Merge branch '1.1'

[jansson.git] / src / load.c

diff --git a/src/load.c b/src/load.c
index 1e2ad2a..4d08139 100644 (file)
--- a/src/load.c
+++ b/src/load.c
@@ -8,12 +8,13 @@
 #define _GNU_SOURCE
 #include <ctype.h>
 #include <errno.h>
+#include <limits.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 #include <stdarg.h>
-#include <unistd.h>
 #include <assert.h>
+#include <stdint.h>
 
 #include <jansson.h>
 #include "jansson_private.h"
@@ -134,10 +135,7 @@ static char stream_get(stream_t *stream, json_error_t *error)
 
         c = stream->buffer[0];
 
-        if(c == EOF && stream->eof(stream->data))
-            return EOF;
-
-        if(c < 0)
+        if((unsigned char)c >= 0x80 && c != (char)EOF)
         {
             /* multi-byte UTF-8 sequence */
             int i, count;
@@ -151,7 +149,7 @@ static char stream_get(stream_t *stream, json_error_t *error)
             for(i = 1; i < count; i++)
                 stream->buffer[i] = stream->get(stream->data);
 
-            if(!utf8_check_full(stream->buffer, count))
+            if(!utf8_check_full(stream->buffer, count, NULL))
                 goto out;
 
             stream->stream_pos += count;
@@ -224,10 +222,10 @@ static void lex_save_cached(lex_t *lex)
 }
 
 /* assumes that str points to 'u' plus at least 4 valid hex digits */
-static int decode_unicode_escape(const char *str)
+static int32_t decode_unicode_escape(const char *str)
 {
     int i;
-    int value = 0;
+    int32_t value = 0;
 
     assert(str[0] == 'u');
 
@@ -257,17 +255,17 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
     lex->value.string = NULL;
     lex->token = TOKEN_INVALID;
 
-    /* skip the " */
     c = lex_get_save(lex, error);
 
     while(c != '"') {
-        if(c == EOF) {
+        if(c == (char)EOF) {
+            lex_unget_unsave(lex, c);
             if(lex_eof(lex))
                 error_set(error, lex, "premature end of input");
             goto out;
         }
 
-        else if(0 <= c && c <= 0x1F) {
+        else if((unsigned char)c <= 0x1F) {
             /* control character */
             lex_unget_unsave(lex, c);
             if(c == '\n')
@@ -328,7 +326,7 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
             if(*p == 'u') {
                 char buffer[4];
                 int length;
-                int value;
+                int32_t value;
 
                 value = decode_unicode_escape(p);
                 p += 5;
@@ -336,14 +334,15 @@ static void lex_scan_string(lex_t *lex, json_error_t *error)
                 if(0xD800 <= value && value <= 0xDBFF) {
                     /* surrogate pair */
                     if(*p == '\\' && *(p + 1) == 'u') {
-                        int value2 = decode_unicode_escape(++p);
+                        int32_t value2 = decode_unicode_escape(++p);
                         p += 5;
 
                         if(0xDC00 <= value2 && value2 <= 0xDFFF) {
                             /* valid second surrogate */
-                            value = ((value - 0xD800) << 10) +
-                                    (value2 - 0xDC00) +
-                                    0x10000;
+                            value =
+                                ((value - 0xD800) << 10) +
+                                (value2 - 0xDC00) +
+                                0x10000;
                         }
                         else {
                             /* invalid second surrogate */
@@ -402,10 +401,11 @@ out:
     free(lex->value.string);
 }
 
-static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
+static int lex_scan_number(lex_t *lex, char c, json_error_t *error)
 {
     const char *saved_text;
     char *end;
+    double value;
 
     lex->token = TOKEN_INVALID;
 
