Slightly simplify creating an ident from a string on an obstack.
[cparser] / format_check.c
index 9c8710b..06e0f7a 100644 (file)
@@ -23,6 +23,7 @@
 #include "format_check.h"
 #include "symbol_t.h"
 #include "ast_t.h"
+#include "entity_t.h"
 #include "diagnostic.h"
 #include "types.h"
 #include "type_t.h"
@@ -82,7 +83,7 @@ static const char* get_length_modifier_name(const format_length_modifier_t mod)
        return names[mod];
 }
 
-static void warn_invalid_length_modifier(const source_position_t pos,
+static void warn_invalid_length_modifier(const source_position_t *pos,
                                          const format_length_modifier_t mod,
                                          const wchar_rep_t conversion)
 {
@@ -142,9 +143,18 @@ static bool atend(vchar_t *self) {
        return self->position + 1 == self->size;
 }
 
-static void check_format_arguments(const call_argument_t *const fmt_arg, const call_argument_t* arg)
+/**
+ * Check printf-style format.
+ */
+static void check_format_arguments(const call_argument_t *arg, unsigned idx_fmt,
+               unsigned idx_param)
 {
-       const expression_t *fmt_expr = fmt_arg->expression;
+       /* find format arg */
+       unsigned idx = 0;
+       for (; idx < idx_fmt; ++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;
        }
@@ -165,8 +175,13 @@ static void check_format_arguments(const call_argument_t *const fmt_arg, const c
        } else {
                return;
        }
-       const source_position_t    pos     = fmt_expr->base.source_position;
-       unsigned fmt = vchar.first(&vchar);
+       /* find the real args */
+       for(; idx < idx_param; ++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;
@@ -175,6 +190,8 @@ static void check_format_arguments(const call_argument_t *const fmt_arg, const c
                if (fmt == '%')
                        continue;
 
+               ++num_fmt;
+
                format_flags_t fmt_flags = FMT_FLAG_NONE;
                if (fmt == '0') {
                        fmt = vchar.next(&vchar);
@@ -207,14 +224,14 @@ static void check_format_arguments(const call_argument_t *const fmt_arg, const c
 
                                        case ' ':
                                                if (fmt_flags & FMT_FLAG_PLUS) {
-                                                       warningf(pos, "' ' is overridden by prior '+' in conversion specification");
+                                                       warningf(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");
+                                                       warningf(pos, "'+' overrides prior ' ' in conversion specification %u", num_fmt);
                                                }
                                                flag = FMT_FLAG_PLUS;
                                                break;
@@ -222,7 +239,7 @@ static void check_format_arguments(const call_argument_t *const fmt_arg, const c
                                        default: goto break_fmt_flags;
                                }
                                if (fmt_flags & flag) {
-                                       warningf(pos, "repeated flag '%c' in conversion specification", (char)fmt);
+                                       warningf(pos, "repeated flag '%c' in conversion specification %u", (char)fmt, num_fmt);
                                }
                                fmt_flags |= flag;
                                fmt = vchar.next(&vchar);
@@ -231,13 +248,14 @@ break_fmt_flags:
 
                        /* minimum field width */
                        if (fmt == '*') {
+                               fmt = vchar.next(&vchar);
                                if (arg == NULL) {
-                                       warningf(pos, "missing argument for '*' field width in conversion specification");
+                                       warningf(pos, "missing argument for '*' field width in conversion specification %u", num_fmt);
                                        return;
                                }
                                const type_t *const arg_type = arg->expression->base.type;
                                if (arg_type != type_int) {
-                                       warningf(pos, "argument for '*' field width in conversion specification is not an 'int', but an '%T'", arg_type);
+                                       warningf(pos, "argument for '*' field width in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type);
                                }
                                arg = arg->next;
                        } else {
@@ -251,13 +269,14 @@ break_fmt_flags:
                if (fmt == '.') {
                        fmt = vchar.next(&vchar);
                        if (fmt == '*') {
+                               fmt = vchar.next(&vchar);
                                if (arg == NULL) {
-                                       warningf(pos, "missing argument for '*' precision in conversion specification");
+                                       warningf(pos, "missing argument for '*' precision in conversion specification %u", num_fmt);
                                        return;
                                }
                                const type_t *const arg_type = arg->expression->base.type;
                                if (arg_type != type_int) {
-                                       warningf(pos, "argument for '*' precision in conversion specification is not an 'int', but an '%T'", arg_type);
+                                       warningf(pos, "argument for '*' precision in conversion specification %u is not an 'int', but an '%T'", num_fmt, arg_type);
                                }
                                arg = arg->next;
                        } else {
@@ -340,7 +359,7 @@ break_fmt_flags:
                        break;
                }
 
