- first version of scanf parameter check
authorMichael Beck <beck@ipd.info.uni-karlsruhe.de>
Wed, 19 Nov 2008 11:43:30 +0000 (11:43 +0000)
committerMichael Beck <beck@ipd.info.uni-karlsruhe.de>
Wed, 19 Nov 2008 11:43:30 +0000 (11:43 +0000)
[r23803]

format_check.c

index 06e0f7a..f45fa07 100644 (file)
@@ -60,6 +60,13 @@ typedef enum format_length_modifier_t {
        FMT_MOD_I64
 } format_length_modifier_t;
 
+typedef struct format_spec_t {
+       const char    *name;     /**< name of the function */
+       format_kind_t  fmt_kind; /**< kind */
+       unsigned       fmt_idx;  /**< index of the format string */
+       unsigned       arg_idx;  /**< index of the first argument */
+} format_spec_t;
+
 static const char* get_length_modifier_name(const format_length_modifier_t mod)
 {
        static const char* const names[] = {
@@ -146,12 +153,11 @@ static bool atend(vchar_t *self) {
 /**
  * Check printf-style format.
  */
-static void check_format_arguments(const call_argument_t *arg, unsigned idx_fmt,
-               unsigned idx_param)
+static void check_printf_format(const call_argument_t *arg, const format_spec_t *spec)
 {
        /* find format arg */
        unsigned idx = 0;
-       for (; idx < idx_fmt; ++idx)
+       for (; idx < spec->fmt_idx; ++idx)
                arg = arg->next;
 
        const expression_t *fmt_expr = arg->expression;
@@ -176,7 +182,7 @@ static void check_format_arguments(const call_argument_t *arg, unsigned idx_fmt,
                return;
        }
        /* find the real args */
-       for(; idx < idx_param; ++idx)
+       for(; idx < spec->arg_idx; ++idx)
                arg = arg->next;
 
        const source_position_t *pos = &fmt_expr->base.source_position;
@@ -536,6 +542,13 @@ eval_fmt_mod_unsigned:
                        type_t *const arg_type           = arg->expression->base.type;
                        type_t *const arg_skip           = skip_typeref(arg_type);
                        type_t *const expected_type_skip = skip_typeref(expected_type);
+
+                       if (fmt == 'p') {
+                               /* allow any pointer type for %p, not just void */
+                               if (is_type_pointer(arg_skip))
+                                       goto next_arg;
+                       }
+
                        if (is_type_pointer(expected_type_skip)) {
                                if (is_type_pointer(arg_skip)) {
                                        type_t *const exp_to = skip_typeref(expected_type_skip->pointer.points_to);
@@ -574,12 +587,327 @@ next_arg:
        }
 }
 
