<html>

<head>
<meta name="GENERATOR" content="Microsoft FrontPage 5.0">
<meta name="ProgId" content="FrontPage.Editor.Document">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<style>
 ins  {background-color: #CCFFCC; text-decoration: none;}
 del  {background-color: #FFCACA; text-decoration: none;}
 a    {text-decoration: none;}
</style>
<title>Filesystem NB Comments C++17</title>
</head>

<body>
<table>
<tr>
  <td>Doc. no.:</td>
  <td>P0492R1</td>
</tr>
<tr>
  <td>Date:</td>
  <td>
  <!--webbot bot="Timestamp" S-Type="EDITED" S-Format="%Y-%m-%d" startspan -->2017-02-06<!--webbot bot="Timestamp" endspan i-checksum="12154" --></td>
</tr>
<tr>
  <td>Reply to:</td>
  <td>Beman Dawes &lt;bdawes at acm dot org&gt;
</tr>
<tr>
  <td>&nbsp;</td>
  <td>S. Davis Herring &lt;herring at lanl dot gov&gt;</td>
</tr>
<tr>
  <td>&nbsp;</td>
  <td>Nicolai Josuttis &lt;nico at josuttis dot de&gt;</td>
</tr>
<tr>
  <td>&nbsp;</td>
  <td>Jason Liu &lt;jasonliu dot development at gmail dot com&gt;</td>
</tr>
<tr>
  <td>&nbsp;</td>
  <td>Billy O&#39;Neal &lt;bion at microsoft dot com&gt;</td>
</tr>
<tr>
  <td>&nbsp;</td>
  <td>P.J. Plauger &lt;pjp at dinkumware dot com&gt;</td>
</tr>
<tr>
  <td>&nbsp;</td>
  <td>Jonathan Wakely &lt;cxx at kayari dot org&gt;</td>
</tr>
<tr>
  <td>Audience:</td>
  <td>Library</td>
</tr>
</table>

<h1>Proposed Resolution of C++17  National Body Comments for Filesystems(R1)</h1>

<p>This document proposes resolutions of C++17 CD National Body Comments from
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0488r0.pdf">
P0488R0</a> related to Filesystems (27.10 [filesystems]). Some &quot;Late&quot; Comments from
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0489r0.pdf">
P0489R0</a> are also considered.</p>

<p>The proposed resolutions in this paper represent the consensus of the Library 
Working Group&#39;s Filesystem small group (SG) unless otherwise indicated.</p>

<p>Proposed wording  is 
relative to the C++ working paper,
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/n4618.pdf">
N4618</a>.</p>

<p><b><a href="#nb-comments">NB Technical and general comments</a></b><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US25">US 25: has_filename() is equivalent to just !empty()</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US31">US 31: Everything is defined in terms of one implicit host system</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US32">US 32: Meaning of 27.10.2.1 unclear</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US33">US 33: Definition of <i>canonical path</i> problematic</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US34">US 34: Are there attributes of a file that are not an aspect of the file system?</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US35">US 35: What synchronization is required to avoid a file system race?</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US36">US 36: Symbolic links themselves are attached to a directory via (hard) links</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US37">US 37: The term “redundant current directory (dot) elements” is not defined</a><br> 
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US40">US 40: Not all directories have a parent.</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US43">US 43: Concerns about <i>encoded character types</i></a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US44">US 44: Definition of path in terms of a string requires leaky abstraction</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US45">US 45: Generic format portability compromised by unspecified root-name</a><br> 
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US46">US 46: <i>filename</i> can be empty so productions for <i>relative-path</i> are redundant</a><br> 
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US48">US 48: Multiple separators are often meaningful in a <i>root-name</i></a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US49">US 49: What does “method of conversion method” mean?</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US51">US 51: Failing to add / when appending empty string prevents useful apps</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US52">US 52: <code>remove_filename()</code> postcondition is not by itself a definition</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US53">US 53: <code>remove_filename()</code>'s name does not correspond to its behavior</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US54">US 54: <code>remove_filename()</code> is broken</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US55">US 55: <code>replace_extension()</code>'s use of <code>path</code> as parameter is inappropriate</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US56">US 56: Remove <code>replace_extension()</code>&#39;s conditional addition of period</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US57">US 57: On Windows, absolute paths will sort in among relative paths</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US58">US 58: <code>parent_path()</code> behavior for root paths is useless</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US59">US 59: <code>filename()</code> returning <code>path</code> for single path components is bizarre</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US60">US 60: <code>path(&quot;/foo/&quot;).filename()==path(&quot;.&quot;)</code> is surprising</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US61">US 61: Leading dots in <code>filename()</code> should not begin an extension</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US62">US 62: It is important that <code>stem()+extension()==filename()</code></a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US63">US 63: <code>lexically_normal()</code> inconsistently treats trailing "/" but not &quot;/..&quot; as directory</a><br> 
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US185">US 185: Fold <code>error_code</code> and non-<code>error_code</code> signatures into one signature</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#FI14">FI 14: <tt>directory_entry</tt> comparisons are members</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US73">US 73, CA 2: <i>root-name</i> is effectively implementation defined</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US74">US 74, CA 3: The term “pathname” is ambiguous in some contexts</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US75">US 75, CA 4: Extra flag in path constructors is needed</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US76">US 76, CA 5: <i>root-name</i> definition is over-specified.</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US77">US 77, CA 6: <code>operator/</code> and other appends not useful if arg has <i>root-name</i></a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US78">US 78, CA 7: Member <code>absolute()</code> in 27.10.4.1 is overspecified for non-POSIX-like O/S</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US79">US 79, CA 8: Some  operation functions are overspecified for implementation-defined file types</a><br>
<b><a href="#editorial">NB comments submitted as editorial</a></b><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US38">US 38: Duplicates §17.3.16</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US39">US 39: Remove note: Dot and dot-dot are not directories</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US41">US 41: The term “parent directory” for a (non-directory) file is unusual</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US42">US 42: Pathname resolution does not always resolve a symlink</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US47">US 47: “.” and “..” already match the name production</a><br>
&nbsp;&nbsp;&nbsp;&nbsp;<a href="#US50">US 50: 27.10.8.1 ¶ 1.4 largely redundant with ¶ 1.3</a><br>
<b>
<a href="#late-comments">Late comments</a></b><br>
&nbsp;&nbsp;&nbsp; <span style="background-color: #FFFF99; font-style:italic">To be supplied</span>

<h2>Revision History</h2>

<h3>P0492R1 Pre-Kona mailing</h3>

<ul>
  <li>Direction remains unchanged.</li>
  <li>Updated WP base document to
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/n4618.pdf">
N4618</a>.</li>
  <li>Modifications to reflect editorial changes to the standard WP. Most 
  importantly, the name of the class <code>path</code> exposition only data 
  member was changed from <code>pathname</code> to <code>pathstring</code>.</li>
  <li>Proposed wording supplied for numerous NB comments that had no proposed 
  wording.</li>
  <li>Numerous minor and editorial corrections.</li>
</ul>

<h3>
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0492r0.html">
P0492R0</a> Post-Issaquah mailing</h3>

<blockquote>

<p>The initial version of this document.</p>

</blockquote>

<h2>Markup</h2>

<p><span style="font-style: italic; background-color: #FFFF99">Notes about work 
that remains to be done, who is planning to do that work, or wording that 
requires specially careful review by the LWG. If the note applies to proposed 
changes to the working paper, the changes may not yet ready to be voted into the 
C++17 working paper.</span><p><span style="font-style: italic; background-color: #E1E9FD">
SG informative notes that will remain in this proposal but do not become part of 
the WP.</span><p><i>Instructions to the project editor.</i><p><del>Text to be 
removed from the WP.</del><p><ins>Text 
to be added to the WP.</ins><h2><a name="nb-comments"></a>NB Technical and general comments</h2>

<h3>Comments related to <code>filename()</code></h3>

<p><span style="font-style: italic; background-color: #E1E9FD">These comments 
are tightly coupled and best resolved as a whole.</span></p>

<h4><a name="US25">US 25</a>; 27.10.8.4.10 [path.query]: <code>has_filename()</code> is equivalent to just 
<code>!empty()<br>
</code>Status: Reject</h4>
<blockquote>
<p>Resolved by <a href="#US52">US-52</a>, <a href="#US53">US-53</a>, <a href="#US54">
US-54</a>, and <a href="#US60">US-60</a>.</p>

</blockquote>

<h4><a name="US37">US 37</a>; 27.10.4.12 [fs.def.normal.form]: The term “redundant current directory (dot) elements” is not defined<br>
Status: Accept with modifications</h4>




<blockquote>

<p><b>Discussion</b></p>




<p>Invariants:</p>




  <ol>
    <li>An empty path remains empty, a non-empty path remains non-empty.</li>
    <li>An absolute path remains absolute, a relative path remains relative.</li>
    <li>The normal form has a trailing <i>directory-separator</i> if and only if 
    the original path had a trailing <i>directory-separator.</i></li>
    <li>No redundant <i>dot</i>, <i>dot-dot</i>, or <i>directory-separator</i>s.</li>
    <li><i>directory-separator</i>s are in the <i>preferred-separator</i> form.</li>
  </ol>

<p>Desired behavior:</p>




  <table border="1" cellpadding="5" cellspacing="0" style="border-collapse: collapse" bordercolor="#111111" height="1101">
    <tr>
      <td><b>Case #</b></td>
      <td><b>Generic<br>
      format<br>
      pathname</b></td>
      <td><b>After<br>
      normalization</b></td>
      <td>Nico&#39;s<br>
      4 Feb<br>
      email</td>
    </tr>
    <tr>
      <td>1</td>
      <td><code>&quot;&quot;</code></td>
      <td><code>&quot;&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>2</td>
      <td><code>&quot;.&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>3</td>
      <td><code>&quot;..&quot;</code></td>
      <td><code>&quot;..&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>4</td>
      <td><code>&quot;/&quot;</code></td>
      <td><code>&quot;/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>5</td>
      <td><code>&quot;//&quot;</code></td>
      <td><code>&quot;//&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>6</td>
      <td>
      <p dir="ltr"><code>&quot;/foo&quot;</code></td>
      <td><code>&quot;/foo&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>7</td>
      <td><code>&quot;/foo/&quot;</code></td>
      <td><code>&quot;/foo/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td bgcolor="#FFFF99">8</td>
      <td bgcolor="#FFFF99"><code>&quot;/foo/.&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/foo&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/foo/&quot;</code></td>
    </tr>
    <tr>
      <td bgcolor="#FFFF99">9</td>
      <td bgcolor="#FFFF99"><code>&quot;/foo/bar/..&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/foo&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/foo/&quot;</code></td>
    </tr>
    <tr>
      <td>10</td>
      <td><code>&quot;/foo/..&quot;</code></td>
      <td><code>&quot;/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>11</td>
      <td><code>&quot;/.&quot;</code></td>
      <td><code>&quot;/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td bgcolor="#FFFF99">12</td>
      <td bgcolor="#FFFF99"><code>&quot;/./&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/./&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/&quot;</code></td>
    </tr>
    <tr>
      <td>13</td>
      <td><code>&quot;/./.&quot;</code></td>
      <td><code>&quot;/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td bgcolor="#FFFF99">14</td>
      <td bgcolor="#FFFF99"><code>&quot;/././&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/./&quot;</code></td>
      <td bgcolor="#FFFF99"><code>&quot;/&quot;</code></td>
    </tr>
    <tr>
      <td>15</td>
      <td><code>&quot;/././.&quot;</code></td>
      <td><code>&quot;/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>16</td>
      <td><code>&quot;./&quot;</code></td>
      <td><code>&quot;./&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>17</td>
      <td><code>&quot;./.&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>18</td>
      <td><code>&quot;././&quot;</code></td>
      <td><code>&quot;./&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>19</td>
      <td><code>&quot;././.&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>20</td>
      <td><code>&quot;./././&quot;</code></td>
      <td><code>&quot;./&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>21</td>
      <td><code>&quot;./././.&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>22</td>
      <td><code>&quot;foo/..&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>23</td>
      <td><code>&quot;foo/../&quot;</code></td>
      <td><code>&quot;./&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>24</td>
      <td><code>&quot;foo/../..&quot;</code></td>
      <td><code>&quot;..&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>25 Windows</td>
      <td><code>&quot;C:bar/..&quot;</code></td>
      <td><code>&quot;C:.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>26 Windows</td>
      <td><code>&quot;C:&quot;</code></td>
      <td><code>&quot;C:&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>27</td>
      <td><code>&quot;//host/bar/..&quot;</code></td>
      <td><code>&quot;//host/&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>28</td>
      <td><code>&quot;//host&quot;</code></td>
      <td><code>&quot;//host&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>29</td>
      <td><code>&quot;foo/../foo/..&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>30</td>
      <td><code>&quot;foo/../foo/../..&quot;</code></td>
      <td><code>&quot;..&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>31</td>
      <td><code>&quot;../foo/../foo/..&quot;</code></td>
      <td><code>&quot;..&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>32</td>
      <td><code>&quot;../.f/../f&quot;</code></td>
      <td><code>&quot;../f&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>33</td>
      <td><code>&quot;../f/../.f&quot;</code></td>
      <td><code>&quot;../.f&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>34</td>
      <td><code>&quot;.././../.&quot;</code></td>
      <td><code>&quot;../..&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    <tr>
      <td>35</td>
      <td><code>&quot;.././.././&quot;</code></td>
      <td><code>&quot;../../&quot;</code></td>
      <td>&nbsp;</td>
    </tr>
    </table>
</blockquote>




