X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=format_check.c;h=5580d9173039bbcfd4e4aa3eb923da8cce9ad37a;hb=c38eb545c4d2f8835354f15c0a67fb2e8c28cc1e;hp=6350e24f3d54df35b1d0644b8ba9bb6f55d8a28e;hpb=db7888dfadd257dc2387362de1c0926ee1dc701c;p=cparser diff --git a/format_check.c b/format_check.c index 6350e24..5580d91 100644 --- a/format_check.c +++ b/format_check.c @@ -21,12 +21,14 @@ #include +#include "adt/strutil.h" #include "adt/util.h" #include "format_check.h" #include "symbol_t.h" #include "ast_t.h" #include "entity_t.h" #include "diagnostic.h" +#include "parser.h" #include "types.h" #include "type_t.h" #include "warning.h" @@ -88,7 +90,7 @@ static const char* get_length_modifier_name(const format_length_modifier_t mod) [FMT_MOD_I32] = "I32", [FMT_MOD_I64] = "I64" }; - assert(mod < lengthof(names)); + assert((size_t)mod < lengthof(names)); return names[mod]; } @@ -96,10 +98,8 @@ static void warn_invalid_length_modifier(const source_position_t *pos, const format_length_modifier_t mod, const utf32 conversion) { - warningf(pos, - "invalid length modifier '%s' for conversion specifier '%%%c'", - get_length_modifier_name(mod), conversion - ); + char const *const lmod = get_length_modifier_name(mod); + warningf(WARN_FORMAT, pos, "invalid length modifier '%s' for conversion specifier '%%%c'", lmod, conversion); } /** @@ -109,7 +109,7 @@ static int internal_check_printf_format(const expression_t *fmt_expr, const call_argument_t *arg, const format_spec_t *spec) { - while (fmt_expr->kind == EXPR_UNARY_CAST_IMPLICIT) { + while (fmt_expr->kind == EXPR_UNARY_CAST) { fmt_expr = fmt_expr->unary.value; } @@ -145,7 +145,7 @@ static int internal_check_printf_format(const expression_t *fmt_expr, fmt = *(++c); if (fmt == '\0') { - warningf(pos, "dangling %% in format string"); + warningf(WARN_FORMAT, pos, "dangling %% in format string"); break; } if (fmt == '%') @@ -186,14 +186,14 @@ static int internal_check_printf_format(const expression_t *fmt_expr, case ' ': if (fmt_flags & FMT_FLAG_PLUS) { - warningf(pos, "' ' is overridden by prior '+' in conversion specification %u", num_fmt); + warningf(WARN_FORMAT, pos, "' ' is overridden by prior '+' in conversion specification %u", num_fmt); } flag = FMT_FLAG_SPACE; break; case '+': if (fmt_flags & FMT_FLAG_SPACE) { - warningf(pos, "'+' overrides prior ' ' in conversion specification %u", num_fmt); + warningf(WARN_FORMAT, pos, "'+' overrides prior ' ' in conversion specification %u", num_fmt); } flag = FMT_FLAG_PLUS; break; @@ -201,7 +201,7 @@ static int internal_check_printf_format(const expression_t *fmt_expr, default: goto break_fmt_flags; } if (fmt_flags & flag) { - warningf(pos, "repeated flag '%c' in conversion specification %u", (char)fmt, num_fmt); + warningf(WARN_FORMAT, pos, "repeated flag '%c' in conversion specification %u", (char)fmt, num_fmt); } fmt_flags |= flag; fmt = *(++c); @@ -213,12 +213,12 @@ break_fmt_flags: ++num_args; fmt = *(++c); if (arg == NULL) { - warningf(pos, "missing argument for '*' field width in conversion specification %u", num_fmt); + warningf(WARN_FORMAT, pos, "missing argument for '*' field width in conversion specification %u", num_fmt); return -1; } const type_t *const arg_type = arg->expression->base.type; if (arg_type != type_int) { - warningf(pos, "argument for '*' field width in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type); + warningf(WARN_FORMAT, pos, "argument for '*' field width in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type); } arg = arg->next; } else { @@ -230,17 +230,21 @@ break_fmt_flags: /* precision */ if (fmt == '.') { + if (fmt_flags & FMT_FLAG_ZERO) { + warningf(WARN_FORMAT, pos, "'0' flag ignored with precision in conversion specification %u", num_fmt); + } + ++num_args; fmt = *(++c); if (fmt == '*') { fmt = *(++c); if (arg == NULL) { - warningf(pos, "missing argument for '*' precision in conversion specification %u", num_fmt); + warningf(WARN_FORMAT, pos, "missing argument for '*' precision in conversion specification %u", num_fmt); return -1; } const type_t *const arg_type = arg->expression->base.type; if (arg_type != type_int) { - warningf(pos, "argument for '*' precision in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type); + warningf(WARN_FORMAT, pos, "argument for '*' precision in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type); } arg = arg->next; } else { @@ -326,17 +330,17 @@ break_fmt_flags: case 'd': case 'i': switch (fmt_mod) { - case FMT_MOD_NONE: expected_type = type_int; break; - case FMT_MOD_hh: expected_type = type_int; break; /* TODO promoted signed char */ - case FMT_MOD_h: expected_type = type_int; break; /* TODO promoted short */ - 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; + 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); @@ -356,8 +360,8 @@ break_fmt_flags: eval_fmt_mod_unsigned: switch (fmt_mod) { case FMT_MOD_NONE: expected_type = type_unsigned_int; break; - case FMT_MOD_hh: expected_type = type_int; break; /* TODO promoted unsigned char */ - case FMT_MOD_h: expected_type = type_int; break; /* TODO promoted unsigned short */ + 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; @@ -468,10 +472,9 @@ eval_fmt_mod_unsigned: break; default: - warningf(pos, "encountered unknown conversion specifier '%%%c' at position %u", fmt, num_fmt); + warningf(WARN_FORMAT, pos, "encountered unknown conversion specifier '%%%c' at position %u", fmt, num_fmt); if (arg == NULL) { - warningf(pos, "too few arguments for format string"); - return -1; + goto too_few_args; } goto next_arg; } @@ -488,15 +491,16 @@ eval_fmt_mod_unsigned: if (wrong_flags & FMT_FLAG_TICK) *p++ = '\''; *p = '\0'; - warningf(pos, "invalid format flags \"%s\" in conversion specification %%%c at position %u", wrong, fmt, num_fmt); + warningf(WARN_FORMAT, pos, "invalid format flags \"%s\" in conversion specification %%%c at position %u", wrong, fmt, num_fmt); } if (arg == NULL) { - warningf(pos, "too few arguments for format string"); +too_few_args: + warningf(WARN_FORMAT, pos, "too few arguments for format string"); return -1; } - { /* create a scope here to prevent warning about the jump to next_arg */ + { /* 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); @@ -518,11 +522,20 @@ eval_fmt_mod_unsigned: } } else if (get_unqualified_type(arg_skip) == expected_type_skip) { goto next_arg; + } else if (arg->expression->kind == EXPR_UNARY_CAST) { + expression_t const *const expr = arg->expression->unary.value; + type_t *const unprom_type = skip_typeref(expr->base.type); + if (get_unqualified_type(unprom_type) == expected_type_skip) { + goto next_arg; + } + if (expected_type_skip == type_unsigned_int && !is_type_signed(unprom_type)) { + goto next_arg; + } } 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); + source_position_t const *const apos = &arg->expression->base.source_position; + char const *const mod = get_length_modifier_name(fmt_mod); + warningf(WARN_FORMAT, apos, "argument type '%T' does not match conversion specifier '%%%s%c' at position %u", arg_type, mod, (char)fmt, num_fmt); } } next_arg: @@ -530,7 +543,7 @@ next_arg: } assert(fmt == '\0'); if (c+1 < string + size) { - warningf(pos, "format string contains '\\0'"); + warningf(WARN_FORMAT, pos, "format string contains '\\0'"); } return num_args; } @@ -563,10 +576,8 @@ static void check_printf_format(call_argument_t const *arg, for (; arg != NULL; arg = arg->next) ++num_args; if (num_args > (size_t)num_fmt) { - warningf(&fmt_expr->base.source_position, - "%u argument%s but only %u format specifier%s", - num_args, num_args != 1 ? "s" : "", - num_fmt, num_fmt != 1 ? "s" : ""); + source_position_t const *const pos = &fmt_expr->base.source_position; + warningf(WARN_FORMAT, pos, "%u argument%s but only %u format specifier%s", num_args, num_args != 1 ? "s" : "", num_fmt, num_fmt != 1 ? "s" : ""); } } @@ -585,7 +596,7 @@ static void check_scanf_format(const call_argument_t *arg, } const expression_t *fmt_expr = arg->expression; - if (fmt_expr->kind == EXPR_UNARY_CAST_IMPLICIT) { + if (fmt_expr->kind == EXPR_UNARY_CAST) { fmt_expr = fmt_expr->unary.value; } @@ -608,15 +619,28 @@ static void check_scanf_format(const call_argument_t *arg, if (fmt != '%') continue; fmt = *(++c); - if (fmt == '\0') { - warningf(pos, "dangling '%%' in format string"); - break; - } if (fmt == '%') continue; ++num_fmt; + bool suppress_assignment = false; + if (fmt == '*') { + fmt = *++c; + suppress_assignment = true; + } + + size_t width = 0; + if ('0' <= fmt && fmt <= '9') { + do { + width = width * 10 + (fmt - '0'); + fmt = *++c; + } while ('0' <= fmt && fmt <= '9'); + if (width == 0) { + warningf(WARN_FORMAT, pos, "field width is zero at format %u", num_fmt); + } + } + /* look for length modifiers */ format_length_modifier_t fmt_mod = FMT_MOD_NONE; switch (fmt) { @@ -679,7 +703,7 @@ static void check_scanf_format(const call_argument_t *arg, } if (fmt == '\0') { - warningf(pos, "dangling % with conversion specififer in format string"); + warningf(WARN_FORMAT, pos, "dangling %% with conversion specififer in format string"); break; } @@ -754,20 +778,32 @@ static void check_scanf_format(const call_argument_t *arg, goto next_arg; } expected_type = type_wchar_t; - break; + goto check_c_width; - case 'c': - expected_type = type_int; + case 'c': { 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_NONE: expected_type = type_char; break; + case FMT_MOD_l: expected_type = type_wchar_t; break; case FMT_MOD_w: expected_type = type_wchar_t; break; default: warn_invalid_length_modifier(pos, fmt_mod, fmt); goto next_arg; } + +check_c_width: + if (width == 0) + width = 1; + if (!suppress_assignment && arg != NULL) { + type_t *const type = skip_typeref(revert_automatic_type_conversion(arg->expression)); + if (is_type_array(type) && + type->array.size_constant && + width > type->array.size) { + warningf(WARN_FORMAT, pos, "target buffer '%T' is too small for %u characters at format %u", type, width, num_fmt); + } + } break; + } case 'S': if (fmt_mod != FMT_MOD_NONE) { @@ -778,7 +814,7 @@ static void check_scanf_format(const call_argument_t *arg, break; case 's': - case '[': + case '[': { switch (fmt_mod) { case FMT_MOD_NONE: expected_type = type_char; break; case FMT_MOD_l: expected_type = type_wchar_t; break; @@ -788,7 +824,19 @@ static void check_scanf_format(const call_argument_t *arg, warn_invalid_length_modifier(pos, fmt_mod, fmt); goto next_arg; } + + if (!suppress_assignment && + width != 0 && + arg != NULL) { + type_t *const type = skip_typeref(revert_automatic_type_conversion(arg->expression)); + if (is_type_array(type) && + type->array.size_constant && + width >= type->array.size) { + warningf(WARN_FORMAT, pos, "target