-               const type_t      *expected_type;
+               type_t            *expected_type;
                type_qualifiers_t  expected_qual = TYPE_QUALIFIER_NONE;
                format_flags_t     allowed_flags;
                switch (fmt) {
@@ -371,7 +390,6 @@ break_fmt_flags:
                        case 'x':
                                allowed_flags = FMT_FLAG_MINUS | FMT_FLAG_HASH | FMT_FLAG_ZERO;
                                goto eval_fmt_mod_unsigned;
-                               break;
 
                        case 'u':
                                allowed_flags = FMT_FLAG_MINUS | FMT_FLAG_ZERO;
@@ -445,7 +463,7 @@ eval_fmt_mod_unsigned:
                                }
                                expected_type = type_wchar_t_ptr;
                                expected_qual = TYPE_QUALIFIER_CONST;
-                               allowed_flags = FMT_FLAG_NONE;
+                               allowed_flags = FMT_FLAG_MINUS;
                                break;
 
                        case 's':
@@ -459,7 +477,7 @@ eval_fmt_mod_unsigned:
                                                goto next_arg;
                                }
                                expected_qual = TYPE_QUALIFIER_CONST;
-                               allowed_flags = FMT_FLAG_NONE;
+                               allowed_flags = FMT_FLAG_MINUS;
                                break;
 
                        case 'p':
@@ -490,13 +508,23 @@ eval_fmt_mod_unsigned:
                                break;
 
                        default:
-                               warningf(pos, "encountered unknown conversion specifier '%%%C'", (wint_t)fmt);
+                               warningf(pos, "encountered unknown conversion specifier '%%%C' at position %u", (wint_t)fmt, num_fmt);
                                goto next_arg;
                }
 
-               if ((fmt_flags & ~allowed_flags) != 0) {
-                       /* TODO better warning message text */
-                       warningf(pos, "invalid format flags in conversion specification");
+               format_flags_t wrong_flags = fmt_flags & ~allowed_flags;
+               if (wrong_flags != 0) {
+                       char wrong[8];
+                       int idx = 0;
+                       if (wrong_flags & FMT_FLAG_HASH)  wrong[idx++] = '#';
+                       if (wrong_flags & FMT_FLAG_ZERO)  wrong[idx++] = '0';
+                       if (wrong_flags & FMT_FLAG_MINUS) wrong[idx++] = '-';
+                       if (wrong_flags & FMT_FLAG_SPACE) wrong[idx++] = ' ';
+                       if (wrong_flags & FMT_FLAG_PLUS)  wrong[idx++] = '+';
+                       if (wrong_flags & FMT_FLAG_TICK)  wrong[idx++] = '\'';
+                       wrong[idx] = '\0';
+
+                       warningf(pos, "invalid format flags \"%s\" in conversion specification %%%c at position %u", wrong, fmt, num_fmt);
                }
 
                if (arg == NULL) {
@@ -505,11 +533,12 @@ eval_fmt_mod_unsigned:
                }
 
