2 * This file is part of cparser.
3 * Copyright (C) 2007-2008 Matthias Braun <matze@braunis.de>
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License
7 * as published by the Free Software Foundation; either version 2
8 * of the License, or (at your option) any later version.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License for more details.
15 * You should have received a copy of the GNU General Public License
16 * along with this program; if not, write to the Free Software
17 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
23 #include "format_check.h"
26 #include "diagnostic.h"
30 #include "lang_features.h"
32 typedef enum format_flag_t {
34 FMT_FLAG_HASH = 1U << 0,
35 FMT_FLAG_ZERO = 1U << 1,
36 FMT_FLAG_MINUS = 1U << 2,
37 FMT_FLAG_SPACE = 1U << 3,
38 FMT_FLAG_PLUS = 1U << 4,
39 FMT_FLAG_TICK = 1U << 5
42 typedef unsigned format_flags_t;
44 typedef enum format_length_modifier_t {
55 /* only in microsoft mode */
60 } format_length_modifier_t;
62 static const char* get_length_modifier_name(const format_length_modifier_t mod)
64 static const char* const names[] = {
75 /* only in microsoft mode */
78 [FMT_MOD_I32] = "I32",
81 assert(mod < sizeof(names) / sizeof(*names));
85 static void warn_invalid_length_modifier(const source_position_t *pos,
86 const format_length_modifier_t mod,
87 const wchar_rep_t conversion)
90 "invalid length modifier '%s' for conversion specifier '%%%c'",
91 get_length_modifier_name(mod), conversion
95 typedef struct vchar_t vchar_t;
97 const void *string; /**< the string */
98 size_t position; /**< current position */
99 size_t size; /**< size of the string */
101 /** return the first character of the string and setthe position to 0. */
102 unsigned (*first)(vchar_t *self);
103 /** return the next character of the string */
104 unsigned (*next)(vchar_t *self);
105 /** return non_zero if the given character is a digit */
106 int (*is_digit)(unsigned vchar);
109 static unsigned string_first(vchar_t *self) {
111 const string_t *string = self->string;
112 return string->begin[0];
115 static unsigned string_next(vchar_t *self) {
117 const string_t *string = self->string;
118 return string->begin[self->position];
121 static int string_isdigit(unsigned vchar) {
122 return isdigit(vchar);
125 static unsigned wstring_first(vchar_t *self) {
127 const wide_string_t *wstring = self->string;
128 return wstring->begin[0];
131 static unsigned wstring_next(vchar_t *self) {
133 const wide_string_t *wstring = self->string;
134 return wstring->begin[self->position];
137 static int wstring_isdigit(unsigned vchar) {
138 return iswdigit(vchar);
141 static bool atend(vchar_t *self) {
142 return self->position + 1 == self->size;
146 * Check printf-style format.