-static const struct {
-       const char    *name;
-       format_kind_t  fmt_kind;
-       unsigned       fmt_idx;
-       unsigned       arg_idx;
-} builtin_table[] = {
+/**
+ * Check scanf-style format.
+ */
+static void check_scanf_format(const call_argument_t *arg, const format_spec_t *spec)
+{
+       /* find format arg */
+       unsigned idx = 0;
+       for (; idx < spec->fmt_idx; ++idx)
+               arg = arg->next;
+
+       const expression_t *fmt_expr = arg->expression;
+       if (fmt_expr->kind == EXPR_UNARY_CAST_IMPLICIT) {
+               fmt_expr = fmt_expr->unary.value;
+       }
+
+       vchar_t vchar;
+       if (fmt_expr->kind == EXPR_WIDE_STRING_LITERAL) {
+               vchar.string   = &fmt_expr->wide_string.value;
+               vchar.size     = fmt_expr->wide_string.value.size;
+               vchar.first    = wstring_first;
+               vchar.next     = wstring_next;
+               vchar.is_digit = wstring_isdigit;
+       } else if (fmt_expr->kind == EXPR_STRING_LITERAL) {
+               vchar.string   = &fmt_expr->string.value;
+               vchar.size     = fmt_expr->string.value.size;
+               vchar.first    = string_first;
+               vchar.next     = string_next;
+               vchar.is_digit = string_isdigit;
+       } else {
+               return;
+       }
+       /* find the real args */
+       for (; idx < spec->arg_idx; ++idx)
+               arg = arg->next;
+
+       const source_position_t *pos = &fmt_expr->base.source_position;
+       unsigned fmt     = vchar.first(&vchar);
+       unsigned num_fmt = 0;
+       for (; fmt != '\0'; fmt = vchar.next(&vchar)) {
+               if (fmt != '%')
+                       continue;
+               fmt = vchar.next(&vchar);
+
+               if (fmt == '%')
+                       continue;
+
+               ++num_fmt;
+
+               /* length modifier */
+               format_length_modifier_t fmt_mod;
+               switch (fmt) {
+                       case 'h':
+                               fmt = vchar.next(&vchar);
+                               if (fmt == 'h') {
+                                       fmt = vchar.next(&vchar);
+                                       fmt_mod = FMT_MOD_hh;
+                               } else {
+                                       fmt_mod = FMT_MOD_h;
+                               }
+                               break;
+
+                       case 'l':
+                               fmt = vchar.next(&vchar);
+                               if (fmt == 'l') {
+                                       fmt = vchar.next(&vchar);
+                                       fmt_mod = FMT_MOD_ll;
+                               } else {
+                                       fmt_mod = FMT_MOD_l;
+                               }
+                               break;
+
+                       case 'L': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_L;    break;
+                       case 'j': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_j;    break;
+                       case 't': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_t;    break;
+                       case 'z': fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_z;    break;
+                       /* microsoft mode */
+                       case 'w':
+                               if (c_mode & _MS) {
+                                       fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_w;
+                               } else {
+                                       fmt_mod = FMT_MOD_NONE;
+                               }
+                               break;
+                       case 'I':
+                               if (c_mode & _MS) {
+                                       fmt = vchar.next(&vchar); fmt_mod = FMT_MOD_I;
+                                       if (fmt == '3') {
+                                               fmt = vchar.next(&vchar);
+                                               if (fmt == '2') {
+                                                       fmt = vchar.next(&vchar);
+                                                       fmt_mod = FMT_MOD_I32;
+                                               } else {
+                                                       /* rewind */
+                                                       --vchar.position;
+                                               }
+                                       } else if (fmt == '6') {
+                                               fmt = vchar.next(&vchar);
+                                               if (fmt == '4') {
+                                                       fmt = vchar.next(&vchar);
+                                                       fmt_mod = FMT_MOD_I64;
+                                               } else {
+                                                       /* rewind */
+                                                       --vchar.position;
+                                               }
+                                       }
+                               } else {
+                                       fmt_mod = FMT_MOD_NONE;
+                               }
+                               break;
+                       default:
+                               fmt_mod = FMT_MOD_NONE;
+                               break;
+               }
+
+               if (fmt == '\0') {
+                       warningf(pos, "dangling %% in format string");
+                       break;
+               }
+
+               type_t            *expected_type;
+               type_qualifiers_t  expected_qual = TYPE_QUALIFIER_NONE;
+               switch (fmt) {
+                       case 'd':
+                       case 'i':
+                               switch (fmt_mod) {
+                                       case FMT_MOD_NONE: expected_type = type_int;         break;
+                                       case FMT_MOD_hh:   expected_type = type_signed_char; break;
+                                       case FMT_MOD_h:    expected_type = type_short;       break;
+                                       case FMT_MOD_l:    expected_type = type_long;        break;
