From 917da0089dcaa013979a69aaeaeff0c08cbc7e26 Mon Sep 17 00:00:00 2001
From: cinap_lenrek <cinap_lenrek@felloff.net>
Date: Thu, 24 Sep 2015 12:23:17 +0200
Subject: [PATCH] cpp: handle 4 byte utf sequences (21-bit runes)

---
 sys/src/cmd/cpp/lex.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/sys/src/cmd/cpp/lex.c b/sys/src/cmd/cpp/lex.c
index e90423e93..226097b35 100644
--- a/sys/src/cmd/cpp/lex.c
+++ b/sys/src/cmd/cpp/lex.c
@@ -29,6 +29,7 @@
 
 #define	UTF2(c)		((c)>=0xA0 && (c)<0xE0)		/* 2-char UTF seq */
 #define	UTF3(c)		((c)>=0xE0 && (c)<0xF0)		/* 3-char UTF seq */
+#define	UTF4(c)		((c)>=0xF0 && (c)<0xF8)		/* 4-char UTF seq */
 
 /* character classes */
 #define	C_WS	1
@@ -259,7 +260,7 @@ expandlex(void)
 			case C_ALPH:
 				for (j=0; j<=256; j++)
 					if ('a'<=j&&j<='z' || 'A'<=j&&j<='Z'
-					  || UTF2(j) || UTF3(j) || j=='_')
+					  || UTF2(j) || UTF3(j) || UTF4(j) || j=='_')
 						bigfsm[j][fp->state] = nstate;
 				continue;
 			case C_NUM:
@@ -274,7 +275,7 @@ expandlex(void)
 	/* install special cases for ? (trigraphs),  \ (splicing), runes */
 	for (i=0; i<MAXSTATE; i++) {
 		for (j=0; j<0xFF; j++)
-			if (j=='?' || j=='\\' || UTF2(j) || UTF3(j)) {
+			if (j=='?' || j=='\\' || UTF2(j) || UTF3(j) || UTF4(j)) {
 				if (bigfsm[j][i]>0)
 					bigfsm[j][i] = ~bigfsm[j][i];
 				bigfsm[j][i] &= ~QBSBIT;
@@ -393,6 +394,10 @@ gettokens(Tokenrow *trp, int reset)
 					runelen = 3;
 					goto reswitch;
 				}
+				if (UTF4(c)) {
+					runelen = 4;
+					goto reswitch;
+				}
 				error(WARNING, "Lexical botch in cpp");
 				ip += runelen;
 				runelen = 1;