Correctly handle problematic nested escapes in URL paths.

url/url_canon_unittest.cc

Index: url/url_canon_unittest.cc
diff --git a/url/url_canon_unittest.cc b/url/url_canon_unittest.cc
index 0ccd6c93009efa94faf0be1a3a95a2b169e96158..4dd58ec06d581cbbdadf064a0bcd09ec04740d2e 100644
--- a/url/url_canon_unittest.cc
+++ b/url/url_canon_unittest.cc
@@ -1000,6 +1000,7 @@ TEST(URLCanonTest, Port) {
 
 TEST(URLCanonTest, Path) {
   DualComponentCase path_cases[] = {
+#if 0

[brettw, 2015/09/21 22:28:34]

Left by mistake?

[Peter Kasting, 2015/09/21 23:15:31]

Yeah, oops :(

     // ----- path collapsing tests -----
     {"/././foo", L"/././foo", "/foo", Component(0, 4), true},
     {"/./.foo", L"/./.foo", "/.foo", Component(0, 5), true},
@@ -1060,6 +1061,24 @@ TEST(URLCanonTest, Path) {
     {"/%7Ffp3%3Eju%3Dduvgw%3Dd", L"/%7Ffp3%3Eju%3Dduvgw%3Dd", "/%7Ffp3%3Eju%3Dduvgw%3Dd", Component(0, 24), true},
       // @ should be passed through unchanged (escaped or unescaped).
     {"/@asdf%40", L"/@asdf%40", "/@asdf%40", Component(0, 9), true},
+      // Nested escape sequences should result in escaping the leading '%' if
+      // unescaping would result in a new escape sequence.
+    {"/%0%30", L"/%0%30", "/%2500", Component(0, 6), true},

[brettw, 2015/09/21 22:28:34]

Can you make these cases use different characters for each possible character
rather than "%30" and "0", to prove we're not doing something silly duplicating
one of them to get the output?

[Peter Kasting, 2015/09/21 23:15:30]

Done.

+    {"/%%300", L"/%%300", "/%2500", Component(0, 6), true},
+    {"/%%30%30", L"/%%30%30", "/%2500", Component(0, 6), true},
+      // Make sure truncated "nested" escapes don't result in reading off the
+      // string end.
+    {"/%%30", L"/%%30", "/%0", Component(0, 3), true},
+      // Don't unescape the leading '%' if unescaping doesn't result in a valid
+      // new escape sequence.
+    {"/%%470", L"/%%470", "/%G0", Component(0, 4), true},
+    {"/%%2D%41", L"/%%2D%41", "/%-A", Component(0, 4), true},
+#endif
+      // Don't erroneously downcast a UTF-16 charater in a way that makes it
+      // look like part of an escape sequence.
+    {NULL, L"/%%30\x0130", "/%0%C4%B0", Component(0, 9), true},
+#if 0
+
 
     // ----- encoding tests -----
       // Basic conversions
@@ -1068,6 +1087,7 @@ TEST(URLCanonTest, Path) {
       // UTF-16 input, so this doesn't happen on 8-bit.
     {"/\xef\xb7\x90zyx", NULL, "/%EF%B7%90zyx", Component(0, 13), true},
     {NULL, L"/\xfdd0zyx", "/%EF%BF%BDzyx", Component(0, 13), false},
+#endif
   };
 
   for (size_t i = 0; i < arraysize(path_cases); i++) {