
-void
-libcgo_sys_thread_start(ThreadStart *ts)
-{
-	// unimplemented
-	*(int*)0 = 0;
-}

src/pkg/runtime/cgo/util.c

@@ -40,7 +40,7 @@ xlibcgo_thread_start(ThreadStart *arg)
 	/* Make our own copy that can persist after we return. */
 	ts = malloc(sizeof *ts);
 	if(ts == nil) {
-		fprintf(stderr, "libcgo: out of memory in thread_start\n");
+		fprintf(stderr, "runtime/cgo: out of memory in thread_start\n");
 		abort();
 	}
 	*ts = *arg;

test/fixedbugs/bug344.go

@@ -0,0 +1,22 @@
+// $G $D/$F.go && $L $F.$A && ./$A.out || echo BUG: bug344
+
+// Copyright 2011 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 main
+
+import "fmt"
+
+func main() {
+	// invalid use of goto.
+	// do whatever you like, just don't crash.
+	i := 42
+	a := []*int{&i, &i, &i, &i}
+	x := a[0]
+	goto start
+	for _, x = range a {
+	start:
+		fmt.Sprint(*x)
+	}
+}