</font>




<blockquote>
<p><i><span style="background-color: #FFFF99">Beman and Nico need to resolve 
their different suggested normalizations for cases 8, 9, 12, and 14. This is 
probably just a synchronization issue rather than a controversy. They are in 
total agreement about invariants and are just trying to settle on the best 
wording to describe the desired results.</span></i></p>
<p><i>Change 27.10.4.12 [fs.def.normal.form]:</i></p>
<blockquote>
<p><b>normal form</b><br>
<del>A path with no redundant current directory (<i>dot</i>) elements, no redundant parent directory (<i>dot-dot</i>) 
elements, and no redundant <i>directory-separators</i>. The normal form for an 
empty path is an empty path. The normal form for a path ending in a
<i>directory-separator</i> that is not the root directory has a current directory (<i>dot</i>) 
element appended. A path in normal form is said to be <i>normalized</i>. The 
process of obtaining a normalized path from a path that is not in normal form is 
called <i>normalization</i>. [ <i>Note:</i> The rule that appends a current 
directory (<i>dot</i>) element supports operating systems like OpenVMS that use 
different syntax for directory names and regular file names. <i>—end note</i> ]</del></p>

<hr>

<p><span style="font-style: italic; background-color: #FFFF99">
<a name="option-a">Option A</a></span></p>

<p><ins>A generic format pathname is in normal form if empty or has:</ins></p>
<ul>
  <li><ins>Only single <i>directory-separator</i> sequences or a leading sequence of two <i>
  directory-separator</i>s.<br>
</ins>&nbsp;</li>
  <li><ins>No <i>dot</i> filenames except:</ins><ul>
  <li><ins>if the pathname would otherwise be empty, the pathname consists of a
  <i>dot</i> filename .</ins></li>
  <li><ins>if the <i>relative-path</i> 
  portion of the pathname would otherwise consist solely of a <i>director-separator, 
  the</i> <i>relative-path</i> portion of the pathname consists of a <i>dot</i> 
  filename followed by a <i>preferred-separator</i>.<br>
&nbsp;</ins></li>
</ul>
  </li>
  <li><ins>No sequences of <i>name</i> <i>directory-separator</i> <i>dot-dot</i>, 
  applied recursively.<br>
&nbsp;</ins></li>
  <li><ins>Only the <i>preferred-separator</i> form of<i> directory-separator</i>s.</ins></li>
</ul>
  <p><ins>A path in normal form is said to be <i>normalized</i>. The 
process of obtaining a normalized path from a path that is not in normal form is 
called <i>normalization</i>.</ins></p>

<hr>

<p><span style="font-style: italic; background-color: #FFFF99">Option B</span></p>

<p><ins>For a non-empty generic format pathname:</ins></p>
<ul>
  <li><ins><i>dot</i> filenames are removed,&nbsp;except when removal would result in an empty pathname or change a relative pathname into 
  an absolute pathname.<br>
  <br>
</ins>&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;<ins>[<i>Example:</i> Normal form of <code>&quot;./././&quot; </code>
  is <code>&quot;./&quot;</code> on POSIX —<i>end example</i>]<br>
</ins>&nbsp;</li>
  <li><ins>
  Sequences of <i>name</i> <i>directory-separator</i> <i>dot-dot</i> are 
  recursively removed. If that would result in a trailing <i>relative-path</i>&nbsp; <i>
  directory-separator</i> becoming the first <i>relative-path</i> character, a <i>dot</i> filename is 
  first inserted at the beginning of 
  the <i>relative-path</i>.<br>
  <br>
</ins>&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;<ins>[<i>Example:</i> Normal form of <code>&quot;/foo/../&quot;
  </code>is <code>&quot;/./&quot;</code> on POSIX &mdash;<i>end example</i>]<br>
</ins>&nbsp;</li>
  <li><ins>
  Sequences of multiple <i>directory-separator</i>s, except two 
  leading <i>
  directory-separator</i>s, 
  are replaced by a single <i>preferred-separator.</i><br>
  <br>
  </ins>&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;<ins>[<i>Example:</i> Normal form of <code>&quot;//foo///bar&quot;
  </code>is <code>&quot;//foo/bar&quot;</code> on POSIX &mdash;<i>end example</i>]<br>
</ins>&nbsp;</li>
  <li><ins><i>slash</i> directory-separators are replaced by <i>preferred-separator</i>s.<br>
  <br>
  </ins>&nbsp;&nbsp;&nbsp; &nbsp;&nbsp;<ins>[<i>Example:</i> Normal form of <code>&quot;//foo/bar&quot;
  </code>is <code>&quot;\\\\foo\\bar&quot;</code> on Windows &mdash;<i>end example</i>]</ins></li>
</ul>
  <p><ins>A path in normal form is said to be <i>normalized</i>. The 
process of obtaining a normalized path from a path that is not in normal form is 
called <i>normalization</i>.</ins></p>


</blockquote>


<hr>


<p><i>Add the following note as a new paragraph at the end of [path.generic]:</i></p>
<blockquote>

<p><span style="font-style: italic; background-color: #E1E9FD">Normal form and 
several other aspects of the class path spec are hard to understand without an 
understanding of the role of a trailing directory-separator. Yet even experts 
are  not always aware of the following. Thus the added note.</span></p>
 <p><ins>[<i>Note:</i> The <i>relative-path</i> portion of a generic format pathname 
  may end with a <i>directory-separator. </i>Its use is:</ins></p>
  <ul>
    <li><ins>Required for operating systems such as z/OS and OpenVMS where 
    native path syntax for directories is different from non-directory native 
    path syntax, and thus directory paths must be lexically distinguished from 
    non-directory paths.</ins></li>
    <li><ins>Required by POSIX utilities for tool dependent purposes such as 
    indicating that a symlink is not to be followed<i>.</i></ins></li>
    <li><ins>Required by some programs to distinguish directory pathnames that are to 
    be handled in some special way.</ins></li>
    <li><ins>Recommended by some coding guidelines to clearly distinguish directory 
    names from non-directory names.</ins></li>
    <li><ins>Recommended occasionally to ensure that an error is reported rather 
    than a file overwritten if a regular file pathname is passed mistakenly as 
    an argument when a directory pathname was intended.</ins></li>
  </ul>
  <p><ins>&nbsp;&mdash;<i>end note</i>]</ins></p>
</blockquote>

<p><span style="background-color: #FFFF99; font-style:italic">Beman supplied the above wording. It 
has not been  reviewed by the small group.</span></p>
</blockquote>

<h4><a name="US51">US 51</a>; 27.10.8.4.3 [path.append] ¶2.3: Failing to add / when appending empty string prevents useful apps<br>
Status: Reject</h4>
<blockquote>

<p>Resolved by <a href="#US74">US-74/CA-3</a></p>
</blockquote>

<h4><a name="US52">US 52</a>; 27.10.8.4.5 [path.modifiers] ¶5: <code>remove_filename()</code> postcondition is not by itself a definition<br>
Status: Accept with modifications</h4>
<blockquote>

<p>Also see LWG
<a href="http://cplusplus.github.io/LWG/lwg-active.html#2665">LWG 2665</a>.</p>
<p>
<i>Change 27.10.8.4.5 [path.modifiers]:</i> </p>
<blockquote>
 <pre>path&amp; remove_filename();</pre>
 <blockquote>
   <p>
 <del><i>Postcondition:</i><code>!has_filename()</code>.</del> </p>
 <p> <ins><i>Effects:</i> 
 If <code>has_filename()</code> then remove the minimum number of trailing 
 characters from generic format pathname such that there is one less element in
 <code>[begin(), end())</code>. Otherwise, no effect.</ins></p>
 <p><i>Returns:</i> <code>*this</code>.</p>
 <p>[ <i>Example:</i></p>
   <blockquote>
<pre>std::cout &lt;&lt; path(&quot;/foo&quot;).remove_filename(); // outputs &quot;/&quot;
std::cout &lt;&lt; path(&quot;/&quot;).remove_filename();    // outputs &quot;<font FACE="LMRoman10-Regular"><code><ins>/</ins></code></font>&quot;</pre>
   </blockquote>
   <p>—<i>end example</i> ]</p>
 </blockquote>
</blockquote>
</blockquote>

<h4><a name="US53">US 53</a>; 27.10.8.4.5 [path.modifiers] ¶4 (was ¶7): <code>remove_filename()</code>'s name does not correspond to its behavior<code><br>
</code>Status: Accept with modifications</h4>

<blockquote>

<p>As a consequence of the <a href="#US52">US-52</a> and <a href="#US60">US-60</a> fixes, the name now corresponds to 
the behavior.</p>

</blockquote>

<h4><a name="US54">US 54</a>; 27.10.8.4.5 [path.modifiers] ¶7 (was ¶10): <code>
replace_filename()</code> is broken<code><br>
</code>Status: Accept with modifications</h4>

<blockquote>

<p>See <a href="#US52">US-52</a> and <a href="#US60">US-60</a> fixes.</p>

<p>
<i>Change 27.10.8.4.5 [path.modifiers]:</i></p>

  <blockquote>
    <pre>path&amp; replace_filename(const path&amp; replacement);</pre>
    <p><i>Effects:</i> Equivalent to:</p>
    <blockquote>
      <pre><ins>if (has_filename()) {</ins>
  remove_filename();
  operator/=(replacement);
<ins>}</ins></pre>
    </blockquote>
    <p><i>Returns:</i> <code>*this</code>.</p>

<p>[ <i>Example:</i></p>

    <blockquote>
      <div dir="ltr">
      <pre><ins>std::cout &lt;&lt; path(&quot;/foo&quot;).replace_filename(&quot;bar&quot;);  // outputs &quot;/bar&quot;</ins>
std::cout &lt;&lt; path(&quot;/&quot;).replace_filename(&quot;bar&quot;);     // outputs &quot;<del>bar</del>&quot;</pre>
      </div>
    </blockquote>

<p><i>—end example</i> ]</p>

  </blockquote>

<p><i>Change 27.10.12.2 [directory_entry.mods]:</i></p>

  <blockquote>
    <pre>void replace_filename(const path&amp; p);</pre>
    <blockquote>

<p><i>Postconditions:</i> <code>path() == <del>x.parent_path() / p</del> <ins>x.replace_filename(p)</ins></code> where <code>
x</code> is the value of <code>path()</code> before the function
is called.</p>

