C code may keep a copy of a Go pointer only as long as the memory it
points to is pinned.

C code may not keep a copy of a Go pointer after the call returns,
unless the memory it points to is pinned with [runtime.Pinner] and the
Pinner is not unpinned while the Go pointer is stored in C memory.
This implies that C code may not keep a copy of a string, slice,

				this.mem.setUint32(offset, ptr, true);
				this.mem.setUint32(offset + 4, 0, true);
				offset += 8;
			});

			// The linker guarantees global data starts from at least wasmMinDataAddr.
			// Keep in sync with cmd/link/internal/ld/data.go:wasmMinDataAddr.
			const wasmMinDataAddr = 4096 + 8192;
			if (offset >= wasmMinDataAddr) {
				throw new Error("total length of command line and environment variables exceeds limit");

			d, ok := decl.(*ast.GenDecl)
			if !ok {
				continue
			}
			for _, spec := range d.Specs {
				if s, ok := spec.(*ast.ImportSpec); ok && s.Path.Value == `"C"` {
					// Replace "C" with _ "unsafe", to keep program valid.
					// (Deleting import statement or clause is not safe if it is followed
					// in the source by an explicit semicolon.)
					f.Edit.Replace(f.offset(s.Path.Pos()), f.offset(s.Path.End()), `_ "unsafe"`)
				}

	back := ss.size - off
	if back > int64(ss.keep) {
		return 0, errDiscardedBytes
	}
	suf := ss.Suffix()
	n = copy(p, suf[len(suf)-int(back):])
	if n != len(p) {
		err = io.EOF
	}
	return
}

func (ss *suffixSaver) Suffix() []byte {
	if len(ss.buf) < ss.keep {
		return ss.buf
	}
	buf := make([]byte, ss.keep)
	n := copy(buf, ss.buf[ss.start:])
	copy(buf[n:], ss.buf[:])

(discussed below).

When preparing any change that is permitted by Go 1 compatibility
but may nonetheless break some existing programs,
we first engineer the change to keep as many existing programs working as possible.
For the remaining programs,
we define a new GODEBUG setting that
allows individual programs to opt back in to the old behavior.

	if _, err := io.ReadFull(r, buf[:4]); err != nil {
		return err
	}
	off := 0
	maybeSig := readBuf(buf[:4])
	if maybeSig.uint32() != dataDescriptorSignature {
		// No data descriptor signature. Keep these four
		// bytes.
		off += 4
	}
	if _, err := io.ReadFull(r, buf[off:12]); err != nil {
		return err
	}
	b := readBuf(buf[:12])
	if b.uint32() != f.CRC32 {
		return ErrChecksum
	}

	tok := in.Stack.Next()
	if tok != scanner.Ident {
		in.expectText("expected identifier after '#'")
	}
	if !in.enabled() {
		// Can only start including again if we are at #else or #endif but also
		// need to keep track of nested #if[n]defs.
		// We let #line through because it might affect errors.
		switch in.Stack.Text() {
		case "else", "endif", "ifdef", "ifndef", "line":
			// Press on.
		default:
			return false
		}

	// is too large we are basically just thrashing the CPU D-cache.
	// So if the result length is larger than an empirically-found
	// limit (8KB), we stop growing the source string once the limit
	// is reached and keep reusing the same source string - that
	// should therefore be always resident in the L1 cache - until we
	// have completed the construction of the result.
	// This yields significant speedups (up to +100%) in cases where

}

// mergePAX merges paxHdrs into hdr for all relevant fields of Header.
func mergePAX(hdr *Header, paxHdrs map[string]string) (err error) {
	for k, v := range paxHdrs {
		if v == "" {
			continue // Keep the original USTAR value
		}
		var id64 int64
		switch k {
		case paxPath:
			hdr.Name = v
		case paxLinkpath:
			hdr.Linkname = v
		case paxUname:
			hdr.Uname = v
		case paxGname:

you are being too clever.
</p>

<p>
Don't be clever.
</p>

<h3 id="overview">Informal Overview</h3>

<p>
Go approaches its memory model in much the same way as the rest of the language,
aiming to keep the semantics simple, understandable, and useful.
This section gives a general overview of the approach and should suffice for most programmers.
The memory model is specified more formally in the next section.
</p>

<p>