char.c: Fix parsing of escapes
PATCH 1/3 char.c: Fix parsing of escapes
When parsing a string or character constant, the parse_escape() function returns a pointer to the character at which to resume parsing. However, in the case of an hex or octal escape, it was returning a one-past-the-end pointer. Thus, a string like:
char str[3] = "\x61\x62\x63";
was being parsed as:
'\x61', 'x', '6', '2', '\x63'
which, in turn, provokes an 'too long initializer' warning.
Also, fix an off-by-one error in get_char_constant() when setting the 'end' pointer for a TOKEN_CHAR or TOKEN_WIDE_CHAR. Despite the name, the string->length of the token is actually the size of the allocated memory (ie len+1), so we need to compensate by using 'token->string->length - 1'.
--- char.c | 9 ++++++--- validation/escapes.c | 12 ++++++++++++ 2 files changed, 18 insertions(+), 3 deletions(-)
char.c +6 -3
@@ -27,9 +27,10 @@ static const char *parse_escape(const char *p, unsigned *val, const char *end, icase 'x' : {unsigned mask = -(1U << (bits -4 ));for (c =0 ; p < end; c = (c <<4 ) + d) { - d =hexval (*p++); + d =hexval (*p);if (d >16 )break ; + p++;if (c & mask) {warning (pos,"hex escape sequence out of range" );
@@ -42,8 +43,10 @@ static const char *parse_escape(const char *p, unsigned *val, const char *end, iif (p +2 < end) end = p +2 ; c -='0' ; -while (p < end && (d = *p++ -'0' ) <8 ) +while (p < end && (d = *p -'0' ) <8 ) { c = (c <<3 ) + d; + p++; + }if ((c &0400 ) && bits <9 )warning (pos,"octal escape sequence out of range" );
@@ -65,7 +68,7 @@ void get_char_constant(struct token *token, unsigned long long *val)case TOKEN_CHAR :case TOKEN_WIDE_CHAR : p = token->string->data; - end = p + token->string->length; + end = p + token->string->length -1 ;break ; case TOKEN_CHAR_EMBEDDED_0 ... TOKEN_CHAR_EMBEDDED_3: end = p + type -TOKEN_CHAR ;
validation/escapes.c +12 -0
@@ -4,6 +4,16 @@ static int e[] = { '\'', '\"', '\?', '\\',static char *s ="\'\"\?\\ \a\b\f\n\r\t\v \377\xcafe" ;static int bad_e[] = {'\c' ,'\0123' ,'\789' ,'\xdefg' }; + +static char a_hex[3 ] ="\x61\x62\x63" ; +static char b_hex[3 ] ="\x61\x62\x63\x64" ; +static char c_hex[3 ] ="\x61\x62" ; +static char d_hex[3 ] ="\x61" ; + +static char a_oct[3 ] ="\141\142\143" ; +static char b_oct[3 ] ="\141\142\143\144" ; +static char c_oct[3 ] ="\141\142" ; +static char d_oct[3 ] ="\141" ;/* * check-name: Character escape sequences *
@@ -16,5 +26,7 @@ escapes.c:6:30: warning: multi-character character constantescapes.c:6:39: warning: multi-character character constant escapes.c:6:47: warning: hex escape sequence out of range escapes.c:6:47: warning: multi-character character constant +escapes.c:9:24: warning: too long initializer-string for array of char +escapes.c:14:24: warning: too long initializer-string for array of char * check-error-end */
Signature · 1 line
1.8.2
Christopher Li
maintainerQuoted 12 lines · Ramsay Jones
When parsing a string or character constant, the
parse_escape()function returns a pointer to the character at which to resume parsing. However, in the case of an hex or octal escape, it was returning a one-past-the-end pointer. Thus, a string like:char str[3] = "\x61\x62\x63";was being parsed as:
'\x61', 'x', '6', '2', '\x63'
Quoted 7 lines
which, in turn, provokes an 'too long initializer' warning.
Also, fix an off-by-one error in
get_char_constant()when setting the 'end' pointer for aTOKEN_CHARorTOKEN_WIDE_CHAR. Despite the name, thestring->lengthof the token is actually the size of the allocated memory (ie len+1), so we need to compensate by using 'token->string->length- 1'.
Great patch. Applied.
Chris