src: support domains with empty labels

Follow the spec of domainToASCII/domainToUnicode in whatwg, and
synchronise WPT url test data.

Refs: web-platform-tests/wpt#5397
PR-URL: #12707
Reviewed-By: James M Snell <[email protected]>
Reviewed-By: Timothy Gu <[email protected]>

Author: watilde

src/node_i18n.cc

@@ -461,6 +461,13 @@ int32_t ToUnicode(MaybeStackBuffer<char>* buf,
                                   &status);
   }
 
+  // UTS #46's ToUnicode operation applies no validation of domain name length
+  // (nor a flag requesting it to do so, like VerifyDnsLength for ToASCII). For
+  // that reason, unlike ToASCII below, ICU4C correctly accepts long domain
+  // names. However, ICU4C still sets the EMPTY_LABEL error in contrary to UTS
+  // #46. Therefore, explicitly filters out that error here.
+  info.errors &= ~UIDNA_ERROR_EMPTY_LABEL;
+
   if (U_FAILURE(status) || (!lenient && info.errors != 0)) {
     len = -1;
     buf->SetLength(0);
@@ -500,6 +507,16 @@ int32_t ToASCII(MaybeStackBuffer<char>* buf,
                                  &status);
   }
 
+  // The WHATWG URL "domain to ASCII" algorithm explicitly sets the
+  // VerifyDnsLength flag to false, which disables the domain name length
+  // verification step in ToASCII (as specified by UTS #46). Unfortunately,
+  // ICU4C's IDNA module does not support disabling this flag through `options`,
+  // so just filter out the errors that may be caused by the verification step
+  // afterwards.
+  info.errors &= ~UIDNA_ERROR_EMPTY_LABEL;
+  info.errors &= ~UIDNA_ERROR_LABEL_TOO_LONG;
+  info.errors &= ~UIDNA_ERROR_DOMAIN_NAME_TOO_LONG;
+
   if (U_FAILURE(status) || (!lenient && info.errors != 0)) {
     len = -1;
     buf->SetLength(0);

test/fixtures/url-idna.js

@@ -182,23 +182,18 @@ module.exports = {
       ascii: 'xn--vitnam-jk8b.icom.museum',
       unicode: 'việtnam.icom.museum'
     },
-    // long URL
-    {
-      ascii: `${`${'a'.repeat(63)}.`.repeat(3)}com`,
-      unicode: `${`${'a'.repeat(63)}.`.repeat(3)}com`
-    }
-  ],
-  invalid: [
     // long label
     {
-      url: `${'a'.repeat(64)}.com`,
-      mode: 'ascii'
+      ascii: `${'a'.repeat(64)}.com`,
+      unicode: `${'a'.repeat(64)}.com`,
     },
     // long URL
     {
-      url: `${`${'a'.repeat(63)}.`.repeat(4)}com`,
-      mode: 'ascii'
-    },
+      ascii: `${`${'a'.repeat(64)}.`.repeat(4)}com`,
+      unicode: `${`${'a'.repeat(64)}.`.repeat(4)}com`
+    }
+  ],
+  invalid: [
     // invalid character
     {
       url: '\ufffd.com',

test/fixtures/url-tests.js

@@ -1,7 +1,7 @@
 'use strict';
 
 /* WPT Refs:
-   https://github.com/w3c/web-platform-tests/blob/3eff1bd/url/urltestdata.json
+   https://github.com/w3c/web-platform-tests/blob/3afae94/url/urltestdata.json
    License: http://www.w3.org/Consortium/Legal/2008/04-testsuite-copyright.html
 */
 module.exports =
@@ -3789,6 +3789,52 @@ module.exports =
     "search": "",
     "hash": ""
   },
+  "Domains with empty labels",
+  {
+    "input": "http://./",
+    "base": "about:blank",
+    "href": "http://./",
+    "origin": "http://.",
+    "protocol": "http:",
+    "username": "",
+    "password": "",
+    "host": ".",
+    "hostname": ".",
+    "port": "",
+    "pathname": "/",
+    "search": "",
+    "hash": ""
+  },
+  {
+    "input": "http://../",
+    "base": "about:blank",
+    "href": "http://../",
+    "origin": "http://..",
+    "protocol": "http:",
+    "username": "",
+    "password": "",
+    "host": "..",
+    "hostname": "..",
+    "port": "",
+    "pathname": "/",
+    "search": "",
+    "hash": ""
+  },
+  {
+    "input": "http://0..0x300/",
+    "base": "about:blank",
+    "href": "http://0..0x300/",
+    "origin": "http://0..0x300",
+    "protocol": "http:",
+    "username": "",
+    "password": "",
+    "host": "0..0x300",
+    "hostname": "0..0x300",
+    "port": "",
+    "pathname": "/",
+    "search": "",
+    "hash": ""
+  },
   "Broken IPv6",
   {
     "input": "http://[www.google.com]/",