}
		b.fill() // buffer is empty
	}
	c := b.buf[b.r]
	b.r++
	b.lastByte = int(c)
	return c, nil
}

// UnreadByte unreads the last byte. Only the most recently read byte can be unread.
//
// UnreadByte returns an error if the most recent method called on the
// [Reader] was not a read operation. Notably, [Reader.Peek], [Reader.Discard], and [Reader.WriteTo] are not
// considered read operations.

	for i := 0; i < len(s); {
		r, size := utf8.DecodeRuneInString(s[i:])
		i += size
		// Officially, ZIP uses CP-437, but many readers use the system's
		// local character encoding. Most encoding are compatible with a large
		// subset of CP-437, which itself is ASCII-like.
		//
		// Forbid 0x7e and 0x5c since EUC-KR and Shift-JIS replace those
		// characters with localized currency and overline characters.

provided

	mpz_t zero;

then cgo would rewrite a reference to C.zero by introducing

	var _C_zero *C.mpz_t

and then replacing all instances of C.zero with (*_C_zero).

Cgo's most interesting translation is for functions.  If xxx is a C
function, then cgo rewrites C.xxx into a new function _C_xxx that
calls the C xxx in a standard pthread.  The new function translates

<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>
A data race is defined as

// the text of the most recently returned token is, and where it was found.
// The underlying scanner elides all spaces except newline, so the input looks like a stream of
// Tokens; original spacing is lost but we don't need it.
type TokenReader interface {
	// Next returns the next token.
	Next() ScanToken
	// The following methods all refer to the most recent token returned by Next.