Allow document.open() to desync session history and document URLs

Closes #6556. In particular, reverts document.open() to only update the document's URL, and not the session history entry's URL, like it did before ae7cf0c. Now that they can mismatch, we need to audit the cases where this might be important, which leads to the following changes:

* Changes location.reload() to reload the current session history entry's URL, instead of the document's URL. This ensures that post-document.open() reload behavior is aligned with WebKit and Gecko, as tested by https://wpt.fyi/results/html/webappapis/dynamic-markup-insertion/opening-the-input-stream/reload.window.html.

* Changes history.pushState()/history.replaceState() with no URL argument to default to the document's URL, instead of the current session history's URL. This ensures that post-document.open() pushState()/replaceState() behavior is aligned with all engines, as tested by web-platform-tests/wpt#28826.

This also modernizes and makes a bit more precise the location.reload() method steps. The user-initiated reload steps remain vague; #6600 will tackle those.

Author: domenic

source

@@ -83619,7 +83619,16 @@ interface <dfn>BarProp</dfn> {
   data-x="struct item">items</span>:</p>
 
   <ul>
-   <li><p><dfn data-x="she-url">URL</dfn>, a <span>URL</span></p></li>
+   <li>
+    <p><dfn data-x="she-url">URL</dfn>, a <span>URL</span></p>
+
+    <p class="warning">This <a href="#warning-document-open-url-mismatch">doesn't necessarily
+    match</a> the <span data-x="she-document">document</span>'s <span
+    data-x="concept-document-URL">URL</span>. In general, to find the "current URL" prefer using the
+    appropriate <code>Document</code>'s <span data-x="concept-document-URL">URL</span>; a
+    <span>session history entry</span>'s <span data-x="she-url">URL</span> is only to be used for
+    session history traversal and restoration.</p>
+   </li>
 
    <li>
     <p><dfn data-x="she-document">document</dfn>, a <code>Document</code> or null</p>
@@ -84178,8 +84187,8 @@ interface <dfn>History</dfn> {
    <li><p>Let <var>serializedData</var> be
    <span>StructuredSerializeForStorage</span>(<var>data</var>). Rethrow any exceptions.</p></li>
 
-   <li><p>Let <var>newURL</var> be the <span>session history</span>'s <span>current entry</span>'s
-   <span data-x="she-url">URL</span>.</p></li>
+   <li><p>Let <var>newURL</var> be <var>document</var>'s <span
+   data-x="concept-document-URL">URL</span>.</p></li>
 
    <li>
     <p>If <var>url</var> is not null, then:</p>
@@ -85086,52 +85095,56 @@ interface <dfn>Location</dfn> { // but see also <a href="#the-location-interface
   <p class="note">The <code data-x="dom-location-replace">replace()</code> method intentionally has
   no security check.</p>
 
-  <p>When the <dfn method for="Location"><code data-x="dom-location-reload">reload()</code></dfn>
-  method is invoked, the user agent must run the appropriate steps from the following list:</p>
-
-  <dl class="switch">
-   <dt>If this <code>Location</code> object's <span>relevant <code>Document</code></span> is
-   null</dt>
-
-   <dd><p>Return.</p></dd>
+  <p>The <dfn method for="Location"><code data-x="dom-location-reload">reload()</code></dfn> method
+  steps are:</p>
 
-   <dt>If this <code>Location</code> object's <span>relevant <code>Document</code></span>'s <span
-   data-x="concept-document-origin">origin</span> is not <span>same origin-domain</span> with the
-   <span>entry settings object</span>'s <span
-   data-x="concept-settings-object-origin">origin</span></dt>
+  <ol>
+   <li><p>Let <var>document</var> be <span>this</span>'s <span>relevant
+   <code>Document</code></span>.</p></li>
 
-   <dd><p>Throw a <span>"<code>SecurityError</code>"</span> <code>DOMException</code>.</p></dd>
+   <li><p>Let <var>browsingContext</var> be <var>document</var>'s <span
+   data-x="concept-document-bc">browsing context</span>.</p></li>
 
-   <dt>If the currently executing <span data-x="concept-task">task</span> is the dispatch of a <code
-   data-x="event-resize">resize</code> event in response to the user resizing the <span>browsing
-   context</span></dt>
+   <li><p>If <var>document</var> is null, then return.</p></li>
 
-   <dd><p>Repaint the <span>browsing context</span> and return.</p></dd> <!-- this
-   theoretically would have no effect but in practice can be useful to work around rendering bugs.
-   -->
+   <li><p>If <var>document</var>'s <span data-x="concept-document-origin">origin</span> is not
+   <span>same origin-domain</span> with the <span>entry settings object</span>'s <span
+   data-x="concept-settings-object-origin">origin</span>, then throw a
+   <span>"<code>SecurityError</code>"</span> <code>DOMException</code>.</p></li>
 