buffer '%T' is too small for %u characters and \\0 at format %u", type, width, num_fmt); + } + } break; + } case 'p': if (fmt_mod != FMT_MOD_NONE) { @@ -798,7 +846,11 @@ static void check_scanf_format(const call_argument_t *arg, expected_type = type_void_ptr; break; - case 'n': + case 'n': { + if (suppress_assignment) { + warningf(WARN_FORMAT, pos, "conversion '%n' cannot be suppressed with '*' at format %u", num_fmt); + } + switch (fmt_mod) { case FMT_MOD_NONE: expected_type = type_int; break; case FMT_MOD_hh: expected_type = type_signed_char; break; @@ -814,23 +866,27 @@ static void check_scanf_format(const call_argument_t *arg, goto next_arg; } break; + } default: - warningf(pos, "encountered unknown conversion specifier '%%%c' at format %u", - fmt, num_fmt); - if (arg == NULL) { - warningf(pos, "too few arguments for format string"); - return; - } + warningf(WARN_FORMAT, pos, "encountered unknown conversion specifier '%%%c' at format %u", fmt, num_fmt); + if (suppress_assignment) + continue; + if (arg == NULL) + goto too_few_args; goto next_arg; } + if (suppress_assignment) + continue; + if (arg == NULL) { - warningf(pos, "too few arguments for format string"); +too_few_args: + warningf(WARN_FORMAT, pos, "too few arguments for format string"); return; } - { /* create a scope here to prevent warning about the jump to next_arg */ + { /* 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); @@ -858,9 +914,9 @@ static void check_scanf_format(const call_argument_t *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); + source_position_t const *const apos = &arg->expression->base.source_position; + char const *const mod = get_length_modifier_name(fmt_mod); + warningf(WARN_FORMAT, apos, "argument type '%T' does not match conversion specifier '%%%s%c' at position %u", arg_type, mod, (char)fmt, num_fmt); } } next_arg: @@ -868,7 +924,7 @@ next_arg: } assert(fmt == '\0'); if (c+1 < string + size) { - warningf(pos, "format string contains '\\0'"); + warningf(WARN_FORMAT, pos, "format string contains '\\0'"); } if (arg != NULL) { unsigned num_args = num_fmt; @@ -876,9 +932,7 @@ next_arg: ++num_args; arg = arg->next; } - warningf(pos, "%u argument%s but only %u format specifier%s", - num_args, num_args != 1 ? "s" : "", - num_fmt, num_fmt != 1 ? "s" : ""); + warningf(WARN_FORMAT, pos, "%u argument%s but only %u format specifier%s", num_args, num_args != 1 ? "s" : "", num_fmt, num_fmt != 1 ? "s" : ""); } } @@ -933,7 +987,7 @@ static const format_spec_t builtin_table[] = { void check_format(const call_expression_t *const call) { - if (!warning.format) + if (!is_warn_on(WARN_FORMAT)) return; const expression_t *const func_expr = call->function; @@ -943,31 +997,27 @@ 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) { - /* the declaration has a GNU format attribute, check it */ - } else { - /* - * For some functions we always check the format, even if it was not - * specified. 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 < lengthof(builtin_table); ++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; - case FORMAT_STRFTIME: - case FORMAT_STRFMON: - /* TODO: implement other cases */ - break; - } + /* + * For some functions we always check the format, even if it was not + * specified. 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 < lengthof(builtin_table); ++i) { + if (streq(name, builtin_table[i].name)) { + 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; + case FORMAT_STRFTIME: + case FORMAT_STRFMON: + /* TODO: implement other cases */ break; } + break; } } }