    </blockquote>
</blockquote>

<p><span style="background-color: #FFFF99; font-style:italic">Beman Dawes supplied the above wording. It 
has not been  reviewed by the small group.</span></p>

</blockquote>

    <font FACE="LMRoman10-Regular">

<h4><a name="US60">US 60</a>; 27.10.8.4.9 [path.decompose] ¶6: <code>path(&quot;/foo/&quot;).filename()==path(&quot;.&quot;)</code> is surprising<code><br>
</code>Status: Accept with modifications</h4>

<blockquote>

<p><b>Discussion:</b></p>

<p>The <code>filename()</code> behavior at issue is specified in 27.10.8.4.9 [path.decompose] ¶6:</p>

<blockquote>

<p><i>Returns</i>: <code>empty() ? 
path() : *--end()</code>.</p>

</blockquote>

<p>So the underlying problem is in 27.10.8.5 path iterators [path.itr]
¶4:</p>

  <blockquote>

<font FACE="LMRoman7-Regular" SIZE="1">
<p>4 </font><font FACE="LMRoman10-Regular">For the 
elements of </font><font FACE="LMMono10-Regular">pathname </font>
<font FACE="LMRoman10-Regular">in the generic format, the forward 
traversal order is as follows:</p>
</font><font FACE="LMRoman7-Regular" SIZE="1">
<p>(4.1) </font><font FACE="LMRoman10-Regular">— The
</font><font FACE="LMRoman10-Italic"><i>root-name </i></font>
<font FACE="LMRoman10-Regular">element, if present.</p>
</font><font FACE="LMRoman7-Regular" SIZE="1">
<p>(4.2) </font><font FACE="LMRoman10-Regular">— The
</font><font FACE="LMRoman10-Italic"><i>root-directory </i></font>
<font FACE="LMRoman10-Regular">element, if present. [</font><font FACE="LMRoman10-Italic"><i>Note:
</i></font><font FACE="LMRoman10-Regular">the generic format is 
required to ensure lexicographical</p>
<p>comparison works correctly. </font>
<font FACE="LMRoman10-Italic"><i>—end note </i></font>
<font FACE="LMRoman10-Regular">]</p>
</font><font FACE="LMRoman7-Regular" SIZE="1">
<p>(4.3) </font><font FACE="LMRoman10-Regular">— Each 
successive </font><font FACE="LMRoman10-Italic"><i>filename </i></font>
<font FACE="LMRoman10-Regular">element, if present.</p>
</font><font FACE="LMRoman7-Regular" SIZE="1">
<p>(4.4) </font><font FACE="LMRoman10-Regular">— </font>
<font FACE="LMRoman10-Italic"><i>dot</i></font><font FACE="LMRoman10-Regular">, 
if one or more trailing non-root </font><font FACE="LMRoman10-Italic">
<i>slash </i></font><font FACE="LMRoman10-Regular">characters are 
present.</p>
</font>

  </blockquote>

<p>The ¶4.4 iterator behavior has always been surprising and controversial, so 
the LWG SG looked at the possible ways of dealing with trailing non-root slash 
characters:</p>

<blockquote>

<p><del> <i>dot</i>, if one or more trailing non-root <i>
slash</i> characters are present.</del></p>

</font>

<blockquote>

<p>Not acceptable because it loses the fact that the path represents a 
directory, yet knowing that is required for some operating systems (e.g. POSIX, 
z/OS, OpenMVS).&nbsp; We want to ensure that reconstruction by iteration yields 
the same path as the original for  definitions of &quot;same&quot;.</p>

</blockquote>

    <font FACE="LMRoman10-Regular">

<p> <i><del>dot</del> <ins>slash</ins></i>, if one or more trailing non-root <i>
slash</i> characters are present.</p>

</font>

<blockquote>

<p>Not acceptable 
because a slash is semantically ambiguous; it is already used to indicate a 
root, so also using it to indicate a directory path would be a trap all too easy 
to fall into.</p>

</blockquote>

    <font FACE="LMRoman10-Regular">

<p> <i><del>dot</del> </i><ins>an empty element</ins>, if one or more trailing non-root <i>
slash</i> characters are present.</p>

</font>

<blockquote>

<p>This works, and was discussed at length by the SG. It appears to resolve a 
number of concerns raised by other NB comments and by the SG during discussion 
of those comments. All the ripple effects of this 
change need to be identified, and  wording changes proposed where needed or desirable.<font FACE="LMRoman10-Regular">&nbsp;</font></p>

<p><font face="LMRoman10-Regular">Because this is a design change, Nico 
presented it to LEWG as part of &quot;Clarify filename&quot;. It was accepted by a vote of 
13/3/0/0/0. </font> </p>

</blockquote>

</blockquote>

<p><b><font face="LMRoman10-Regular">Proposed resolution:</font></b></p>

<p><i><font face="LMRoman10-Regular">Change 27.10.8.5 [path.itr]:</font></i></p>

  <blockquote>

<p>

    <font FACE="LMRoman10-Regular">

<font FACE="LMRoman7-Regular" SIZE="1">
(4.4) </font>— <i><del>dot</del> </i><ins>an empty element</ins>, 
if one or more trailing non-root <font FACE="LMRoman10-Italic">
<i>slash </i></font>characters are 
present.</font></p>

  </blockquote>

<p><span style="font-style: italic; background-color: #FFFF99">Also see LWG 2665 remove_filename() post condition is 
incorrect</span></p>

<p><span style="font-style: italic; background-color: #FFFF99">Although not a NB comment,
<a href="http://cplusplus.github.io/LWG/lwg-active.html#2665">LWG 2665</a> needs 
to be coordinated with this set of comments, particularly US-52. It is mentioned 
here as a reminder to the LWG.</span></p>

</blockquote>

<h3>Other comments</h3>

<h4><a name="US31">US 31</a>; 27.10 [filesystems]: Everything is defined in terms of one implicit host system<code><br>
</code>Status: Reject</h4>
<blockquote>
<p>That&#39;s by design, aimed at achieving portable syntax and portable behavior. 
And that objective has been largely achieved in practical real-world use for 
many years.&nbsp; There is no proposed wording, and without proposed wording 
there is no consensus for change.&nbsp; Will re-open if a paper or issue appears 
with proposed wording.</p>

</blockquote>

<h4><a name="US32">US 32</a>; 27.10.2.1 [fs.conform.9945] ¶3: Meaning of POSIX 
conformance unclear<code><br>
</code>Status: Accept with modifications</h4>

<blockquote>

<p>The comment requests &quot;Clarify that ¶2 governs and an error must be 
reported in such cases&quot;.</p>

<p><i>Change 27.10.2.1 [fs.conform.9945]:</i></p>
  <blockquote>
    <p>Implementations are not required to provide behavior that is not 
    supported by a particular file system. [ <i>Example:</i> The FAT file system used 
    by some memory cards, camera memory, and floppy disks does not support hard 
    links, symlinks, and many other features of more capable file systems, so 
    implementations are not required to support those features on the FAT file 
    system <ins>but instead are required to report an error as described above</ins>. —<i>end example</i> ]</p>
  </blockquote>
    <p><i><span style="background-color: #FFFF99">Jonathan Wakely may supply improved wording. 
    The above change, however, is probably sufficient to resolve the NB comment.</span></i></p>
</blockquote>
<h4><a name="US33">US 33</a>; 27.10.4.2 [fs.def.canonical.path]: Definition of <i>canonical path</i> problematic<br>
Status: Accept</h4>
<blockquote>
<p>The comment appears to be correct in that the definition of canonical path is 
not used in the WP.</p>
<p><i>Remove 27.10.4.2 [fs.def.canonical.path]:</i></p>
<blockquote>
<p><del><b>canonical path</b><br>
An absolute path that has no elements that are symbolic links, and no dot or 
dot-dot elements (27.10.8.1).</del></p>
</blockquote>
</blockquote>
<h4><a name="US34">US 34</a>; 27.10.4.5 [fs.def.filesystem]: Are there attributes of a file that are not an aspect of the file system?<br>
Status: Accept with modifications</h4>
<blockquote>
<p><i>Change 27.10.4.5 [fs.def.filesystem]:</i></p>
  <blockquote>
<p>A collection of files and <del>certain of</del> their attributes.</p>
  </blockquote>
</blockquote>
<h4><a name="US35">US 35</a>; 27.10.4.6 [fs.def.race]: What synchronization is required to avoid a file system race?<code><br>
</code>Status: Reject</h4>
<blockquote>
<p>27.10.4 Terms and definitions [fs.definitions] is not the proper place to 
describe file system race behavior. 27.10.2.3 File system race behavior [fs.race.behavior] 
would be the proper place. There is no proposed wording, and without proposed 
wording there is no consensus for change.&nbsp; Will re-open if a paper or issue 
appears with proposed wording.</p>
</blockquote>
<h4><a name="US36">US 36</a>; 27.10.4.9 [fs.def.link]: Symbolic links themselves are attached to a directory via (hard) links<br>
Status: Accept with modifications</h4>
<blockquote>
<p>The SG provides wording, taken from POSIX definitions 3.130, Directory Entry 
(or Link).</p>
<h4><span style="font-weight: 400; background-color:#FFFF99; font-style:italic">
Davis Herring plans to open an issue on a related concern that does not affect 
the resolution of this NB comment below. </span></h4>
<p><i>Change 27.10.4.9 [fs.def.link]:</i></p>
  <blockquote>
<p><b>link</b><br>
<del>A directory entry that associates a filename with a file. A link is either 
a hard link (27.10.4.8) or a symbolic link (27.10.4.21).</del> <ins>An object 
that associates a filename with a file. Several links can associate names 
with the same file.</ins></p>
  </blockquote>
</blockquote>
<h4><a name="US40">US 40</a>; 27.10.4.15 [fs.def.parent]: Not all directories 
have a parent<br>
Status: Reject</h4>

<blockquote>
<p>No proposed wording. The definition in question is taken directly from 
the POSIX
<font face="Times New Roman">§</font>3.268 <i>Parent Directory</i> definition. </p>

</blockquote>

<h4><a name="US43">US 43</a>; 27.10.5 [fs.req] ¶4: Concerns about <i>encoded character types</i><br>
Status: Accept with modifications</h4>

<blockquote>

<p><i>Change 27.10.5 
Requirements [fs.req] ¶4:</i></p>

  <blockquote>

<p>[<i>Note:</i> Use of an encoded character type implies an associated <ins>character set and</ins> encoding.
Since <code>signed char</code> and <code>unsigned char</code> have 
no implied <ins>character set and</ins> encoding, they are not included as permitted types. 
<i>—end note</i> ]</p>

  </blockquote>

<p><i>Add a sentence to 27.10.8.2.2 path type and encoding conversions [path.type.cvt]:</i></p>

  <blockquote>

<p>If the encoding being converted to has no representation for source 
characters, the resulting converted characters, if any, are unspecified. 
<ins>Implementations are strongly encouraged  not to mutate member function arguments 
if already of type <code>path::value_type</code>.</ins>
<span style="background-color: #FFFF99; font-style:italic">Billy O&#39;Neal suggested this addition 
and Beman Dawes supports it, but it 
has not been reviewed by the rest of the filesystem small group.</span></p>

  </blockquote>

</blockquote>

<h4><a name="US44">US 44</a>; 27.10.8 [class.path]: Definition of path in terms of a string requires leaky abstraction<code><br>
</code>Status: Accept with modifications</h4>
<blockquote>

<p>Proposed wording, including 
a note, has been applied to LWG 2734 to partially resolve US 44.</p>

<p>The portion of this comment not resolved by LWG 2734 is resolved by
<a href="#US74">US-74/CA-3</a> </p>

<p> <span style="background-color: #FFFF99; font-style:italic">Beman Dawes will check 2734, and 
coordinate with Marshall Clow if needed, or add a comment to 2734 </span></p>

