Index: net/base/net_util.cc |
=================================================================== |
--- net/base/net_util.cc (revision 223643) |
+++ net/base/net_util.cc (working copy) |
@@ -324,7 +324,7 @@ |
UErrorCode status = U_ZERO_ERROR; |
#ifdef U_WCHAR_IS_UTF16 |
icu::UnicodeSet dangerous_characters(icu::UnicodeString( |
- L"[[\\ \u00bc\u00bd\u01c3\u0337\u0338" |
+ L"[[\\ \u00ad\u00bc\u00bd\u01c3\u0337\u0338" |
L"\u05c3\u05f4\u06d4\u0702\u115f\u1160][\u2000-\u200b]" |
L"[\u2024\u2027\u2028\u2029\u2039\u203a\u2044\u205f]" |
L"[\u2154-\u2156][\u2159-\u215b][\u215f\u2215\u23ae" |
@@ -341,7 +341,7 @@ |
0, status); |
#else |
icu::UnicodeSet dangerous_characters(icu::UnicodeString( |
- "[[\\u0020\\u00bc\\u00bd\\u01c3\\u0337\\u0338" |
+ "[[\\u0020\\u00ad\\u00bc\\u00bd\\u01c3\\u0337\\u0338" |
"\\u05c3\\u05f4\\u06d4\\u0702\\u115f\\u1160][\\u2000-\\u200b]" |
"[\\u2024\\u2027\\u2028\\u2029\\u2039\\u203a\\u2044\\u205f]" |
"[\\u2154-\\u2156][\\u2159-\\u215b][\\u215f\\u2215\\u23ae" |
@@ -398,6 +398,29 @@ |
return false; |
} |
+struct uidna_wrapper { |
Peter Kasting
2013/09/19 20:55:36
Struct names should be CamelCase.
jungshik at Google
2013/09/20 21:33:41
Done.
|
+ uidna_wrapper() { |
+ UErrorCode err = U_ZERO_ERROR; |
+ // This is the option closest to what we had in the past with IDNA 2003 |
Peter Kasting
2013/09/19 20:55:36
Never write in a comment about what "used to" happ
jungshik at Google
2013/09/20 21:33:41
Done. I just kept TODO comment here and explained
|
+ // API and matches what IE 10 does except for BiDi check. |
+ // IDNA 2003 always checks BiDi. We used to allow unassigned code |
+ // points. However, with our Unicode DB pretty up to date, we'd not |
+ // need to turn this on. |
+ // We didn't use STD3 rules and we continue not to. |
+ // TODO(jungshik) : Change options as different parties (browsers, |
+ // registrars, search engines) converge toward a consensus. |
+ int32_t options = UIDNA_CHECK_BIDI; |
Peter Kasting
2013/09/19 20:55:36
Nit: Don't make a temp for this, just inline it.
jungshik at Google
2013/09/20 21:33:41
Done.
|
+ value = uidna_openUTS46(options, &err); |
+ if (U_FAILURE(err)) |
+ value = NULL; |
+ } |
+ |
+ UIDNA* value; |
+}; |
+ |
+static base::LazyInstance<uidna_wrapper>::Leaky |
+ g_uidna = LAZY_INSTANCE_INITIALIZER; |
+ |
// Converts one component of a host (between dots) to IDN if safe. The result |
// will be APPENDED to the given output string and will be the same as the input |
// if it is not IDN or the IDN is unsafe to display. Returns whether any |
@@ -414,26 +437,26 @@ |
static const base::char16 kIdnPrefix[] = {'x', 'n', '-', '-'}; |
if ((comp_len > arraysize(kIdnPrefix)) && |
!memcmp(comp, kIdnPrefix, arraysize(kIdnPrefix) * sizeof(base::char16))) { |
- // Repeatedly expand the output string until it's big enough. It looks like |
- // ICU will return the required size of the buffer, but that's not |
- // documented, so we'll just grow by 2x. This should be rare and is not on a |
- // critical path. |
+ |
+ UIDNA* uidna = g_uidna.Get().value; |
+ DCHECK(uidna != NULL); |
size_t original_length = out->length(); |
- for (int extra_space = 64; ; extra_space *= 2) { |
+ for (int output_length = 64; ; ) { |
Peter Kasting
2013/09/19 20:55:36
How about writing the loop this way:
// Try t
jungshik at Google
2013/09/20 21:33:41
Done with a slight change because UErrorCode has t
|
UErrorCode status = U_ZERO_ERROR; |
- out->resize(out->length() + extra_space); |
- int output_chars = uidna_IDNToUnicode(comp, |
- static_cast<int32_t>(comp_len), &(*out)[original_length], extra_space, |
- UIDNA_DEFAULT, NULL, &status); |
- if (status == U_ZERO_ERROR) { |
+ UIDNAInfo info = UIDNA_INFO_INITIALIZER; |
+ out->resize(original_length + output_length); |
+ output_length = uidna_labelToUnicode(uidna, comp, |
+ static_cast<int32_t>(comp_len), &(*out)[original_length], |
+ output_length, &info, &status); |
+ if (U_SUCCESS(status) && info.errors == 0) { |
// Converted successfully. |
- out->resize(original_length + output_chars); |
- if (IsIDNComponentSafe(out->data() + original_length, output_chars, |
+ out->resize(original_length + output_length); |
+ if (IsIDNComponentSafe(out->data() + original_length, output_length, |
languages)) |
return true; |
} |
- if (status != U_BUFFER_OVERFLOW_ERROR) |
+ if (status != U_BUFFER_OVERFLOW_ERROR || info.errors != 0) |
break; |
} |
// Failed, revert back to original string. |