-   <dt>If the <span>browsing context</span>'s <span>active document</span> is <span>an
-   <code>iframe</code> <code data-x="attr-iframe-srcdoc">srcdoc</code> document</span></dt>
+   <li><p>If <span>this</span>'s <span>relevant agent</span>'s <span
+   data-x="concept-agent-event-loop">event loop</span>'s <span>currently running task</span> is the
+   dispatch of a <code data-x="event-resize">resize</code> event in response to the user resizing
+   the <span>browsing context</span>, then repaint the <span>browsing context</span> and
+   return.</p></li>
+   <!-- this theoretically would have no effect but in practice can be useful to work around
+   rendering bugs. -->
 
-   <dd><p><span data-x="process the iframe attributes">Reprocess the <code>iframe</code>
-   attributes</span> of the <span>browsing context</span>'s <span
-   data-x="bc-container">container</span>.</p></dd>
+   <li><p>If <var>document</var> is <span>an <code>iframe</code> <code
+   data-x="attr-iframe-srcdoc">srcdoc</code> document</span>, then <span data-x="process the iframe
+   attributes">reprocess the <code>iframe</code> attributes</span> of <var>browsingContext</var>'s
+   <span data-x="bc-container">container</span> and return.</p></li>
 
-   <dt>Otherwise</dt>
+   <li>
+    <p><span>Navigate</span><!--DONAV location.reload()--> <var>browsingContext</var> to
+    <var>browsingContext</var>'s <span>session history</span>'s <span>current entry</span>'s <span
+    data-x="she-url">URL</span>, with <var><span>exceptionsEnabled</span></var> set to true, <var
+    data-x="navigation-hh">historyHandling</var> set to "<code data-x="hh-reload">reload</code>",
+    and the <span>source browsing context</span> set to <var>browsingContext</var>.</p>
 
-   <dd><p><span>Navigate</span><!--DONAV location.reload()--> the <span>browsing context</span> to
-   this <code>Location</code> object's <span>relevant <code>Document</code></span>'s <span
-   data-x="concept-document-url">URL</span>, with <var><span>exceptionsEnabled</span></var> set to
-   true, <var data-x="navigation-hh">historyHandling</var> set to "<code
-   data-x="hh-reload">reload</code>", and the <span>source browsing context</span> set to the
-   <span>browsing context</span> being navigated.</p></dd>
+    <p class="note">This is different than navigating to <var>document</var>'s <span
+    data-x="concept-document-url">URL</span>, due to the potential for <a
+    href="#warning-document-open-url-mismatch">mismatches caused by <code
+    data-x="dom-document-open">document.open()</code></a>.</p>
 
-   <!-- It appears that document.reload() always uses GET and does not, e.g., re-POST. Thus the
-   difference between using the document's URL here, and "the same resource as that Document" below
-   in the user-triggered reload section. -->
+    <!-- It appears that document.reload() always uses GET and does not, e.g., re-POST. Thus the
+    difference between using the document's URL here, and "the same resource as that Document" below
+    in the user-triggered reload section. -->
 
-   <!-- https://software.hixie.ch/utilities/js/live-dom-viewer/?saved=141 defends the source browsing context claim; submit, eval document.referrer, reload, eval document.referrer. -->
-  </dl>
+    <!-- https://software.hixie.ch/utilities/js/live-dom-viewer/?saved=141 defends the source
+    browsing context claim; submit, eval document.referrer, reload, eval document.referrer. -->
+   </li>
+  </ol>
 
   <p>When a user requests that the <span>active document</span> of a <span>browsing context</span>
   be reloaded through a user interface element, the user agent should <span>navigate</span><!--DONAV
@@ -94075,6 +94088,28 @@ document.body.appendChild(frame)</code></pre>
      <li><p>If <var>entryDocument</var> is not <var>document</var>, then set <var>newURL</var>'s
      <span data-x="concept-url-fragment">fragment</span> to null.</p></li>
 
+     <li>
+      <p>Set <var>document</var>'s <span data-x="concept-document-url">URL</span> to
+      <var>newURL</var>.</p>
+
+      <div class="note">
+       <p>The main purpose of this step is to allow code such as the following:</p>
+
+       <pre><code class="js">const w = window.open();
+w.document.open();
+w.document.write("&lt;img src='image.png'>");</code></pre>
+
+       <p>to resolve the <code data-x="">image.png</code> URL relative to the caller, instead of
+       relative to <code>about:blank</code>.</p>
+      </div>
+
+      <p class="warning" id="warning-document-open-url-mismatch">This can cause
+      <var>document</var>'s <span data-x="concept-document-url">URL</span> to mismatch its
+      corresponding <span>session history entry</span>'s <span data-x="she-url">URL</span>. This
+      mismatch is required for web compatibility; see <a
+      href="https://github.com/whatwg/html/issues/6556">issue #6556</a>.</p>
+     </li>
+
      <li><p>Run the <span>URL and history update steps</span> with <var>document</var> and
      <var>newURL</var>.</p></li>
     </ol>