</blockquote>

<h4><a name="US45">US 45</a>; 27.10.8.1 [path.generic]: Generic format portability compromised by unspecified 
<i>root-name</i><code><br>
</code>Status: Accept with modifications</h4>
<blockquote>
<p>Resolved by <a href="#US73">US-73/CA-2</a></p>

</blockquote>

<h4><a name="US46">US 46</a>; 27.10.8.1 [path.generic]: <i>filename</i> can be empty so productions for <i>relative-path</i> are redundant<br>
Status: Accept with modifications</h4>
<blockquote>
<p><i>Change [path.generic]:</i></p>

<blockquote>
<p><code>relative-path:<br>
&nbsp; filename<br>
&nbsp; relative-path directory-separator<br>
&nbsp; relative-path directory-separator filename<br></code>
&nbsp;
&nbsp; <ins>an empty path</ins>&nbsp;&nbsp;&nbsp; <i>
<span style="background-color: #C0C0C0">Note to editor: this line is 
deliberately not in code font</span></i></p>

<p>...</p>

<p><i>name:</i><br>
&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; A <ins>non-empty</ins> sequence 
of characters other than directory-separator characters.</p>

</blockquote>

<p><span style="background-color: #FFFF99; font-style:italic">Jonathan Wakely supplied the above wording. It 
has been reviewed by Beman Dawes, but not the rest of the small group.</span></p>

</blockquote>

<h4><a name="US48">US 48</a>; 27.10.8.1 [path.generic] ¶1: Multiple separators are often meaningful in a <i>root-name</i><br>
Status: Accept with modifications</h4>
<blockquote>

<p><i>Change 27.10.8.1 [path.generic] ¶1:</i></p>

  <blockquote>
<p><ins>Except in <i>root-name</i>, m</ins><del>M</del>ultiple successive <i>
directory-separator</i> characters are considered to be the same as one <i>
directory-separator </i>character.</p>

  </blockquote>

</blockquote>

<h4><a name="US49">US 49</a>; 27.10.8.2.2 [path.type.cvt]: What does “method of conversion method” mean?<br>
Status: Editorial</h4>
<blockquote>

<p>Fixed in the post-Issaquah working paper.</p>
</blockquote>

<font FACE="LMRoman10-Regular">

<h4><a name="US55">US 55</a>; 27.10.8.4.5 [path.modifiers] ¶11: <code>replace_extension()</code>'s use of <code>path</code> as parameter is inappropriate<br>
Status: Accept with modifications</h4>

<blockquote>
<p>The SG believes <code>path</code> is the appropriate parameter type, but that 
the standard needs to clarify class <code>path</code> usage.</p>
<p><i>Change 27.10.8 [class.path] ¶1:</i></p>

<blockquote>
<p>An object of class path represents a path (27.10.4.17)
 and contains a pathname (27.10.4.18). Such an object
is concerned only with the lexical and syntactic aspects of a path. The path does not necessarily exist in
external storage, and the pathname is not necessarily valid for the current operating system or for a particular
file system.</p>

<p><ins>[<i>Note:</i> Class path is used to support the differences between the string types
 used by different operating systems to represent pathnames, and to perform 
conversions between encodings when necessary. &mdash; <i>end note</i>]</ins></p>

</blockquote>
 
</blockquote>
 
<h4><a name="US56">US 56</a>; 27.10.8.4.5 [path.modifiers] ¶11.2: Remove <code>
replace_extension()</code>&#39;s conditional addition of period<br>
Status: Reject</h4>

<blockquote>
 
<p>Nico presented this to the Library Evolution Working group in 
Issaquah, including a number of examples (See Nico&#39;s <i>Filesystem: Filename 
Issues</i> slides on the Issaquah LEWG wiki.) There was no 
consensus for change; the vote was 3/3/3/7/2. See the Issaquah LEWG notes.</p>

</blockquote>

<h4><a name="US57">US 57</a>; 27.10.8.4.8 [path.compare] ¶2: On Windows, absolute paths will sort in among relative paths<br>
Status: Reject</h4>

<blockquote>

<p>Insufficient motivation for change.</p>

</blockquote>

<h4><a name="US58">US 58</a>; 27.10.8.4.9 [path.decompose] ¶5: <code>parent_path()</code> behavior for root paths is useless<br>
Status: Accept with modifications</h4>

<blockquote>

<p><i>Change [path.decompose]:</i></p>

  <blockquote>
    <p><tt>path parent_path() const;</tt> </p>
    <p><ins>-?- Let <tt>parentEnd</tt> be an instance of <tt>path::iterator</tt>, 
    initialized as follows:</ins> </p>
    <ul>
      <li><ins>if <tt>has_relative_path()</tt>, from <tt>--end()</tt></ins></li>
      <li><ins>otherwise, from <tt>end()</tt></ins></li>
    </ul>
  </blockquote>
  <p></p>
  <blockquote>
    <p>-5- <em>Returns:</em> <del><tt>(empty() || begin() == --end()) ? path() :
    </tt><em>pp</em>, where <em>pp</em> is </del><ins>a <tt>path</tt> </ins>
    constructed as if by starting with an empty <tt>path</tt> and successively 
    applying <tt>operator/=</tt> for each element in the range <tt>[begin(),
    <del>--end()</del><ins>parentEnd</ins>)</tt>.</p>
  </blockquote>

</blockquote>

<h4><a name="US59">US 59</a>; 27.10.8.4.9 [path.decompose] ¶6: <code>filename()</code> returning <code>path</code> for single path components is bizarre<br>
Status: Reject</h4>

<blockquote>

<p> <code>path</code> is filesystem's
vocabulary type for both full paths and for path components, in the same sense that
<code>std::string</code> might be used as the vocabulary type for both sentences and 
the words in the
sentences. See <a href="#US55">US 55</a> for further discussion.</p>

</blockquote>

<h4><a name="US61">US 61</a>; 27.10.8.4.9 [path.decompose] ¶8: Leading dots in <code>filename()</code> should not begin an extension<code><br>
</code>Status: Accept with modifications</h4>

<blockquote>

<p>Nico Josuttis presented this to the Library Evolution Working group in 
Issaquah, including a number of examples (See Nico&#39;s <i>Filesystem: Filename 
Issues</i> slides on the Issaquah LEWG wiki.) Support was unanimous. See the 
Issaquah LEWG notes.</p>

<p>Nico presented the following cases; the only change from the TS is for the
<code>&quot;.profile&quot;</code> case:</p>

<blockquote>

  <table border="1" cellpadding="5" cellspacing="0" style="border-collapse: collapse" bordercolor="#111111">
    <tr>
      <td><i><b><code>path p</code></b></i></td>
      <td><i><b><code>p.stem()</code></b></i></td>
      <td><i><b><code>p.extension()</code></b></i></td>
    </tr>
    <tr>
      <td><code>&quot;/foo/bar.txt&quot;</code></td>
      <td><code>&quot;bar&quot;</code></td>
      <td><code>&quot;.txt&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;.profile&quot;</code></td>
      <td><code><del>&quot;&quot;</del> <ins>&quot;.profile&quot;</ins></code></td>
      <td><code><del>&quot;.profile&quot;</del> <ins>&quot;&quot;</ins></code></td>
    </tr>
    <tr>
      <td><code>&quot;.profile.old&quot;</code></td>
      <td><code>&quot;.profile&quot;</code></td>
      <td><code>&quot;.old&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;..abc&quot;</code></td>
      <td><code>&quot;..abc&quot;</code></td>
      <td><code>&quot;&quot;</code></td>
    </tr>
    <tr>
      <td>&quot;<code>...abc</code>&quot;</td>
      <td><code>&quot;...abc&quot;</code></td>
      <td><code>&quot;&quot;</code></td>
    </tr>
    <tr>
      <td>&quot;<code>abc..def</code>&quot;</td>
      <td><code>&quot;abc.&quot;</code></td>
      <td><code>&quot;.def&quot;</code></td>
    </tr>
    <tr>
      <td>&quot;<code>abc...def</code>&quot;</td>
      <td><code>&quot;abc..&quot;</code></td>
      <td><code>&quot;.def&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;abc.&quot;</code></td>
      <td><code>&quot;abc&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;abc..&quot;</code></td>
      <td><code>&quot;abc.&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;abc.d.&quot;</code></td>
      <td><code>&quot;abc.d&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;..&quot;</code></td>
      <td><code>&quot;..&quot;</code></td>
      <td><code>&quot;&quot;</code></td>
    </tr>
    <tr>
      <td><code>&quot;.&quot;</code></td>
      <td><code>&quot;.&quot;</code></td>
      <td><code>&quot;&quot;</code></td>
    </tr>
  </table>
</blockquote>

<p><i>Change 27.10.8.4.9 [path.decompose] ¶ 8:</i></p>

<blockquote>

<p><code>path stem() const;</code></p>

  <blockquote>

<p><i>Returns:</i> if <code>filename()</code> contains a period <ins>other than 
at the beginning and</ins> <del>but</del> does not 
consist solely of one or two periods, returns the substring of <code>filename()</code> 
starting at its beginning and ending with the character before the last period.
Otherwise, returns <code>filename()</code>.</p>

  </blockquote>
</blockquote>

<p><i>Change 27.10.8.4.9 [path.decompose] ¶ 10:</i></p>

<blockquote>

<p><code>path extension() const;</code></p>

  <blockquote>

<p><i>Returns:</i> if <code>filename()</code> contains a period <ins>other than 
at the beginning and</ins> <del>but</del> does not consist solely of one or two 
periods, returns the substring of <code>filename()</code> 
starting at the rightmost period and for the remainder of the path. Otherwise, 
returns an empty <code>path</code> object.</p>

  </blockquote>
</blockquote>

<p><span style="background-color: #FFFF99"><i>Beman Dawes supplied the above 
wording per the LEWG vote. It has not been reviewed by the small group.</i></span></p>

</blockquote>

<h4><a name="US62">US 62</a>; 27.10.8.4.9 [path.decompose] ¶13 (was ¶11): It is important that <code>stem()+extension()==filename()<br>
</code>Status: Reject</h4>

<blockquote>

<p>Yes, it is important. But the note in ¶13 already says exactly that, so
the question becomes "Should we make that part of the note normative?"</p>

<p>Saying that normatively would be a second way of saying 
what is already specified behavior, and saying the same normative thing twice is 
not good standardese. No proposed wording. </p>

</blockquote>

<h4><a name="US63">US 63</a>; 27.10.8.4.11 [path.gen] ¶1: <code>lexically_normal()</code> inconsistently treats trailing "/" but not &quot;/..&quot; as directory<code><br>
</code>Status: Accept with modifications</h4>

<blockquote>

<p>See <a href="#US37">US-37</a> and the <a href="#path.gen">change to 
[path.gen]</a> in <a href="#US74">US-74/CA-3</a>.</p>

</blockquote>

<h4><b><a name="US185">US 185</a>; 27.10.7 [fs.err.report]: </b>Fold <code>error_code</code> and non-<code>error_code</code> signatures into one signature<br>
Status: Reject</h4>

<blockquote>

<p> Does not handle one signature being <code>
noexcept(true)</code> and the other <code>noexcept(false)</code>, and many other 
problems. Even the submitter is no longer in favor. See LWG mailing list thread 
&quot;[filesystem] US-185 Eliminating dual signatures for operational functions&quot;, 
October 25, 2016.</p>

</blockquote>

<h4><a name="FI 14">FI 14</a>; 27.10.12.3 [directory_entry.obs]: <tt>directory_entry</tt> comparisons are members<br>
Status: Reject</h4>



<blockquote>

<p>This is LWG issue
<a href="http://cplusplus.github.io/LWG/lwg-closed.html#2761">2761</a>, which 
has been closed as NAD. </p>

</blockquote>



<h4><a name="US73">US 73</a>, CA 2; 27.10.8.1 [path.generic]: <i>root-name</i> is effectively implementation defined<br>
Status: Accept with modifications</h4>

