When we Japanese write some Japanese text in spec file like
this: Summary(ja): xxxxxxxx, rpm warns this file as binary
and exits if we use JIS kanji code for this text.
In Japan, commonly used kanji code for Linux is EUC-JP and
no problem occurs when we use EUC-JP in spec file, but
someone want to use JIS kanji code because JIS (ISO-2022-JP)
is also used widely, especially in mail, and EUC-JP itself
is based on JIS.
So please omit checking out ESC (0x1b) at this point:

diff -uNr rpm-2.5.5/build.c rpm-2.5.5.new/build.c
--- rpm-2.5.5/build.c	Fri Sep 25 05:23:16 1998
+++ rpm-2.5.5.new/build.c	Sat Apr  3 15:28:13 1999
@@ -128,7 +128,7 @@
     close(fd);
     s = buf;
     while(count--) {
-	if (! (isprint(*s) || isspace(*s))) {
+	if (! (isprint(*s) || isspace(*s) || (*s == 0x1b))) {
 	    fprintf(stderr, _("File contains non-printable
characters(%c): %s\n"), *s,
 		    specfile);
 	    ret