cmd/gc: skip over reported BOMs

This keeps the BOM runes from causing other errors.

R=golang-dev, remyoudompheng
CC=golang-dev
https://golang.org/cl/6625062

diff --git a/src/cmd/gc/lex.c b/src/cmd/gc/lex.c
index 601f1829978d729d2f09818c275be655dbd54891..703bb127df39357cd669c2969bc839a998906d3c 100644 (file)
--- a/src/cmd/gc/lex.c
+++ b/src/cmd/gc/lex.c
@@ -1532,7 +1532,7 @@ yylex(void)
 static int
 getc(void)
 {
-       int c;
+       int c, c1, c2;
 
        c = curio.peekc;
        if(c != 0) {
@@ -1545,8 +1545,20 @@ getc(void)
                c = *curio.cp & 0xff;
                if(c != 0)
                        curio.cp++;
-       } else
+       } else {
+       loop:
                c = Bgetc(curio.bin);
+               if(c == 0xef) {
+                       c1 = Bgetc(curio.bin);
+                       c2 = Bgetc(curio.bin);
+                       if(c1 == 0xbb && c2 == 0xbf) {
+                               yyerrorl(lexlineno, "Unicode (UTF-8) BOM in middle of file");
+                               goto loop;
+                       }
+                       Bungetc(curio.bin);
+                       Bungetc(curio.bin);
+               }
+       }
 
 check:
        switch(c) {
@@ -1597,10 +1609,6 @@ loop:
        if(!fullrune(str, i))
                goto loop;
        c = chartorune(&rune, str);
-       if(rune == BOM) {
-               lineno = lexlineno;
-               yyerror("Unicode (UTF-8) BOM in middle of file");
-       }
        if(rune == Runeerror && c == 1) {
                lineno = lexlineno;
                yyerror("illegal UTF-8 sequence");