<blockquote>



<p><i>Change [path.generic]:</i></p>

  <blockquote>
    <p><i>root-name:</i><br>
    An operating system dependent name that identifies the starting location for 
    absolute paths. <ins>If the operating system does not define at least one <i>
    root-name</i>, then the implementation defines a <i>root-name</i>. 
    Implementations are permitted to define additional <i>root-name</i>s.</ins> 
    [ <i>Note:</i> Many operating systems define a name beginning with two <i>
    directory-separator</i> characters as a <i>root-name</i> that identifies 
    network or other resource locations. Some operating systems define a single 
    letter followed by a colon as a drive specifier – a <i>root-name</i> 
    identifying a specific device such as a disk drive. —<i>end note</i> ]</p>
  </blockquote>
  <p><i>Add a new paragraph at the end of [path.generic]:</i></p>
  <blockquote>
    <p><ins>If <i>root-name</i> is otherwise ambiguous, the possibility with the 
    longest sequence of characters is chosen. [ Note: on a POSIX-like operating 
    system, it is impossible to have a <i>root-name</i> and a <i>relative-path</i> 
    without an intervening <i>root-directory element</i>. -- end note ]</ins></p>
  </blockquote>
</blockquote>

<h4><a name="US74">US 74</a>, CA 3; 27.10.8 [class.path]: The term “pathname” is ambiguous in some contexts<br>
Status: Accept with modifications</h4>

<p><span style="background-color: #FFFF99"><i>Davis Herring supplied the wording 
at the request of the filesystem small group in Issaquah. The wording has been 
reviewed by Beman Dawes but not the rest of the small group.</i></span></p>

<p><span style="background-color: #FFFF99"><i>This NB comment is a showstopper 
for C++17. These specification improvements enable the resolution of many other 
filesystem NB comments.</i></span></p>

<p><b>Discussion: </b></p>
<p>Many operations on <code>filesystem::path</code> are defined in terms of its 
exposition-only <code>pathname</code> member, which is defined implicitly by
<code>path::native()</code> as the native format path. However, the native 
format need not be compatible with lexical operations like, for example, <code>
path::root_name()</code> and <code>path::extension()</code>. </p>
<p>This same concern was also raised by
<a href="http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p0430r0.pdf">
P0430R0</a> Section 2.1, which provided additional rationale.</p>
<p>The proposed resolution below clarifies the distinct role played by the <code>
+= operator</code> as mentioned in US 44 and addresses US 74/CA 3 (with a more 
extensive resolution). </p>

<p><b>Proposed resolution:</b> </p>
<p><i>Remove In the [class.path] synopsis, remove the private (exposition-only) member from 
class path:</i></p>
<blockquote>
  <pre><del>private:
  string_type pathstring; // exposition only</del></pre>
</blockquote>
<p><i>Change [path.fmt.cvt]:</i></p>
<p><span style="background-color: #FFFF99; font-style:italic">Option A; from Davis: </span> </p>
<blockquote>
<p>[ <i>Note:</i> The format conversions described in this section are not applied on 
POSIX-<del> or Windows-</del>based operating systems because on these systems:</p>
<blockquote>
<p>— The generic format is acceptable as a native path.<br>
— There is no need to distinguish between native format and generic format in 
function arguments.<br>
— Paths for regular files and paths for directories share the same syntax.</p>
</blockquote>
<p>—end note ]</p>
</blockquote>
<p><span style="background-color: #FFFF99; font-style:italic">Option B; from Beman:</span></p>
<blockquote>
<p>[ <i>Note:</i> The format conversions described in this section are <ins>
usually</ins> not applied on 
POSIX- or Windows-based operating systems because on these systems:</p>
<blockquote>
<p>— The generic format is acceptable as a native path.<br>
— There is <ins>usually</ins> no need to distinguish between native format and 
generic format in function arguments.<br>
— Paths for regular files and paths for directories share the same syntax.</p>
</blockquote>
<p><ins>Even on these operating systems, the format conversions may be applied 
for pathnames with implementation-defined <i>root-name</i>s ([path.generic]).</ins> </p>
<p>—end note ]</p>
</blockquote>
<p><span style="font-style: italic; background-color: #FFFF99">We need to choose between Option A 
or Option B.</span></p>
<blockquote>
  <p><ins>Several functions are defined to accept <i>detected-format</i> character 
  sequence arguments that represent paths using either the generic pathname 
  format grammar ([path.generic]) or the native pathname format ([fs.def.native]). Such an 
  argument is taken to be in the generic format if and only if it matches the 
  generic format but is not acceptable to the operating system as a native path.
  </ins>
  </p>
  <p><del>Function arguments that take character sequences representing paths 
  may use the generic pathname format grammar (27.10.8.1) or the native pathname 
  format (27.10.4.11). If and only if such arguments are in the generic format 
  and the generic format is not acceptable to the operating system as a native 
  path, conversion to native format shall be performed during the processing of 
  the argument.</del></p>
<p>[ Note: Some operating systems may have no unambiguous way to 
distinguish between native format and generic format arguments. This is by 
design as it simplifies use for operating systems that do not require 
disambiguation. An implementation for an operating system where disambiguation 
is required is permitted to distinguish between the formats. —end note ]</p>
<p><span style="background-color: #FFFF99"><a name="US74markup"></a><i>Beman 
says his code font markup of text from Davis in the following paragraph is almost 
certainly wrong:</i></span></p>
<ul>
  <li>
<p><span style="background-color: #FFFF99; font-style:italic">Would s/Define G(n) and N(g) 
as the functions/Let G and N be the implementation&#39;s functions/ be clearer?</span></p>
  </li>
  <li>
<p><span style="background-color: #FFFF99; font-style:italic">Are some of the <code>&#39;=&#39;</code> 
supposed to be <code>&#39;==&#39;?</code></span></p>
  </li>
</ul>
<p><ins>Pathnames are converted as needed between the generic and native formats 
in an operating-system-dependent manner. Define <code>G(n)</code> and <code>N(g)</code> 
as the functions that convert to the generic and native formats respectively. If
<code>g=G(n)</code> for some <code>n</code>, then <code>G(N(g))=g</code>; if
<code>n=N(g)</code> for some <code>g</code>, then <code>N(G(n))=n</code>. [ <i>
Note:</i> Neither <code>G</code> nor <code>N</code> need be invertible. —<i>end 
note</i> ] </ins></p>
<p>If the native format requires paths for regular files to be formatted 
differently from paths for directories, the path shall be treated as a directory 
path if its last element is a directory-separator, otherwise it shall be treated 
as a path to a regular file.</p>
  <p><ins>When a path is constructed from or is assigned a single representation 
  separate from any path, the other representation is selected by the 
  appropriate conversion function (<code>G</code> or <code>N</code>). </ins> </p>
  <p><ins>When the (new) value <code>p</code> of one representation of a path is derived from the 
  representation of that or another path, a value <code>q</code> is chosen for the other 
  representation. The value <code>q</code> converts to <code>p</code> (by <code>G</code> or 
  <code>N</code> as appropriate) if any 
  such value does so; <code>q</code> is otherwise unspecified. [ <i>Note:</i> If 
  <code>q</code> is the result of 
  converting any path at all, it is the result of converting <code>p</code>.<i>—end note</i> ]
  </ins>
  </p>
<p><i><span style="background-color: #E1E9FD">[SG note: P0430R1 (for US 75/CA 4) 
adds support for explicitly specifying the format to use, at least in the 
constructor.]</span></i><span style="background-color: #E1E9FD"> </span></p>
</blockquote>
<p><i>Change [path.construct]: </i> </p>
<blockquote>
  <pre>path(const path&amp; p);
path(path&amp;&amp; p) noexcept;</pre>
  <blockquote>
  <p><i>Effects:</i> Constructs an object of class <code>path</code>
  <ins>having the same native and generic format pathnames as <code>p</code></ins> 
  <del>with <code>pathstring</code> having the original value of <code>p.pathstring</code></del>. In
  the second form, <code>p</code> is left in a valid but unspecified state.</p>
  </blockquote>
  <pre>path(string_type&amp;&amp; source);</pre>
  <blockquote>
    <p><i>Effects:</i> Constructs an object of class <code>path</code>
     <ins>for which the pathname in the detected-format&nbsp; of 
  <code>source</code> has</ins> <del>with
    <code>pathstring</code> having</del> the original value of <code>source</code>
     <ins>([path.fmt.cvt])</ins>.
    <code>source</code> is left in a valid but unspecified state.</p>
  </blockquote>
  <pre>template &lt;class Source&gt;
path(const Source&amp; source);
template &lt;class InputIterator&gt;
path(InputIterator first, InputIterator last);</pre>
  <blockquote>
    <p><i>Effects:</i> <ins>Let <code>s</code> be the effective range of <code>
    source</code> ([path.req]) or the range <code>[first, last)</code>, with the 
    encoding converted if required ([path.cvt]). Finds the detected-format of
    <code>s</code> ([path.fmt.cvt]) and constructs an object of class <code>path</code> 
    for which the pathname in that format is <code>s</code>.</ins>&nbsp; <del>Constructs an object of class <code>path</code>, storing 
    the effective range of source (27.10.8.3) or the range <code>[first, last)</code> 
    in <code>pathstring</code>, converting format and encoding if required 
    (27.10.8.2).</del></p>
  </blockquote>
  <pre>template &lt;class Source&gt;
path(const Source&amp; source, const locale&amp; loc);
template &lt;class InputIterator&gt;
path(InputIterator first, InputIterator last, const locale&amp; loc);</pre>
  <blockquote>
    <p><i>Requires:</i> The value type of <code>Source</code> and <code>
    InputIterator</code> is <code>char</code>. </p>
    <p>&nbsp;<i>Effects:&nbsp; </i> <del>Constructs an object of class <code>path</code>, 
    storing the effective range of source or the range <code>[first, last)</code> 
    in <code>pathstring</code>, after converting format if required and</del> <ins>
    Let <code>s</code> be the effective range of <code>source</code> or the 
    range <code>[first, last)</code>,</ins><i> </i>after 
    converting the encoding as follows:</p>
    <blockquote>
    <p>— If <code>value_type</code> is <code>wchar_t</code>, converts to the 
    native wide encoding (27.10.4.10) using the <code>codecvt&lt;wchar_t, char, 
    mbstate_t&gt;</code> facet of <code>loc</code>.</p>
    <p>&nbsp;— Otherwise a conversion is performed using the <code>
    codecvt&lt;wchar_t, char, mbstate_t&gt;</code> facet of <code>loc</code>, and then 
    a second conversion to the current narrow encoding.</p>
    </blockquote>
    <p><ins>Finds&nbsp;the detected-format of <code>s</code> ([path.fmt.cvt]) and constructs an object of class 
    <code>path</code> for which the pathname in that format is <code>s</code>.</ins></p>
  </blockquote>
</blockquote>
<p><i>In the example after [path.fmt.cvt] ¶ 7.2, change:</i></p>
<blockquote>
<p>For POSIX-based operating systems, the path is constructed by first 
using <code>latin1_facet</code> to convert ISO/IEC 8859-1 encoded <code>
latin1_string</code> to a wide character string in the native wide encoding 
(27.10.4.10). The resulting wide string is then converted to a narrow character
<code><del>pathstring</del></code> <ins>pathname</ins> string in the current 
native narrow encoding. If the native wide encoding is UTF-16 or UTF-32, and the 
current native narrow encoding is UTF-8, all of the characters in the ISO/IEC 
8859-1 character set will be converted to their Unicode representation, but for 
other native narrow encodings some characters may have no representation.</p>
<p>For Windows-based operating systems, the path is constructed by 
using <code>latin1_facet</code> to convert ISO/IEC 8859-1 encoded <code>
latin1_string</code> to a UTF-16 encoded wide character <code><del>pathstring</del></code>
<ins>pathname</ins> string. All of the characters in the ISO/IEC 8859-1 
character set will be converted to their Unicode representation.</p>
</blockquote>
<p><i>Change [path.assign]: </i> </p>
<blockquote>
  <pre>path&amp; operator=(const path&amp; p);</pre>
  <blockquote>
<p><i>Effects:</i> If <code>*this</code> and <code>p</code> are the same object, 
has no effect. Otherwise, <ins>sets the native and generic format pathnames to 
those of <code>p</code></ins> <del>modifies <code>pathstring</code> to have the original 
value of <code>p.pathstring</code></del>.</p>
<p><i>Returns:</i> <code>*this</code>.</p>
  </blockquote>
  <pre>path&amp; operator=(path&amp;&amp; p) noexcept;</pre>
  <blockquote>
<p><i>Effects:</i> If <code>*this</code> and <code>p</code> are the same object, 
has no effect. Otherwise, <ins>sets the native and generic format pathnames to 
those of <code>p</code></ins> <del>modifies <code>pathstring</code> to have the original 
value of <code>p.pathstring</code></del>. <code>p</code> is left in a valid but 
unspecified state. [ <i>Note:</i> A valid implementation is <code>swap(p)</code>. 
<i>—end 
note</i> ]</p>
<p><i>Returns:</i> <code>*this</code>.</p>
  </blockquote>
  <pre>path&amp; operator=(string_type&amp;&amp; source);
path&amp; assign(string_type&amp;&amp; source);</pre>
  <blockquote>
    <p><i>Effects:</i> <ins>Sets the native and generic format pathnames to 
those of <code>source</code></ins> <del>Modifies <code>pathstring</code> to have the original value 
    of <code>source</code></del>. <code>source</code> is left in a valid but 
    unspecified state.</p>
    <p><i>Returns:</i> <code>*this</code>.</p>
  </blockquote>
  <pre>template &lt;class Source&gt;
path&amp; operator=(const Source&amp; source);
template &lt;class Source&gt;
path&amp; assign(const Source&amp; source);
template &lt;class InputIterator&gt;
path&amp; assign(InputIterator first, InputIterator last);</pre>
  <blockquote>
    <p><i>Effects: </i><ins>Let <code>s</code> be the effective range of <code>source</code> ([path.req]) or the 
    range <code>[first, last)</code>, with the encoding converted if required ([path.cvt]). 
    Finds the detected-format of <code>s</code> and sets the pathname in that format 
    to <code>s</code> ([path.fmt.cvt]).</ins> <del>Stores the effective range of <code>source</code> 
    (27.10.8.3) or the range <code>[first, last)</code> in <code>pathstring</code>, 
    converting format and encoding if required (27.10.8.2).</del></p>
    <p><i>Returns:</i> <code>*this</code>.</p>
  </blockquote>
</blockquote>
<p><i>Change [path.append]