+                                       case FMT_MOD_ll:   expected_type = type_long_long;   break;
+                                       case FMT_MOD_j:    expected_type = type_intmax_t;    break;
+                                       case FMT_MOD_z:    expected_type = type_ssize_t;     break;
+                                       case FMT_MOD_t:    expected_type = type_ptrdiff_t;   break;
+                                       case FMT_MOD_I:    expected_type = type_ptrdiff_t;   break;
+                                       case FMT_MOD_I32:  expected_type = type_int32;       break;
+                                       case FMT_MOD_I64:  expected_type = type_int64;       break;
+
+                                       default:
+                                               warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                               goto next_arg;
+                               }
+                               break;
+
+                       case 'o':
+                       case 'X':
+                       case 'x':
+                               goto eval_fmt_mod_unsigned;
+
+                       case 'u':
+eval_fmt_mod_unsigned:
+                               switch (fmt_mod) {
+                                       case FMT_MOD_NONE: expected_type = type_unsigned_int;       break;
+                                       case FMT_MOD_hh:   expected_type = type_unsigned_char;      break;
+                                       case FMT_MOD_h:    expected_type = type_unsigned_short;     break;
+                                       case FMT_MOD_l:    expected_type = type_unsigned_long;      break;
+                                       case FMT_MOD_ll:   expected_type = type_unsigned_long_long; break;
+                                       case FMT_MOD_j:    expected_type = type_uintmax_t;          break;
+                                       case FMT_MOD_z:    expected_type = type_size_t;             break;
+                                       case FMT_MOD_t:    expected_type = type_uptrdiff_t;         break;
+                                       case FMT_MOD_I:    expected_type = type_size_t;             break;
+                                       case FMT_MOD_I32:  expected_type = type_unsigned_int32;     break;
+                                       case FMT_MOD_I64:  expected_type = type_unsigned_int64;     break;
+
+                                       default:
+                                               warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                               goto next_arg;
+                               }
+                               break;
+
+                       case 'A':
+                       case 'a':
+                       case 'E':
+                       case 'e':
+                       case 'F':
+                       case 'f':
+                       case 'G':
+                       case 'g':
+                               switch (fmt_mod) {
+                                       case FMT_MOD_l:    /* l modifier is ignored */
+                                       case FMT_MOD_NONE: expected_type = type_double;      break;
+                                       case FMT_MOD_L:    expected_type = type_long_double; break;
+
+                                       default:
+                                               warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                               goto next_arg;
+                               }
+                               break;
+
+                       case 'C':
+                               if (fmt_mod != FMT_MOD_NONE) {
+                                       warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                       goto next_arg;
+                               }
+                               expected_type = type_wchar_t;
+                               break;
+
+                       case 'c':
+                               expected_type = type_int;
+                               switch (fmt_mod) {
+                                       case FMT_MOD_NONE: expected_type = type_int;     break; /* TODO promoted char */
+                                       case FMT_MOD_l:    expected_type = type_wint_t;  break;
+                                       case FMT_MOD_w:    expected_type = type_wchar_t; break;
+
+                                       default:
+                                               warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                               goto next_arg;
+                               }
+                               break;
+
+                       case 'S':
+                               if (fmt_mod != FMT_MOD_NONE) {
+                                       warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                       goto next_arg;
+                               }
+                               expected_type = type_wchar_t_ptr;
+                               expected_qual = TYPE_QUALIFIER_CONST;
+                               break;
+
+                       case 's':
+                               switch (fmt_mod) {
+                                       case FMT_MOD_NONE: expected_type = type_char_ptr;    break;
+                                       case FMT_MOD_l:    expected_type = type_wchar_t_ptr; break;
+                                       case FMT_MOD_w:    expected_type = type_wchar_t_ptr; break;
+
+                                       default:
+                                               warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                               goto next_arg;
+                               }
+                               expected_qual = TYPE_QUALIFIER_CONST;
+                               break;
+
+                       case 'p':
+                               if (fmt_mod != FMT_MOD_NONE) {
+                                       warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                       goto next_arg;
+                               }
+                               expected_type = type_void_ptr;
+                               break;
+
+                       case 'n':
