<html>
<head>
<base href="https://bugs.llvm.org/">
</head>
<body><table border="1" cellspacing="0" cellpadding="8">
<tr>
<th>Bug ID</th>
<td><a class="bz_bug_link
bz_status_NEW "
title="NEW - pointer_traits for __wrap_iter types sometimes produces nonsense"
href="https://bugs.llvm.org/show_bug.cgi?id=32006">32006</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>pointer_traits for __wrap_iter types sometimes produces nonsense
</td>
</tr>
<tr>
<th>Product</th>
<td>libc++
</td>
</tr>
<tr>
<th>Version</th>
<td>4.0
</td>
</tr>
<tr>
<th>Hardware</th>
<td>All
</td>
</tr>
<tr>
<th>OS</th>
<td>All
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>enhancement
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>All Bugs
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedclangbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>arthur.j.odwyer@gmail.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org, mclow.lists@gmail.com
</td>
</tr></table>
<p>
<div>
<pre>This is more of an enhancement request than a bug, I think, because I don't see
anywhere that the Standard actually specifies what counts as a "pointer-like
type". However, here's the problem:
In general, it is true (by accident, not by specification) that
pointer_traits<T>::element_type ==
remove_reference_t<decltype(*declval<T>())>
But when you set T == std::vector<int>::iterator, you find that
pointer_traits<T>::element_type == T::pointer_type
instead.
<a href="http://melpon.org/wandbox/permlink/pscCG6PVumWexxeN">http://melpon.org/wandbox/permlink/pscCG6PVumWexxeN</a>
This is because for any type of the form X<Y>, the standard mandates that
pointer_traits<X<Y>>::element_type == Y. So when you feed in
std::vector<int>::iterator a.k.a. __wrap_iter<int*>, element_type comes out as
(int*).
Additionally, this nerfs the libc++ extension function "__to_raw_pointer" for
contiguous iterators.
int *p = std::__to_raw_pointer(myvector.end());
does not compile.
<a href="http://melpon.org/wandbox/permlink/LL61wlsNYdEtQX1X">http://melpon.org/wandbox/permlink/LL61wlsNYdEtQX1X</a>
This could be resolved by
- closing as NOTABUG since iterator types are not "pointer-like"; or
- adding an element_type typedef to libc++'s __wrap_iter; or
- adding a partial specialization for pointer_traits<__wrap_iter<Y>>.
The discussion that led to this "discovery" was
<a href="https://groups.google.com/a/isocpp.org/forum/#!topic/std-proposals/PBETQcQHB-0">https://groups.google.com/a/isocpp.org/forum/#!topic/std-proposals/PBETQcQHB-0</a></pre>
</div>
</p>
<hr>
<span>You are receiving this mail because:</span>
<ul>
<li>You are on the CC list for the bug.</li>
</ul>
</body>
</html>