</font>




<font FACE="Times New Roman">

¶2 &amp; 3</font><font FACE="LMRoman10-Regular">:</font></i><font FACE="LMRoman10-Regular"></p>
<blockquote>
  <pre>path&amp; operator/=(const path&amp; p);</pre>
  <blockquote>
<p><i>Effects:</i> <ins>If equivalent to <code>has_filename() || !has_root_directory() &amp;&amp; is_absolute()</code>, 
then appends <code>path::preferred_separator</code> to the generic format pathname. [ <i>
Example:</i> <code>path(&quot;//host&quot;)/&quot;foo&quot;</code> and <code>path(&quot;//host/&quot;)/&quot;foo&quot;</code> both
 equal <code>&quot;//host/foo&quot;</code>. <i>—end example</i> ]</ins></p>
<p>&nbsp;<del>Appends <code>path::preferred_separator</code> to <code>pathstring</code> 
unless:</del></p>
    <blockquote>
<p><del>— an added <i>directory-separator</i> would be redundant, or</del></p>
<p><del>— an added <i>directory-separator</i> would change a relative path into an 
absolute path [<i>Note:</i> An empty path is relative.<i>—end note</i> ] , or</del></p>
<p><del>— <code>p.empty()</code> is <code>true</code>, or</del></p>
<p><del>— <code>*p.native().cbegin()</code> is a directory-separator.</del></p>
    </blockquote>
<p>Then appends <code>p.native()</code> to <ins>the native format pathname</ins>&nbsp; <code>
<del>pathstring</del></code>.</p>
<p><i>Returns:</i> <code>*this</code>.</p>
  </blockquote>
<p><i><span style="background-color: #E1E9FD">[SG note: this proposed wording depends on LWG 2732 
and the recently proposed definition of has_filename() and will change again to 
address LWG 2664 and US 77/CA 6. It already, simply by ignoring p entirely until 
such time, addresses US 51.] </span></i></p>
</blockquote>
<p><i>Change [path.modifiers]:</i></p>
<blockquote>
  <pre>path&amp; make_preferred();</pre>
  <blockquote>
  <p><i>Effects:</i> Each <i>directory-separator</i> is converted to
  <i>preferred-separator</i> in the <ins>generic format pathname</ins>. </p>
  </blockquote>
</blockquote>
<p><i><span style="background-color: #E1E9FD">[</span><span style="background-color: #E1E9FD">SG</span><span style="background-color: #E1E9FD"> 
note: the existing definition of </span><span style="background-color: #E1E9FD">
remove_filename</span><span style="background-color: #E1E9FD">() is broken; it 
is expected to be rewritten.] </span></i></p>
<p><i>Change [path.modifiers]:</i></p>
<blockquote>
  <pre>path&amp; replace_extension(const path&amp; replacement = path());</pre>
  <blockquote>
  <p><i>Effects:</i> <del><code>pathstring</code> (the stored path) is modified as follows:</del></p>
  <blockquote>
<p>— Any existing <code>extension()</code>(27.10.8.4.9) is removed from the <ins>generic format pathname</ins>
<del>stored path</del>, then</p>
<p>— If <code>replacement</code> is not empty and does not begin with a <i>dot</i> character, a 
<i>dot</i> 
character is appended to the <ins>generic format pathname</ins> <del>stored path</del>, then</p>
<p>— <ins><code>operator+=(replacement);</code></ins> <code><del>replacement</del></code><del> is concatenated to the stored path</del>.</p>
  </blockquote>
<p><i>Returns</i> <code>*this</code>.</p>
  </blockquote>
</blockquote>
<p><i>Change [path.modifiers]:</i></p>
<blockquote>
  <pre>void swap(path&amp; rhs) noexcept;</pre>
  <blockquote>
  <p><i>Effects:</i> Swaps the contents <ins>(in all formats)</ins> of the two paths
  <code><del>pathstring</del></code><del> and </del><code><del>rhs.pathstring</del></code>. </p>
  </blockquote>
</blockquote>
<p><i>Change [path.native.obs]:</i></p>
<blockquote>
  <pre>const string_type&amp; native() const noexcept;</pre>
  <blockquote>
    <p><i>Returns: </i><ins>The native format pathname</ins> <code><del>
    pathstring</del></code>.</p>
  </blockquote>
  <pre>const value_type* c_str() const noexcept;</pre>
  <blockquote>
    <p><i>Returns: </i><ins>Equivalent to <code>native().c_str()</code></ins>
    <code><del>pathstring.c_str()</del></code>.</p>
  </blockquote>
  <pre>operator string_type() const;</pre>
  <blockquote>
    <p><i>Returns:</i> <ins><code>native()</code></ins> <code><del>pathstring</del></code>.</p>
<p>[ <i>Note:</i> Conversion to string_type is provided so that an object of 
class path can be given as an argument to existing standard library file stream 
constructors and open functions. —<i>end note</i> ]</p>
  </blockquote>
  <pre>template &lt;class EcharT, class traits = char_traits&lt;EcharT&gt;,
          class Allocator = allocator&lt;EcharT&gt;&gt;
  basic_string&lt;EcharT, traits, Allocator&gt;
    string(const Allocator&amp; a = Allocator()) const;</pre>
  <blockquote>
    <p><i>Returns:</i> <ins><code>native()</code></ins> <code><del>pathstring</del></code>.</p>
<p><i>Remarks:</i> All memory allocation, including for the return value, shall 
be performed by a. Conversion, if any, is specified by 27.10.8.2.</p>
  </blockquote>
</blockquote>
<p><i>Change [path.generic.obs]:</i></p>
<blockquote>
  <pre>template &lt;class EcharT, class traits = char_traits&lt;EcharT&gt;,
          class Allocator = allocator&lt;EcharT&gt;&gt;
  basic_string&lt;EcharT, traits, Allocator&gt;
    generic_string(const Allocator&amp; a = Allocator()) const;</pre>
  <blockquote>
<p><i>Returns: </i><ins>The generic format pathname</ins> <del><code>pathstring</code>, 
reformatted according to the generic pathname format (27.10.8.1)</del>.</p>
<p><i>Remarks:</i> All memory allocation, including for the return value, shall 
be performed by a. Conversion, if any, is specified by 27.10.8.2.</p>
  </blockquote>
  <pre>std::string generic_string() const;
std::wstring generic_wstring() const;
std::string generic_u8string() const;
std::u16string generic_u16string() const;
std::u32string generic_u32string() const;</pre>
  <blockquote>
<p><i>Returns:</i> <ins>The generic format pathname</ins> <del><code>pathstring</code>, 
reformatted according to the generic pathname format (27.10.8.1)</del>.</p>
<p><i>Remarks:</i> Conversion, if any, is specified by 27.10.8.2. The encoding 
of the string returned by <code>generic_u8string()</code> is always UTF-8.</p>
  </blockquote>
</blockquote>
<p><i>Change [path.decompose]:</i></p>
<blockquote>
  <pre>path root_name() const;</pre>
  <blockquote>
    <p><i>Returns:</i> <i>root-name</i>, if <ins>the generic format 
    pathname</ins> <del><code>pathstring</code></del> includes <i>root-name</i>, 
    otherwise <code>path()</code>. </p>
  </blockquote>
  <pre>path root_directory() const;</pre>
  <blockquote>
    <p><i>Returns:</i> <i>root-directory</i>, if <ins>the generic format 
    pathname</ins> <del><code>pathstring</code></del> includes <i>root-directory</i>, 
    otherwise <code>path()</code>.</p>
  </blockquote>
  <p>...</p>
  <pre>path relative_path() const;</pre>
  <blockquote>
    <p><i>Returns:</i> A path composed from <ins>the generic format pathname</ins>
    <del><code>pathstring</code></del>, if <code>!empty()</code>, beginning with 
    the first <i>filename</i> after <i>root-path</i>. Otherwise, <code>path()</code>.</p>
  </blockquote>
  <p>...</p>
  <pre>path filename() const;</pre>
  <blockquote>
    <p><i>Returns:</i> <ins><code>relative_path().</ins>empty() ? path() : *--end()</code>.</p>
    <p>[ <i>Example:</i></p>
    <blockquote>
      <pre>std::cout &lt;&lt; path(&quot;/foo/bar.txt&quot;).filename(); // outputs &quot;bar.txt&quot;
<ins>std::cout &lt;&lt; path(&quot;/foo/bar&quot;).filename();     // outputs &quot;bar&quot;</ins>
<ins>std::cout &lt;&lt; path(&quot;/foo/bar/&quot;).filename();    // outputs &quot;&quot;</ins>
std::cout &lt;&lt; path(&quot;/&quot;).filename();            // outputs &quot;<del>/</del>&quot;
<ins>std::cout &lt;&lt; path(&quot;//host&quot;).filename();       // outputs &quot;&quot;</ins>
std::cout &lt;&lt; path(&quot;.&quot;).filename();            // outputs &quot;.&quot;
std::cout &lt;&lt; path(&quot;..&quot;).filename();           // outputs &quot;..&quot;</pre>
    </blockquote>
    <p>—<i>end example</i> ]</p>
  </blockquote>
  <pre>path stem() const;</pre>
  <blockquote>
    <p><del><i>Returns:</i> if <code>filename()</code> contains a period but does not consist 
    solely of one or two periods, returns the substring of <code>filename()</code> starting 
    at its beginning and ending with the character before the last period. 
    Otherwise, returns <code>filename()</code>.</del></p>
  <p><ins><i>Returns:</i> Let <code>f</code> be the generic format pathname for 
  <code>filename()</code>. Returns a 
  path whose generic format pathname is </ins></p>
  <blockquote>
  <p><ins>- <code>f</code>, if it contains no periods other than a leading 
  period or consists solely of one or two periods;</ins>
  </p>
  <p><ins>- otherwise, the prefix of <code>f</code> ending before its last period.</ins> </p>
  </blockquote>
  <p><i><span style="background-color: #FFFF99">Beman Dawes inserted &quot;other than 
  a leading period&quot; to reflect LEWG guidance from Issaquah. That wording has not 
  been reviewed by the filesystem small group.</span></i></p>
<p>[ Example:</p>
    <blockquote>
      <pre>std::cout &lt;&lt; path(&quot;/foo/bar.txt&quot;).stem(); // outputs &quot;bar&quot;
path p = &quot;foo.bar.baz.tar&quot;;
for (; !p.extension().empty(); p = p.stem())
std::cout &lt;&lt; p.extension() &lt;&lt; ’\n’;
// outputs: .tar
// .baz
// .bar</pre>
    </blockquote>
    <p>—end example ]</p>
  </blockquote>
  <pre>path extension() const;</pre>
  <blockquote>
    <p><i>Returns: </i><ins>a path whose generic format pathname is the 
    suffix of the generic format pathname for <code>filename()</code> not included in 
    <code>stem()</code>.</ins> 
    <del>if <code>filename()</code> contains a period but does not consist solely of 
    one or two periods, returns the substring of <code>filename()</code> 
    starting at the rightmost period and for the remainder of the path. 
    Otherwise, returns an empty path object.</del></p>
    <p><i>Remarks:</i> Implementations are permitted to define additional 
    behavior for file systems which append additional elements to extensions, 
    such as alternate data streams or partitioned dataset names. </p>
    <p>[ <i>Example:</i></p>
    <pre>std::cout &lt;&lt; path(&quot;/foo/bar.txt&quot;).extension(); // outputs &quot;.txt&quot;
<ins>std::cout &lt;&lt; path(&quot;/foo/bar&quot;).extension();     // outputs &quot;&quot;</ins></pre>
    <p><i>—end example</i> ]</p>
    <p>[ <i>Note:</i> The period is included in the return value so that it is 
    possible to distinguish between no extension and an empty extension. Also 
    note that for a path <code>p</code>, <code>p.stem()+p.extension() == 
    p.filename()</code>.<i> —end note </i>]</p>
  </blockquote>
</blockquote>
<p><i>Change [path.query]:</i></p>
<blockquote>
  <pre>bool empty() const noexcept;</pre>
  <blockquote>
<p><i>Returns: </i><ins><code>true</code> if the generic format pathname is 
empty, else <code>false</code></ins><i> </i><del><code>pathstring.empty()</code></del>.</p>
  </blockquote>
<p>...</p>
  <pre>bool is_absolute() const;</pre>
  <blockquote>
<p><i>Returns:</i> <code>true</code> if <ins>the native format pathname</ins>
<code><del>pathstring</del></code> contains an absolute path (27.10.4.1), else 
false.</p>
<p>[ Example: <code>path(&quot;/&quot;).is_absolute()</code> is true for POSIX-based 
operating systems, and false for Windows-based operating systems. —end example ]</p>
  </blockquote>
</blockquote>
<p><i>Change [<a name="path.gen">path.gen</a>]:</i></p>
<blockquote>
  <pre>path lexically_normal() const;</pre>
  <blockquote>
<p><i>Returns: </i><ins>a path whose generic format pathname is the normal form 
([fs.def.normal.form]) of the generic format pathname of <code>*this</code></ins><i>
</i><del><code>*this</code> in normal form ([fs.def.normal.form])</del>.</p>