148 static void check_format_arguments(const call_argument_t *arg, unsigned idx_fmt,
151 /* find format arg */
153 for (; idx < idx_fmt; ++idx)
156 const expression_t *fmt_expr = arg->expression;
157 if (fmt_expr->kind == EXPR_UNARY_CAST_IMPLICIT) {
158 fmt_expr = fmt_expr->unary.value;
162 if (fmt_expr->kind == EXPR_WIDE_STRING_LITERAL) {
163 vchar.string = &fmt_expr->wide_string.value;
164 vchar.size = fmt_expr->wide_string.value.size;
165 vchar.first = wstring_first;
166 vchar.next = wstring_next;
167 vchar.is_digit = wstring_isdigit;
168 } else if (fmt_expr->kind == EXPR_STRING_LITERAL) {
169 vchar.string = &fmt_expr->string.value;
170 vchar.size = fmt_expr->string.value.size;
171 vchar.first = string_first;
172 vchar.next = string_next;
173 vchar.is_digit = string_isdigit;
177 /* find the real args */
178 for(; idx < idx_param; ++idx)
181 const source_position_t *pos = &fmt_expr->base.source_position;
182 unsigned fmt = vchar.first(&vchar);
183 unsigned num_fmt = 0;
184 for (; fmt != '\0'; fmt = vchar.next(&vchar)) {
187 fmt = vchar.next(&vchar);
194 format_flags_t fmt_flags = FMT_FLAG_NONE;
196 fmt = vchar.next(&vchar);
197 fmt_flags |= FMT_FLAG_ZERO;
200 /* argument selector or minimum field width */
201 if (vchar.is_digit(fmt)) {
203 fmt = vchar.next(&vchar);
204 } while (vchar.is_digit(fmt));
206 /* digit string was ... */
208 /* ... argument selector */
209 fmt_flags = FMT_FLAG_NONE; /* reset possibly set 0-flag */
213 /* ... minimum field width */
219 case '#': flag = FMT_FLAG_HASH; break;
220 case '0': flag = FMT_FLAG_ZERO; break;
221 case '-': flag = FMT_FLAG_MINUS; break;
222 case '\'': flag = FMT_FLAG_TICK; break;
225 if (fmt_flags & FMT_FLAG_PLUS) {
226 warningf(pos, "' ' is overridden by prior '+' in conversion specification %u", num_fmt);
228 flag = FMT_FLAG_SPACE;
232 if (fmt_flags & FMT_FLAG_SPACE) {
233 warningf(pos, "'+' overrides prior ' ' in conversion specification %u", num_fmt);
235 flag = FMT_FLAG_PLUS;
238 default: goto break_fmt_flags;
240 if (fmt_flags & flag) {
241 warningf(pos, "repeated flag '%c' in conversion specification %u", (char)fmt, num_fmt);
244 fmt = vchar.next(&vchar);
248 /* minimum field width */
250 fmt = vchar.next(&vchar);
252 warningf(pos, "missing argument for '*' field width in conversion specification %u", num_fmt);
255 const type_t *const arg_type = arg->expression->base.type;
256 if (arg_type != type_int) {
257 warningf(pos, "argument for '*' field width in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type);
261 while (vchar.is_digit(fmt)) {
262 fmt = vchar.next(&vchar);
269 fmt = vchar.next(&vchar);
271 fmt = vchar.next(&vchar);
273 warningf(pos, "missing argument for '*' precision in conversion specification %u", num_fmt);
276 const type_t *const arg_type = arg->expression->base.type;
277 if (arg_type != type_int) {
278 warningf(pos, "argument for '*' precision in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type);
282 /* digit string may be omitted */
283 while (vchar.is_digit(fmt)) {
284 fmt = vchar.next(&vchar);
289 /* length modifier */
290 format_length_modifier_t fmt_mod;
293 fmt = vchar.next(&vchar);