+                               switch (fmt_mod) {
+                                       case FMT_MOD_NONE: expected_type = type_int;         break;
+                                       case FMT_MOD_hh:   expected_type = type_signed_char; break;
+                                       case FMT_MOD_h:    expected_type = type_short;       break;
+                                       case FMT_MOD_l:    expected_type = type_long;        break;
+                                       case FMT_MOD_ll:   expected_type = type_long_long;   break;
+                                       case FMT_MOD_j:    expected_type = type_intmax_t;    break;
+                                       case FMT_MOD_z:    expected_type = type_ssize_t;     break;
+                                       case FMT_MOD_t:    expected_type = type_ptrdiff_t;   break;
+
+                                       default:
+                                               warn_invalid_length_modifier(pos, fmt_mod, fmt);
+                                               goto next_arg;
+                               }
+                               break;
+
+                       default:
+                               warningf(pos, "encountered unknown conversion specifier '%%%C' at position %u", (wint_t)fmt, num_fmt);
+                               goto next_arg;
+               }
+
+               if (arg == NULL) {
+                       warningf(pos, "too few arguments for format string");
+                       return;
+               }
+
+               {       /* create a scope here to prevent warning about the jump to next_arg */
+                       type_t *const arg_type           = arg->expression->base.type;
+                       type_t *const arg_skip           = skip_typeref(arg_type);
+                       type_t *const expected_type_skip = skip_typeref(expected_type);
+
+                       if (! is_type_pointer(arg_skip))
+                               goto error_arg_type;
+                       type_t *const ptr_skip = skip_typeref(arg_skip->pointer.points_to);
+
+                       if (fmt == 'p') {
+                               /* allow any pointer type for %p, not just void */
+                               if (is_type_pointer(ptr_skip))
+                                       goto next_arg;
+                       }
+
+                       if (is_type_pointer(expected_type_skip)) {
+                               if (is_type_pointer(ptr_skip)) {
+                                       type_t *const exp_to = skip_typeref(expected_type_skip->pointer.points_to);
+                                       type_t *const arg_to = skip_typeref(ptr_skip->pointer.points_to);
+                                       if ((arg_to->base.qualifiers & ~expected_qual) == 0 &&
+                                               get_unqualified_type(arg_to) == exp_to) {
+                                               goto next_arg;
+                                       }
+                               }
+                       } else {
+                               if (get_unqualified_type(ptr_skip) == expected_type_skip) {
+                                       goto next_arg;
+                               }
+                       }
+error_arg_type:
+                       if (is_type_valid(arg_skip)) {
+                               warningf(pos,
+                                       "argument type '%T' does not match conversion specifier '%%%s%c' at position %u",
+                                       arg_type, get_length_modifier_name(fmt_mod), (char)fmt, num_fmt);
+                       }
+               }
+next_arg:
+               arg = arg->next;
+       }
+       if (!atend(&vchar)) {
+               warningf(pos, "format string contains NUL");
+       }
+       if (arg != NULL) {
+               unsigned num_args = num_fmt;
+               while (arg != NULL) {
+                       ++num_args;
+                       arg = arg->next;
+               }
+               warningf(pos, "%u argument%s but only %u format string%s",
+                       num_args, num_args != 1 ? "s" : "",
+                       num_fmt, num_fmt != 1 ? "s" : "");
+       }
+}
+
+static const format_spec_t builtin_table[] = {
        { "printf",        FORMAT_PRINTF,   0, 1 },
        { "wprintf",       FORMAT_PRINTF,   0, 1 },
        { "sprintf",       FORMAT_PRINTF,   1, 2 },
@@ -640,7 +968,7 @@ void check_format(const call_expression_t *const call)
        const entity_t        *const entity = func_expr->reference.entity;
        const call_argument_t *      arg    = call->arguments;
 
-       if(false) {
+       if (false) {
                /* the declaration has a GNU format attribute, check it */
        } else {
                /*
@@ -648,12 +976,15 @@ void check_format(const call_expression_t *const call)
                 * This allows to check format even in MS mode or without header included.
                 */
                const char *const name = entity->base.symbol->string;
-               for(size_t i = 0; i < sizeof(builtin_table) / sizeof(builtin_table[0]); ++i) {
-                       if(strcmp(name, builtin_table[i].name) == 0) {
-                               if(builtin_table[i].fmt_kind == FORMAT_PRINTF) {
-                                       check_format_arguments(arg,
-                                                              builtin_table[i].fmt_idx,
-                                                              builtin_table[i].arg_idx);
+               for (size_t i = 0; i < sizeof(builtin_table) / sizeof(builtin_table[0]); ++i) {
+                       if (strcmp(name, builtin_table[i].name) == 0) {
+                               switch (builtin_table[i].fmt_kind) {
+                               case FORMAT_PRINTF:
+                                       check_printf_format(arg, &builtin_table[i]);
+                                       break;
+                               case FORMAT_SCANF:
+                                       check_scanf_format(arg, &builtin_table[i]);
+                                       break;
                                }
                                break;
                        }