summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--src/pugixml.cpp51
-rw-r--r--tests/test_parse.cpp48
2 files changed, 79 insertions, 20 deletions
diff --git a/src/pugixml.cpp b/src/pugixml.cpp
index 242202e..68a68d0 100644
--- a/src/pugixml.cpp
+++ b/src/pugixml.cpp
@@ -1374,37 +1374,48 @@ namespace
{
unsigned int ucsc = 0;
- ++stre;
-
- if (*stre == 'x') // &#x... (hex code)
+ if (stre[1] == 'x') // &#x... (hex code)
{
- ++stre;
-
- while (*stre)
+ stre += 2;
+
+ char_t ch = *stre;
+
+ if (ch == ';') return stre;
+
+ for (;;)
{
- if (*stre >= '0' && *stre <= '9')
- ucsc = 16 * ucsc + (*stre++ - '0');
- else if (*stre >= 'A' && *stre <= 'F')
- ucsc = 16 * ucsc + (*stre++ - 'A' + 10);
- else if (*stre >= 'a' && *stre <= 'f')
- ucsc = 16 * ucsc + (*stre++ - 'a' + 10);
- else if (*stre == ';')
+ if (static_cast<unsigned int>(ch - '0') <= 9)
+ ucsc = 16 * ucsc + (ch - '0');
+ else if (static_cast<unsigned int>((ch | ' ') - 'a') <= 5)
+ ucsc = 16 * ucsc + ((ch | ' ') - 'a' + 10);
+ else if (ch == ';')
break;
else // cancel
return stre;
- }
- if (*stre != ';') return stre;
-
+ ch = *++stre;
+ }
+
++stre;
}
else // &#... (dec code)
{
- while (*stre >= '0' && *stre <= '9')
- ucsc = 10 * ucsc + (*stre++ - '0');
+ char_t ch = *++stre;
- if (*stre != ';') return stre;
-
+ if (ch == ';') return stre;
+
+ for (;;)
+ {
+ if (static_cast<unsigned int>(ch - '0') <= 9)
+ ucsc = 10 * ucsc + (ch - '0');
+ else if (ch == ';')
+ break;
+ else // cancel
+ return stre;
+
+ ch = *++stre;
+ }
+
++stre;
}
diff --git a/tests/test_parse.cpp b/tests/test_parse.cpp
index c2f56e5..50f8867 100644
--- a/tests/test_parse.cpp
+++ b/tests/test_parse.cpp
@@ -281,6 +281,47 @@ TEST(parse_escapes_code)
CHECK_STRING(doc.child_value(STR("node")), STR("\01 "));
}
+TEST(parse_escapes_code_exhaustive_dec)
+{
+ xml_document doc;
+ CHECK(doc.load(STR("<node>&#/;&#01;&#2;&#3;&#4;&#5;&#6;&#7;&#8;&#9;&#:;&#a;&#A;&#XA;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&#/;\x1\x2\x3\x4\x5\x6\x7\x8\x9&#:;&#a;&#A;&#XA;"));
+}
+
+TEST(parse_escapes_code_exhaustive_hex)
+{
+ xml_document doc;
+ CHECK(doc.load(STR("<node>&#x/;&#x01;&#x2;&#x3;&#x4;&#x5;&#x6;&#x7;&#x8;&#x9;&#x:;&#x@;&#xA;&#xB;&#xC;&#xD;&#xE;&#xF;&#xG;&#x`;&#xa;&#xb;&#xc;&#xd;&#xe;&#xf;&#xg;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&#x/;\x1\x2\x3\x4\x5\x6\x7\x8\x9&#x:;&#x@;\xa\xb\xc\xd\xe\xf&#xG;&#x`;\xa\xb\xc\xd\xe\xf&#xg;"));
+}
+
+TEST(parse_escapes_code_restore)
+{
+ xml_document doc;
+ CHECK(doc.load(STR("<node>&#1&#32;&#x1&#32;&#1-&#32;&#x1-&#32;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&#1 &#x1 &#1- &#x1- "));
+}
+
+TEST(parse_escapes_char_restore)
+{
+ xml_document doc;
+
+ CHECK(doc.load(STR("<node>&q&#32;&qu&#32;&quo&#32;&quot&#32;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&q &qu &quo &quot "));
+
+ CHECK(doc.load(STR("<node>&a&#32;&ap&#32;&apo&#32;&apos&#32;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&a &ap &apo &apos "));
+
+ CHECK(doc.load(STR("<node>&a&#32;&am&#32;&amp&#32;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&a &am &amp "));
+
+ CHECK(doc.load(STR("<node>&l&#32;&lt&#32;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&l &lt "));
+
+ CHECK(doc.load(STR("<node>&g&#32;&gt&#32;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&g &gt "));
+}
+
TEST(parse_escapes_unicode)
{
xml_document doc;
@@ -314,6 +355,13 @@ TEST(parse_escapes_error)
CHECK(!doc.load(STR("<node id='&apos")));
}
+TEST(parse_escapes_code_invalid)
+{
+ xml_document doc;
+ CHECK(doc.load(STR("<node>&#;&#x;&;&#x-;&#-;</node>"), parse_minimal | parse_escapes));
+ CHECK_STRING(doc.child_value(STR("node")), STR("&#;&#x;&;&#x-;&#-;"));
+}
+
TEST(parse_attribute_spaces)
{
xml_document doc;