295 fmt = vchar.next(&vchar);
296 fmt_mod = FMT_MOD_hh;
303 fmt = vchar.next(&vchar);
305 fmt = vchar.next(&vchar);
306 fmt_mod = FMT_MOD_ll;
312 case 'L': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_L; break;
313 case 'j': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_j; break;
314 case 't': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_t; break;
315 case 'z': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_z; break;
316 case 'q': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_q; break;
320 fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_w;
322 fmt_mod = FMT_MOD_NONE;
327 fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_I;
329 fmt = vchar.next(&vchar);
331 fmt = vchar.next(&vchar);
332 fmt_mod = FMT_MOD_I32;
337 } else if (fmt == '6') {
338 fmt = vchar.next(&vchar);
340 fmt = vchar.next(&vchar);
341 fmt_mod = FMT_MOD_I64;
348 fmt_mod = FMT_MOD_NONE;
352 fmt_mod = FMT_MOD_NONE;
357 warningf(pos, "dangling %% in format string");
361 type_t *expected_type;
362 type_qualifiers_t expected_qual = TYPE_QUALIFIER_NONE;
363 format_flags_t allowed_flags;
368 case FMT_MOD_NONE: expected_type = type_int; break;
369 case FMT_MOD_hh: expected_type = type_int; break; /* TODO promoted signed char */
370 case FMT_MOD_h: expected_type = type_int; break; /* TODO promoted short */
371 case FMT_MOD_l: expected_type = type_long; break;
372 case FMT_MOD_ll: expected_type = type_long_long; break;
373 case FMT_MOD_j: expected_type = type_intmax_t; break;
374 case FMT_MOD_z: expected_type = type_ssize_t; break;
375 case FMT_MOD_t: expected_type = type_ptrdiff_t; break;
376 case FMT_MOD_I: expected_type = type_ptrdiff_t; break;
377 case FMT_MOD_I32: expected_type = type_int32; break;
378 case FMT_MOD_I64: expected_type = type_int64; break;
381 warn_invalid_length_modifier(pos, fmt_mod, fmt);
384 allowed_flags = FMT_FLAG_MINUS | FMT_FLAG_SPACE | FMT_FLAG_PLUS | FMT_FLAG_ZERO;
390 allowed_flags = FMT_FLAG_MINUS | FMT_FLAG_HASH | FMT_FLAG_ZERO;
391 goto eval_fmt_mod_unsigned;
394 allowed_flags = FMT_FLAG_MINUS | FMT_FLAG_ZERO;
395 eval_fmt_mod_unsigned:
397 case FMT_MOD_NONE: expected_type = type_unsigned_int; break;
398 case FMT_MOD_hh: expected_type = type_int; break; /* TODO promoted unsigned char */
399 case FMT_MOD_h: expected_type = type_int; break; /* TODO promoted unsigned short */
400 case FMT_MOD_l: expected_type = type_unsigned_long; break;
401 case FMT_MOD_ll: expected_type = type_unsigned_long_long; break;
402 case FMT_MOD_j: expected_type = type_uintmax_t; break;
403 case FMT_MOD_z: expected_type = type_size_t; break;
404 case FMT_MOD_t: expected_type = type_uptrdiff_t; break;
405 case FMT_MOD_I: expected_type = type_size_t; break;
406 case FMT_MOD_I32: expected_type = type_unsigned_int32; break;
407 case FMT_MOD_I64: expected_type = type_unsigned_int64; break;
410 warn_invalid_length_modifier(pos, fmt_mod, fmt);
424 case FMT_MOD_l: /* l modifier is ignored */
425 case FMT_MOD_NONE: expected_type = type_double; break;
426 case FMT_MOD_L: expected_type = type_long_double; break;
429 warn_invalid_length_modifier(pos, fmt_mod, fmt);
