New issue from Arthur O'Dwyer: "bitset::reference should be const-assignable"

Author: Dani-Hub

xml/issue4187.xml

@@ -0,0 +1,89 @@
+<?xml version='1.0' encoding='utf-8' standalone='no'?>
+<!DOCTYPE issue SYSTEM "lwg-issue.dtd">
+
+<issue num="4187" status="New">
+<title>`bitset::reference` should be const-assignable</title>
+<section><sref ref="[template.bitset]"/></section>
+<submitter>Arthur O'Dwyer</submitter>
+<date>21 Dec 2024</date>
+<priority>99</priority>
+
+<discussion>
+<p>
+LWG <iref ref="3638"/>, which proposes changes to <tt>vector&lt;bool&gt;::reference</tt>, is related. 
+Should <tt>vector&lt;bool&gt;::reference</tt> and <tt>bitset&lt;N&gt;::reference</tt> behave differently 
+in any respect? I think there's no good reason for them to behave differently, and good technical 
+incentives to permit them to behave the same. We already have implementation divergence: libc++ makes 
+`bitset::reference` const-assignable, whereas libstdc++ and MS STL do not. This means that libc++'s 
+`bitset::reference` successfully avoids false positives from Arthur's proposed -Wassign-to-class-rvalue 
+diagnostic, while libstdc++'s does not (See <a href="https://godbolt.org/z/4e96j7rvv">Godbolt</a>).
+<p/>
+The proposed resolution applies <a href="https://wg21.link/P2321#vectorboolreference">P2321</a>'s approach. We 
+can't just insert `const` into the existing spec, because ABI. But also, since our goal is consistency 
+with the post-P2321 <tt>vector&lt;bool&gt;::reference</tt>, we should do the same thing here as P2321, not invent anything novel.
+<p/>
+Open questions related to the current P/R:
+</p>
+<ol>
+<li><p>LWG 3638 proposes to add these three `swap` overloads to <tt>vector&lt;bool&gt;::reference</tt>. 
+Should we also, consistently, add them to `bitset::reference`? I think we should.</p>
+<blockquote><pre>
+friend constexpr void swap(reference x, reference y) noexcept;
+friend constexpr void swap(reference x, bool&amp; y) noexcept;
+friend constexpr void swap(bool&amp; x, reference y) noexcept;
+</pre></blockquote>
+</li>
+<li><p>Both <tt>vector&lt;bool&gt;::reference</tt> and `bitset::reference` right now are specified with</p>
+<blockquote><pre>
+constexpr reference(const reference&amp;) = default;
+</pre></blockquote>
+<p>which is meaningless because we don't know the data members of `reference`. So this isn't actually 
+specifying that the constructor is trivial, let alone that it's `noexcept`. I think we should re-specify 
+both types' copy constructors as simply `constexpr` and `noexcept`; and then if we want them to be trivial, 
+we should say so in English prose.</p>
+</li>
+</ol>
+</discussion>
+
+<resolution>
+<p>
+This wording is relative to <paper num="N5001"/>.
+</p>
+
+<ol>
+<li><p>Modify <sref ref="[template.bitset.general]"/> as indicated:</p>
+
+<blockquote>
+<pre>
+namespace std {
+  template&lt;size_t N&gt; class bitset {
+  public:
+    // <i>bit reference</i>
+    class reference {
+    public:
+      constexpr reference(const reference&amp;) = default;
+      constexpr ~reference();
+      constexpr reference&amp; operator=(bool x) noexcept;           // <i>for b[i] = x;</i>
+      constexpr reference&amp; operator=(const reference&amp;) noexcept; // <i>for b[i] = b[j];</i>
+      <ins>constexpr const reference&amp; operator=(bool x) const noexcept;</ins>
+      constexpr bool operator~() const noexcept;                 // <i>flips the bit</i>
+      constexpr operator bool() const noexcept;                  // <i>for x = b[i];</i>
+      constexpr reference&amp; flip() noexcept;                      // <i>for b[i].flip();</i>
+      <ins>friend constexpr void swap(reference x, reference y) noexcept;
+      friend constexpr void swap(reference x, bool&amp; y) noexcept;
+      friend constexpr void swap(bool&amp; x, reference y) noexcept;</ins>      
+    };
+    [&hellip;]
+  };
+  [&hellip;]
+}
+</pre>
+</blockquote>
+
+
+</li>
+
+</ol>
+</resolution>
+
+</issue>