build: omit named test source operands

cmd/wcc/lex.c

@@ -358,6 +358,78 @@ skipws(Lex *l)
 	}
 }
 
+/* Return the next initial import token without ever lexing the first ordinary
+ * declaration token. Go's named-file reader consumes header trivia while
+ * looking for another import, so NUL/comment diagnostics in that trivia remain
+ * visible; a malformed UTF-8 byte or non-leading BOM that itself begins the
+ * ordinary body is outside the header and must not be diagnosed here. */
+int
+lexheaderimport(Lex *l, Tok *out)
+{
+	for (;;) {
+		if (l->pos >= l->srclen)
+			return 0;
+		unsigned char c = (unsigned char)l->src[l->pos];
+		if (c == 0) {
+			/* The loader necessarily reaches this byte while deciding whether
+			 * another import follows. Preserve the source decoder diagnostic. */
+			lskipnul(l);
+			return 0;
+		}
+		if ((c >= 0x80 && !utf8bytevalid(l->src, l->srclen, l->pos))
+		    || bomat(l->src, l->srclen, l->pos))
+			return 0;
+		if (c == ' ' || c == '\t' || c == '\r' || c == '\n') {
+			lget(l);
+			continue;
+		}
+		if (c == '/' && l->pos + 1 < l->srclen
+		    && l->src[l->pos + 1] == '/') {
+			lget(l); lget(l);
+			linecomment(l);
+			if (l->errs != 0)
+				return 0;
+			continue;
+		}
+		if (c == '/' && l->pos + 1 < l->srclen
+		    && l->src[l->pos + 1] == '*') {
+			lget(l); lget(l);
+			int prev = -1;
+			for (;;) {
+				int x = lget(l);
+				if (x < 0) {
+					Pos p = lpos(l);
+					errorf(p, "unterminated /* comment");
+					l->errs++;
+					return 0;
+				}
+				if (prev == '*' && x == '/')
+					break;
+				prev = x;
+			}
+			if (l->errs != 0)
+				return 0;
+			continue;
+		}
+		break;
+	}
+
+	/* A compiler-owned source-boundary directive is an ordinary declaration
+	 * boundary here, just as it was when lexnext eagerly exposed the token. */
+	if (l->modreset || l->modpath != NULL)
+		return 0;
+	static const char kw[] = "import";
+	const u64 n = sizeof kw - 1;
+	if (l->srclen - l->pos < n
+	    || memcmp(l->src + l->pos, kw, n) != 0)
+		return 0;
+	if (l->srclen - l->pos > n
+	    && isidcont((unsigned char)l->src[l->pos + n]))
+		return 0;
+	*out = lexnext(l);
+	return out->kind == TK_USE;
+}
+
 static u64
 parseint(const char *s, u64 n, int base, int *ok)
 {