                {       /* create a scope here to prevent warning about the jump to next_arg */
-                       type_t *const arg_type = arg->expression->base.type;
-                       if (is_type_pointer(expected_type)) {
-                               type_t *const arg_skip = skip_typeref(arg_type);
+                       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(expected_type_skip)) {
                                if (is_type_pointer(arg_skip)) {
-                                       type_t *const exp_to = skip_typeref(expected_type->pointer.points_to);
+                                       type_t *const exp_to = skip_typeref(expected_type_skip->pointer.points_to);
                                        type_t *const arg_to = skip_typeref(arg_skip->pointer.points_to);
                                        if ((arg_to->base.qualifiers & ~expected_qual) == 0 &&
                                                get_unqualified_type(arg_to) == exp_to) {
@@ -517,14 +546,14 @@ eval_fmt_mod_unsigned:
                                        }
                                }
                        } else {
-                               if (get_unqualified_type(skip_typeref(arg_type)) == expected_type) {
+                               if (get_unqualified_type(arg_skip) == expected_type_skip) {
                                        goto next_arg;
                                }
                        }
-                       if (is_type_valid(arg_type)) {
+                       if (is_type_valid(arg_skip)) {
                                warningf(pos,
-                                       "argument type '%T' does not match conversion specifier '%%%s%c'",
-                                       arg_type, get_length_modifier_name(fmt_mod), (char)fmt);
+                                       "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:
@@ -534,10 +563,71 @@ next_arg:
                warningf(pos, "format string contains NUL");
        }
        if (arg != NULL) {
-               warningf(pos, "too many arguments for format string");
+               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 struct {
+       const char    *name;
+       format_kind_t  fmt_kind;
+       unsigned       fmt_idx;
+       unsigned       arg_idx;
+} builtin_table[] = {
+       { "printf",        FORMAT_PRINTF,   0, 1 },
+       { "wprintf",       FORMAT_PRINTF,   0, 1 },
+       { "sprintf",       FORMAT_PRINTF,   1, 2 },
+       { "swprintf",      FORMAT_PRINTF,   1, 2 },
+       { "snprintf",      FORMAT_PRINTF,   2, 3 },
+       { "snwprintf",     FORMAT_PRINTF,   2, 3 },
+       { "fprintf",       FORMAT_PRINTF,   1, 2 },
+       { "fwprintf",      FORMAT_PRINTF,   1, 2 },
+       { "snwprintf",     FORMAT_PRINTF,   2, 3 },
+       { "snwprintf",     FORMAT_PRINTF,   2, 3 },
+
+       { "scanf",         FORMAT_SCANF,    0, 1 },
+       { "wscanf",        FORMAT_SCANF,    0, 1 },
+       { "sscanf",        FORMAT_SCANF,    1, 2 },
+       { "swscanf",       FORMAT_SCANF,    1, 2 },
+       { "fscanf",        FORMAT_SCANF,    1, 2 },
+       { "fwscanf",       FORMAT_SCANF,    1, 2 },
+
+       { "strftime",      FORMAT_STRFTIME, 3, 4 },
+       { "wcstrftime",    FORMAT_STRFTIME, 3, 4 },
+
+       { "strfmon",       FORMAT_STRFMON,  3, 4 },
+
+       /* MS extensions */
+       { "_snprintf",     FORMAT_PRINTF,   2, 3 },
+       { "_snwprintf",    FORMAT_PRINTF,   2, 3 },
+       { "_scrintf",      FORMAT_PRINTF,   0, 1 },
+       { "_scwprintf",    FORMAT_PRINTF,   0, 1 },
+       { "printf_s",      FORMAT_PRINTF,   0, 1 },
+       { "wprintf_s",     FORMAT_PRINTF,   0, 1 },
+       { "sprintf_s",     FORMAT_PRINTF,   3, 4 },
+       { "swprintf_s",    FORMAT_PRINTF,   3, 4 },
+       { "fprintf_s",     FORMAT_PRINTF,   1, 2 },
+       { "fwprintf_s",    FORMAT_PRINTF,   1, 2 },
+       { "_sprintf_l",    FORMAT_PRINTF,   1, 3 },
+       { "_swprintf_l",   FORMAT_PRINTF,   1, 3 },
+       { "_printf_l",     FORMAT_PRINTF,   0, 2 },
+       { "_wprintf_l",    FORMAT_PRINTF,   0, 2 },
+       { "_fprintf_l",    FORMAT_PRINTF,   1, 3 },
+       { "_fwprintf_l",   FORMAT_PRINTF,   1, 3 },
+       { "_printf_s_l",   FORMAT_PRINTF,   0, 2 },
+       { "_wprintf_s_l",  FORMAT_PRINTF,   0, 2 },
+       { "_sprintf_s_l",  FORMAT_PRINTF,   3, 5 },
+       { "_swprintf_s_l", FORMAT_PRINTF,   3, 5 },
+       { "_fprintf_s_l",  FORMAT_PRINTF,   1, 3 },
+       { "_fwprintf_s_l", FORMAT_PRINTF,   1, 3 },
+};
+
 void check_format(const call_expression_t *const call)
 {
        if (!warning.format)
@@ -547,17 +637,26 @@ void check_format(const call_expression_t *const call)
        if (func_expr->kind != EXPR_REFERENCE)
                return;
 
-       const char            *const name = func_expr->reference.symbol->string;
-       const call_argument_t *      arg  = call->arguments;
-       if (strcmp(name, "wprintf") == 0) { /* TODO gammlig */
-               check_format_arguments(arg, arg->next);
-       } else if (strcmp(name, "printf") == 0) {
-               check_format_arguments(arg, arg->next);
-       } else if (strcmp(name, "swprintf") == 0) {
-               arg = arg->next->next; /* skip destination buffer and size */
-               check_format_arguments(arg, arg->next);
-       } else if (strcmp(name, "sprintf") == 0) {
-               arg = arg->next->next; /* skip destination buffer and size */
-               check_format_arguments(arg, arg->next);
+       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 < 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);
+                               }
+                               break;
+                       }
+               }
        }
 }