<p>[ <i>Example:</i><br>
<code>&nbsp;&nbsp;&nbsp; assert(path(&quot;foo/./bar/..&quot;).lexically_normal() == &quot;foo&quot;);<br>
&nbsp;&nbsp;&nbsp; assert(path(&quot;foo/.///bar/../&quot;).lexically_normal() == &quot;foo/<del>.</del>&quot;);<br>
</code><br>
The above assertions will succeed. <del>The second example ends with a current 
directory (dot) element appended to support operating systems that use different 
syntax for directory names and regular file names.</del></p>

<p>On Windows, the returned path’s directory-separator characters will be 
backslashes rather than slashes, but that does not affect path equality. —<i>end 
example</i> ]</p>

  </blockquote>
</blockquote>
<p><i>Change [path.itr]:</i></p>
<blockquote>
<p>Path iterators iterate over the elements of <ins>the generic format pathname</ins>
<del><code>pathstring</code> in the generic format</del> (27.10.8.1).</p>
<p>A <code>path::iterator</code> is a constant iterator satisfying all the 
requirements of a bidirectional iterator (24.2.6) except that, for 
dereferenceable iterators <code>a</code> and <code>b</code> of type <code>
path::iterator</code> with <code>a == b</code>, there is no requirement that
<code>*a</code> and <code>*b</code> are bound to the same object. Its <code>
value_type</code> is <code>path</code>.</p>
<p>Calling any non-const member function of a path object invalidates all 
iterators referring to elements of that object.</p>
<p>For the elements of <ins>the generic format pathname</ins> <del><code>
pathstring</code> in the generic format</del>, the forward traversal order is as 
follows:</p>
<p>...</p>
</blockquote>
<p><i>Change [fs.op.canonical]:</i></p>
<blockquote>
  <pre>path canonical(const path&amp; p, error_code&amp; ec);
path canonical(const path&amp; p, const path&amp; base, error_code&amp; ec);</pre>
  <blockquote>
    <p><i>Effects:</i> Converts <code>p</code>, which must exist, to an absolute 
    path that has no symbolic link, <ins><i>dot</i>, or <i>dot-dot</i> elements 
    in its generic format pathname</ins> <del>&quot;.&quot;, or &quot;..&quot; elements</del>.</p>
    <p>...</p>
  </blockquote>
</blockquote>
<p><i>Change [fs.op.current_path]:</i></p>
<blockquote>
  <pre>path current_path();
path current_path(error_code&amp; ec);</pre>
  <blockquote>
<p><i>Returns:</i> The absolute path of the current working directory, obtained
<ins>in the native format</ins> as if by POSIX <code>getcwd()</code>. The 
signature with argument <code>ec</code> returns <code>path()</code> if an error 
occurs.</p>
<p>...</p>
  </blockquote>
<p><span style="font-style: italic; background-color: #E1E9FD">Rationale: </span>
</p>
<p><span style="font-style: italic; background-color: #E1E9FD">The lexical 
operation specifications leave some room for variation among implementations, 
but not as much as may be required for native formats; further relaxation would 
compromise the utility of the operations for portable, predictable path 
manipulation. Instead, we specify a usable but flexible mapping between the two 
formats and define each operation in terms of one (or, occasionally, both).
</span></p>
<p><span style="font-style: italic; background-color: #E1E9FD">The second bullet 
item in the first note in [path.fmt.cvt], which says for POSIX- or Windows-based 
operating systems &quot;There is no need to distinguish between native format and 
generic format in function arguments.&quot;, may not be true on Windows even if the 
other two bullet items are true.</span></p>

</blockquote>

<h4><a name="US75">US 75</a>, CA 4; 27.10.8.4.1 [path.construct]: Extra flag in path constructors is needed<br>
Status: Accept with modifications</h4>

<blockquote>

<p>See P0430R1 
Section 2.2 for wording.</p>

</blockquote>

<h4><a name="US76">US 76</a>, CA 5; 27.10.8.1 [path.generic]: <i>root-name</i> 
definition is over-specified<br>
Status: Accept with modifications</h4>

<blockquote>

<p>See P0430R1 
Section 2.3.1 for wording.</p>

</blockquote>

<h4><a name="US77">US 77</a>, CA 6; 27.10.8.4.3 [path.append]: <code>operator/</code> and other appends not useful if arg has <i>root-name</i><br>
Status: Accept with modifications</h4>

<blockquote>

<p>See P0430R1 Section 2.3.2 for wording.</p>

</blockquote>

<h4><a name="US78">US 78</a>, CA 7; 27.10.15.1 [fs.op.absolute]: Member <code>absolute()</code> in 27.10.4.1 is overspecified for non-POSIX-like O/S<br>
Status: Accept with modifications</h4>

<blockquote>

<p>See P0430R1 
Section 2.4.1 for wording.</p>

</blockquote>

<h4><a name="US79">US 79</a>, CA 8; Several: Some  operation functions are overspecified for implementation-defined file types<br>
Status: Accept with modifications</h4>

<blockquote>

<p>See P0430R1 
Section 2.4.2 for wording.</p>

</blockquote>



<h2><a name="editorial"></a>NB  comments submitted as editorial</h2>





<p>There was concern that some of these might involve substantive changes, so 
the SG processed them as ordinary technical comments.</p>





<h4><a name="US38">US 38</a>; 27.10.4.13 [fs.def.symlink]:&nbsp; Duplicates 
§17.3.16<br>
Status: Accept</h4>



<blockquote>

<p> Project editor will make change.</p>

</blockquote>



<h4><a name="US39">US 39</a>; 27.10.4.15 [fs.def.parent]: Remove note: Dot and 
dot-dot are not directories<br>
Status: Accept</h4>



<blockquote>

<p> Project editor will make change.</p>

</blockquote>



<h4><a name="US41">US 41</a>; 27.10.4.16 [fs.def.parent.other]: The term “parent 
directory” for a (non-directory) file is unusual<br>
Status: Reject</h4>


<blockquote>
<p>Insufficient motivation for change. &quot;parent directory&quot; is the term we use to 
describe the containing directory; we&#39;re (already) following POSIX there, even 
though the POSIX definition is a bit weird.</p>
</blockquote>

<h4><a name="US42">US 42</a>; 27.10.4.21 [fs.def.symlink]: Pathname resolution 
does not always resolve a symlink<br>
Status: Reject</h4>



<blockquote>

<p>There is an apparent corner case contradiction between the POSIX definition and other 
places in the POSIX standard. The [fs.def.symlink] definition is identical to 
the POSIX 3.381 Symbolic Link definition () and the SG prefers to keep it that 
way.</p>

</blockquote>



<h4><a name="US47">US 47</a>; 27.10.8.1 [path.generic]: “.” and “..” already 
match the name production<br>
Status: Accept with modifications </h4>



<blockquote>

<p> <span style="font-style: italic; background-color: #FFFF99">This is not editorial, but we know the fix. Davis Herring 
to supply updated proposed wording.</span></p>

</blockquote>



<h4><a name="US50">US 50</a>;&nbsp; 27.10.8.3 [path.req]: <code>path</code> 
requirements ¶ 1.4 largely redundant with ¶ 1.3<br>
Status: Reject</h4>


<blockquote>

<p><i><span style="background-color: #FFFF99">Beman Dawes to open an issue to expose this concern to additional experts. 
This is not a priority issue for shipping C++17.</span></i></p>

</blockquote>



<h2><a name="late-comments"></a>Late comments from P0489R0</h2>



<p>Quoting P0489R0, &quot;These comments were not submitted in time to 
be registered as National Body Comments, but should be considered as possible 
issues against SC 22, N3151, ISO/IEC CD 14882.&quot; </p>




</font>




<p><span style="background-color: #E1E9FD; font-style:italic">Late comments 15-47 apply to 27.10 [filesystems]. 
Except for as noted below, the submitter of these late comments has agreed to 
open LWG issues for any late comments that are still a concern.</span></p>