432 allowed_flags = FMT_FLAG_MINUS | FMT_FLAG_SPACE | FMT_FLAG_PLUS | FMT_FLAG_HASH | FMT_FLAG_ZERO;
436 if (fmt_mod != FMT_MOD_NONE) {
437 warn_invalid_length_modifier(pos, fmt_mod, fmt);
440 expected_type = type_wchar_t;
441 allowed_flags = FMT_FLAG_NONE;
445 expected_type = type_int;
447 case FMT_MOD_NONE: expected_type = type_int; break; /* TODO promoted char */
448 case FMT_MOD_l: expected_type = type_wint_t; break;
449 case FMT_MOD_w: expected_type = type_wchar_t; break;
452 warn_invalid_length_modifier(pos, fmt_mod, fmt);
455 allowed_flags = FMT_FLAG_NONE;
459 if (fmt_mod != FMT_MOD_NONE) {
460 warn_invalid_length_modifier(pos, fmt_mod, fmt);
463 expected_type = type_wchar_t_ptr;
464 expected_qual = TYPE_QUALIFIER_CONST;
465 allowed_flags = FMT_FLAG_MINUS;
470 case FMT_MOD_NONE: expected_type = type_char_ptr; break;
471 case FMT_MOD_l: expected_type = type_wchar_t_ptr; break;
472 case FMT_MOD_w: expected_type = type_wchar_t_ptr; break;
475 warn_invalid_length_modifier(pos, fmt_mod, fmt);
478 expected_qual = TYPE_QUALIFIER_CONST;
479 allowed_flags = FMT_FLAG_MINUS;
483 if (fmt_mod != FMT_MOD_NONE) {
484 warn_invalid_length_modifier(pos, fmt_mod, fmt);
487 expected_type = type_void_ptr;
488 allowed_flags = FMT_FLAG_NONE;
493 case FMT_MOD_NONE: expected_type = type_int_ptr; break;
494 case FMT_MOD_hh: expected_type = type_signed_char_ptr; break;
495 case FMT_MOD_h: expected_type = type_short_ptr; break;
496 case FMT_MOD_l: expected_type = type_long_ptr; break;
497 case FMT_MOD_ll: expected_type = type_long_long_ptr; break;
498 case FMT_MOD_j: expected_type = type_intmax_t_ptr; break;
499 case FMT_MOD_z: expected_type = type_ssize_t_ptr; break;
500 case FMT_MOD_t: expected_type = type_ptrdiff_t_ptr; break;
503 warn_invalid_length_modifier(pos, fmt_mod, fmt);
506 allowed_flags = FMT_FLAG_NONE;
510 warningf(pos, "encountered unknown conversion specifier '%%%C' at position %u", (wint_t)fmt, num_fmt);
514 format_flags_t wrong_flags = fmt_flags & ~allowed_flags;
515 if (wrong_flags != 0) {
518 if (wrong_flags & FMT_FLAG_HASH) wrong[idx++] = '#';
519 if (wrong_flags & FMT_FLAG_ZERO) wrong[idx++] = '0';
520 if (wrong_flags & FMT_FLAG_MINUS) wrong[idx++] = '-';
521 if (wrong_flags & FMT_FLAG_SPACE) wrong[idx++] = ' ';
522 if (wrong_flags & FMT_FLAG_PLUS) wrong[idx++] = '+';
523 if (wrong_flags & FMT_FLAG_TICK) wrong[idx++] = '\'';
526 warningf(pos, "invalid format flags \"%s\" in conversion specification %%%c at position %u", wrong, fmt, num_fmt);
530 warningf(pos, "too few arguments for format string");
534 { /* create a scope here to prevent warning about the jump to next_arg */
535 type_t *const arg_type = arg->expression->base.type;
536 type_t *const arg_skip = skip_typeref(arg_type);
537 type_t *const expected_type_skip = skip_typeref(expected_type);
538 if (is_type_pointer(expected_type_skip)) {
539 if (is_type_pointer(arg_skip)) {
540 type_t *const exp_to = skip_typeref(expected_type_skip->pointer.points_to);
541 type_t *const arg_to = skip_typeref(arg_skip->pointer.points_to);
542 if ((arg_to->base.qualifiers & ~expected_qual) == 0 &&
543 get_unqualified_type(arg_to) == exp_to) {