@@ -419,21 +419,37 @@ static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
             goto out;
         }
     }
-    else /* c != '0' */ {
+    else if(isdigit(c)) {
         c = lex_get_save(lex, error);
         while(isdigit(c))
             c = lex_get_save(lex, error);
     }
+    else {
+      lex_unget_unsave(lex, c);
+      goto out;
+    }
 
     if(c != '.' && c != 'E' && c != 'e') {
+        long value;
+
         lex_unget_unsave(lex, c);
-        lex->token = TOKEN_INTEGER;
 
         saved_text = strbuffer_value(&lex->saved_text);
-        lex->value.integer = strtol(saved_text, &end, 10);
+        value = strtol(saved_text, &end, 10);
         assert(end == saved_text + lex->saved_text.length);
 
-        return;
+        if((value == LONG_MAX && errno == ERANGE) || value > INT_MAX) {
+            error_set(error, lex, "too big integer");
+            goto out;
+        }
+        else if((value == LONG_MIN && errno == ERANGE) || value < INT_MIN) {
+            error_set(error, lex, "too big negative integer");
+            goto out;
+        }
+
+        lex->token = TOKEN_INTEGER;
+        lex->value.integer = (int)value;
+        return 0;
     }
 
     if(c == '.') {
@@ -463,14 +479,29 @@ static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
     }
 
     lex_unget_unsave(lex, c);
-    lex->token = TOKEN_REAL;
 
     saved_text = strbuffer_value(&lex->saved_text);
-    lex->value.real = strtod(saved_text, &end);
+    value = strtod(saved_text, &end);
     assert(end == saved_text + lex->saved_text.length);
 
+    if(value == 0 && errno == ERANGE) {
+        error_set(error, lex, "real number underflow");
+        goto out;
+    }
+
+    /* Cannot test for +/-HUGE_VAL because the HUGE_VAL constant is
+       only defined in C99 mode. So let's trust in sole errno. */
+    else if(errno == ERANGE) {
+        error_set(error, lex, "real number overflow");
+        goto out;
+    }
+
+    lex->token = TOKEN_REAL;
+    lex->value.real = value;
+    return 0;
+
 out:
-    return;
+    return -1;
 }
 
 static int lex_scan(lex_t *lex, json_error_t *error)
@@ -480,8 +511,8 @@ static int lex_scan(lex_t *lex, json_error_t *error)
     strbuffer_clear(&lex->saved_text);
 
     if(lex->token == TOKEN_STRING) {
-      free(lex->value.string);
-      lex->value.string = NULL;
+        free(lex->value.string);
+        lex->value.string = NULL;
     }
 
     c = lex_get(lex, error);
@@ -493,7 +524,7 @@ static int lex_scan(lex_t *lex, json_error_t *error)
         c = lex_get(lex, error);
     }
 
-    if(c == EOF) {
+    if(c == (char)EOF) {
         if(lex_eof(lex))
             lex->token = TOKEN_EOF;
         else
@@ -509,8 +540,10 @@ static int lex_scan(lex_t *lex, json_error_t *error)
     else if(c == '"')
         lex_scan_string(lex, error);
 
-    else if(isdigit(c) || c == '-')
-        lex_scan_number(lex, c, error);
+    else if(isdigit(c) || c == '-') {
+        if(lex_scan_number(lex, c, error))
+            goto out;
+    }
 
     else if(isupper(c) || islower(c)) {
         /* eat up the whole identifier for clearer error messages */
@@ -717,7 +750,7 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
             break;
 
         case '{':
-          json = parse_object(lex, error);
+            json = parse_object(lex, error);
             break;
 
         case '[':
@@ -739,7 +772,7 @@ static json_t *parse_value(lex_t *lex, json_error_t *error)
     return json;
 }
 
-json_t *parse_json(lex_t *lex, json_error_t *error)
+static json_t *parse_json(lex_t *lex, json_error_t *error)
 {
     error_init(error);
 
@@ -836,6 +869,8 @@ json_t *json_load_file(const char *path, json_error_t *error)
     json_t *result;
     FILE *fp;
 
+    error_init(error);
+
     fp = fopen(path, "r");
     if(!fp)
     {