<font FACE="LMRoman10-Regular">

<h4><a name="Late15">Late 15</a>; 27.10.8.4.11 ¶3.1: A “mismatched element” cannot be equal to an iterator</h4>

<h4><a name="Late16">Late 16</a>; 27.10.8.4.11 ¶3.3.1: <code>lexically_relative("..\\foo")</code> produces nonsense.</h4>

<h4><a name="Late17">Late 17</a>; 27.10.8.4.11 ¶3.3.2: Behavior not always 
appropriate</h4>

<h4><a name="Late18">Late 18</a>; 27.10.8.4.11 ¶4: Behavior wrong in case corresponding to  comment Late17</h4>

<h4><a name="Late19">Late 19</a>; 27.10.8.5 ¶2: It is surprising that a path is a container of paths</h4>

<blockquote>

<p>Recommend NAD. See <a href="#US55">US 55</a>, <a href="#US59">US 59</a>.</p>

</blockquote>

<h4><a name="Late20">Late 20</a>; 27.10.11 ¶1: Anemic status structure</h4>

<h4><a name="Late21">Late 21</a>; 27.10.12: <code>directory_entry</code> is just a trivial wrapper for <code>path</code></h4>

<blockquote>

<p>Subsumed by LWG <a href="http://cplusplus.github.io/LWG/lwg-active.html#2663">
2663</a> and <a href="http://cplusplus.github.io/LWG/lwg-active.html#2677">2677</a></p>

</blockquote>

<h4><a name="Late22">Late 22</a>; 27.10.13 ¶6: <code>directory_iterator</code> cumbersome to assemble full path with iterator’s directory</h4>

<h4><a name="Late23">Late 23</a>; 27.10.14.1 ¶28: <code>disable_recursion_pending()</code> name is ugly implementation detail</h4>

<h4><a name="Late24">Late 24</a>; 27.10.15.1 ¶1: <code>absolute()</code> can produce nonsense result</h4>

<h4><a name="Late25">Late 25</a>; 27.10.15.4 ¶4.2: What attributes does <code>copy_file()</code> copy?</h4>

<h4><a name="Late26">Late 26</a>; 27.10.15.6 ¶5: <code>create_directories()</code> complexity assumes syscalls take constant time</h4>

<h4><a name="Late27">Late 27</a>; 27.10.15.7 ¶5: <code>create_directory()</code> specification prevents sensible security measures</code></h4>

<h4><a name="Late28">Late 28</a>; 27.10.15.10: <code>create_symlink()</code> might misbehave if <code>to</code> is a directory</h4>

<h4><a name="Late29">Late 29</a>; 27.10.15.13: Access to file ID is superior to <code>equivalent()</code></h4>

<h4><a name="Late30">Late 30</a>; 27.10.15.13 ¶3: <code>equivalent()</code>'s <code>s1==s2</code> check is ill-formed and could race</h4>

<h4><a name="Late31">Late 31</a>; 27.10.15.13 ¶5: Surprising that <code>equivalent()</code>'s error_code overload 
can throw</h4>

<h4><a name="Late32">Late 32</a>; 27.10.15.13 ¶5 : <code>equivalent()</code> should not reject special files outright</h4>

<h4><a name="Late33">Late 33</a>; 27.10.15.21: <code>is_other()</code> result surprising</h4>

<h4><a name="Late34">Late 34</a>; 27.10.15.25 ¶5: Could <code>last_write_time()</code> guarantee the error direction</code></h4>

<blockquote>

<p>Recommend submitting a paper that researches whether this is in fact 
possible. If it is possible, providing proposed wording would be helpful.</p>

</blockquote>

<h4><a name="Late35">Late 35</a>; 27.10.15.26: Why is there no way to read permissions?</h4>

<blockquote>

<p>Recommend NAD. They are available in <code>file_status</code>. If anyone 
thinks a separate function would be useful, they should submit an issue or a 
paper.</p>

</blockquote>

<h4><a name="Late36">Late 36</a>; 27.10.15.26: The <code>permissions() error_code</code> overload should be <code>noexcept</code><br>
Status: Accept</h4>

<blockquote>

<p><i>Change 27.10.15.26 [fs.op.permissions] and 27.10.6 Header &lt;filesystem&gt; 
synopsis [fs.filesystem.syn]:</i></p>

  <blockquote>

<p><code>void permissions(const path&amp; p, perms prms, error_code&amp; ec) <ins>noexcept</ins>;</code></p>

  </blockquote>

</blockquote>

<h4><a name="Late37">Late 37</a>; 27.10.15.26 ¶2:<code>permissions()</code> actions should be separate parameter<br>
Status: Accept with modifications</h4>


<blockquote>

<p><span style="font-style: italic; background-color: #FFFF99">Beman Dawes 
supplied the wording, which has  been reviewed by Davis but not others. Beman comments: 
Matt Austern pointed this out, with the same suggested fix, prior to the TS. It 
got dropped on the floor in the TS shipping rush and has never been fixed. It is 
an embarrassment that needs fixing before shipping C++17.</span></p>  
<p><i>Change [fs.filesystem.syn]:</i></p>
<blockquote>
  <pre>// 27.10.10, enumerations
enum class file_type;
enum class perms;
<ins>enum class perm_options;</ins>
enum class copy_options;
enum class directory_options;</pre>
<p>...</p>
  <pre>void permissions(const path&amp; p, perms prms<ins>, perm_options opts=perm_options::replace</ins>);
void permissions(const path&amp; p, perms prms, error_code&amp; ec);
<ins>void permissions(const path&amp; p, perms prms, perm_options opts, error_code&amp; ec);</ins></pre>
</blockquote>

  <p dir="ltr"><i>Strike the following rows in [fs.enum] Table nnn — Enum class
  <code>perms</code>:</i></p>
<ul>
  <li><i><code>add_perms</code></i></li>
  <li><i><code>remove_perms</code></i></li>
  <li><i><code>symlink_nofollow</code></i></li>
</ul>

  <p><i>Insert a new sub-section after [enum.perms]:</i></p>
<blockquote>

  <p><b><i><ins>27.10.n.n </i>Enum class <code>perm_options </code>[enum.perm_options]</b></ins></p>

  <p><ins>The <code>enum class</code> type <code>perm_options</code> is 
  a bitmask type (bitmask.types) that specifies bitmask constants used to 
  control the semantics of permissions operations.</ins></p>
  <p align="center"><ins>Table <i>nnn</i> &mdash; Enum class <code>perm_options</code></ins></p>
  <table border="1" cellpadding="5" cellspacing="0" style="border-collapse: collapse" bordercolor="#111111">
    <tr>
      <td bgcolor="#CCFFCC"><b>Name</b></td>
      <td bgcolor="#CCFFCC"><b>Meaning</b></td>
    </tr>
    <tr>
      <td bgcolor="#CCFFCC"><code>replace</code></td>
      <td bgcolor="#CCFFCC">(Default) <code>permissions()</code> shall replace the file's  permission bits with the <code>perm</code> argument’s permission bits.</td>
    </tr>
    <tr>
      <td bgcolor="#CCFFCC"><code>add</code></td>
      <td bgcolor="#CCFFCC">
      <p dir="ltr"><code>permissions()</code> shall bitwise <i>or</i> the <code>
      perm</code> argument’s permission bits to the file’s current permission 
      bits.</td>
    </tr>
    <tr>
      <td bgcolor="#CCFFCC"><code>remove</code></td>
      <td bgcolor="#CCFFCC">
      <p dir="ltr"><code>permissions()</code> shall bitwise <i>and</i> the 
      complement of the <code>perm</code> argument’s permission bits to the 
      file’s current permission bits.</td>
    </tr>
    <tr>
      <td bgcolor="#CCFFCC"><code>nofollow</code></td>
      <td bgcolor="#CCFFCC"><code>permissions()</code> shall change the permissions of a symbolic 
      link itself rather than the permissions of the file the link resolves to. <code>nofollow</code> 
      is a bit-mask value.</td>
    </tr>
  </table>
</blockquote>

<p><i>Change [fs.op.permissions]:</i></p>
<blockquote>
<pre>void permissions(const path&amp; p, perms prms<ins>, perm_options opts=perm_opts::replace</ins>);
void permissions(const path&amp; p, perms prms, error_code&amp; ec);
<ins>void permissions(const path&amp; p, perms prms, perm_options opts, error_code&amp; ec);</ins></pre>
<blockquote>
  <p><i>Requires:</i> <del><code>!((prms &amp; perms::add_perms) != perms::none &amp;&amp; (prms 
  &amp; perms::remove_perms) != perms::none)</code>.</del> <ins>One and 
  only one of the <code>perm_options</code> constants <code>replace</code>,
  <code>add</code>, or <code>remove</code> is present in <code>opts</code>.</ins></p>
  <p><ins><i>Remarks:</i> The second signature behaves as if it had an additional 
  argument&nbsp; <code>perm_options opts</code> with a value of <code>
  perm_options::replace</code>.</ins></p>
  <p><del><i>Effects:</i> Applies the effective permissions bits from <code>prms</code> 
  to the file <code>p</code> resolves to, or if that file is a symbolic link and
  <code>symlink_nofollow</code> is not set in <code>prms</code>, the file that 
  it points to, as if by POSIX <code>fchmodat()</code>. The effective permission 
  bits are determined as specified in Table 127, where <code>s</code> is the 
  result of <code>(prms &amp; perms::symlink_nofollow) != perms::none ? 
  symlink_status(p) : status(p)</code>.</del></p>
  <p><ins><i>Effects:</i> Applies the action specified by <code>opts</code> 
  to the file <code>p</code> resolves to, or to file <code>p</code> itself if <code>p</code> is a symbolic link and
  <code>perm_options::nofollow</code> is set in <code>opts</code>. The action is 
  applied as if by POSIX <code>fchmodat()</code>.</ins></p>
  <p>[ <i>Note:</i> Conceptually permissions are viewed as bits, but the actual 
  implementation may use some other mechanism. <i>—end note</i> ]</p>
  <p><i>Throws:</i> As specified in 27.10.7.</p>
</blockquote>

  </blockquote>

</blockquote>

<h4><a name="Late38">Late 38</a>; 27.10.15.26 ¶2: Unimplementable implication that <code>permissions()</code> is atomic</h4>

<h4><a name="Late39">Late 39</a>; 27.10.15.29 ¶3: Suprising <code>relative()</code> behavior for symlinks</h4>

<h4><a name="Late40">Late 40</a>; 27.10.15.31 ¶1: <code>remove_all()</code> unclear for <code>symlink/</code></h4>

<h4><a name="Late41">Late 41</a>; 27.10.15.31 ¶3: Would be useful for <code>remove_all()</code> to report successful removal count on error</h4>

<h4><a name="Late42">Late 42</a>; 27.10.15.33 ¶1: <code>resize_file()</code> <i>Postcondition</i> missing argument<br>
Status: Editorial</h4>

<blockquote>

<p>Fixed in the post-Issaquah working 
paper</p>

</blockquote>

<h4><a name="Late43">Late 43</a>; 27.10.15.33 ¶3: Add POSIX <code>ftruncate()</code> equivalent function</h4>

<blockquote>

<p>Recommend opening issue. Such a feature might be useful in 27.9 [file.streams] for C++2n.</p> 

</blockquote>

<h4><a name="Late44">Late 44</a>; 27.10.15.38: <code>system_complete()</code> name inconsistent with similar functions in subclause</h4>

<h4><a name="Late45">Late 45</a>; 27.10.15.38 ¶6: Remove <code>system_complete()</code> unless it can be made reliable</h4>

<h4><a name="Late46">Late 46</a>; 27.10.15.40 ¶2: Question about <code>weakly_canonical()</code> <i>Effects</i></code></h4>

<h4><a name="Late47">Late 47</a>; 27.10.15.40 ¶4: <code>weakly_canonical()</code> suggested caching is incorrect</h4>


<hr>

</body>

</html>