548 if (get_unqualified_type(arg_skip) == expected_type_skip) {
552 if (is_type_valid(arg_skip)) {
554 "argument type '%T' does not match conversion specifier '%%%s%c' at position %u",
555 arg_type, get_length_modifier_name(fmt_mod), (char)fmt, num_fmt);
561 if (!atend(&vchar)) {
562 warningf(pos, "format string contains NUL");
565 unsigned num_args = num_fmt;
566 while (arg != NULL) {
570 warningf(pos, "%u argument%s but only %u format string%s",
571 num_args, num_args != 1 ? "s" : "",
572 num_fmt, num_fmt != 1 ? "s" : "");
576 static const struct {
578 format_kind_t fmt_kind;
581 } builtin_table[] = {
582 { "printf", FORMAT_PRINTF, 0, 1 },
583 { "wprintf", FORMAT_PRINTF, 0, 1 },
584 { "sprintf", FORMAT_PRINTF, 1, 2 },
585 { "swprintf", FORMAT_PRINTF, 1, 2 },
586 { "snprintf", FORMAT_PRINTF, 2, 3 },
587 { "snwprintf", FORMAT_PRINTF, 2, 3 },
588 { "fprintf", FORMAT_PRINTF, 1, 2 },
589 { "fwprintf", FORMAT_PRINTF, 1, 2 },
590 { "snwprintf", FORMAT_PRINTF, 2, 3 },
591 { "snwprintf", FORMAT_PRINTF, 2, 3 },
593 { "scanf", FORMAT_SCANF, 0, 1 },
594 { "wscanf", FORMAT_SCANF, 0, 1 },
595 { "sscanf", FORMAT_SCANF, 1, 2 },
596 { "swscanf", FORMAT_SCANF, 1, 2 },
597 { "fscanf", FORMAT_SCANF, 1, 2 },
598 { "fwscanf", FORMAT_SCANF, 1, 2 },
600 { "strftime", FORMAT_STRFTIME, 3, 4 },
601 { "wcstrftime", FORMAT_STRFTIME, 3, 4 },
603 { "strfmon", FORMAT_STRFMON, 3, 4 },
606 { "_snprintf", FORMAT_PRINTF, 2, 3 },
607 { "_snwprintf", FORMAT_PRINTF, 2, 3 },
608 { "_scrintf", FORMAT_PRINTF, 0, 1 },
609 { "_scwprintf", FORMAT_PRINTF, 0, 1 },
610 { "printf_s", FORMAT_PRINTF, 0, 1 },
611 { "wprintf_s", FORMAT_PRINTF, 0, 1 },
612 { "sprintf_s", FORMAT_PRINTF, 3, 4 },
613 { "swprintf_s", FORMAT_PRINTF, 3, 4 },
614 { "fprintf_s", FORMAT_PRINTF, 1, 2 },
615 { "fwprintf_s", FORMAT_PRINTF, 1, 2 },
616 { "_sprintf_l", FORMAT_PRINTF, 1, 3 },
617 { "_swprintf_l", FORMAT_PRINTF, 1, 3 },
618 { "_printf_l", FORMAT_PRINTF, 0, 2 },
619 { "_wprintf_l", FORMAT_PRINTF, 0, 2 },
620 { "_fprintf_l", FORMAT_PRINTF, 1, 3 },
621 { "_fwprintf_l", FORMAT_PRINTF, 1, 3 },
622 { "_printf_s_l", FORMAT_PRINTF, 0, 2 },
623 { "_wprintf_s_l", FORMAT_PRINTF, 0, 2 },
624 { "_sprintf_s_l", FORMAT_PRINTF, 3, 5 },
625 { "_swprintf_s_l", FORMAT_PRINTF, 3, 5 },
626 { "_fprintf_s_l", FORMAT_PRINTF, 1, 3 },
627 { "_fwprintf_s_l", FORMAT_PRINTF, 1, 3 },
630 void check_format(const call_expression_t *const call)
635 const expression_t *const func_expr = call->function;
636 if (func_expr->kind != EXPR_REFERENCE)
639 const declaration_t *const decl = func_expr->reference.declaration;
640 const call_argument_t * arg = call->arguments;
643 /* the declaration has a GNU format attribute, check it */
646 * For some functions we always check the format, even if it was not specified.
647 * This allows to check format even in MS mode or without header included.
649 const char *const name = decl->symbol->string;
650 for(size_t i = 0; i < sizeof(builtin_table) / sizeof(builtin_table[0]); ++i) {
651 if(strcmp(name, builtin_table[i].name) == 0) {
652 if(builtin_table[i].fmt_kind == FORMAT_PRINTF) {
653 check_format_arguments(arg,
654 builtin_table[i].fmt_idx,
655 builtin_table[i].arg_idx);