<html>
<head>
<base href="https://llvm.org/bugs/" />
</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 --- - incomplete support for -fdata-sections in clang/llvm for C++ static data members"
href="https://llvm.org/bugs/show_bug.cgi?id=26570">26570</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>incomplete support for -fdata-sections in clang/llvm for C++ static data members
</td>
</tr>
<tr>
<th>Product</th>
<td>libraries
</td>
</tr>
<tr>
<th>Version</th>
<td>trunk
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>Linux
</td>
</tr>
<tr>
<th>Status</th>
<td>NEW
</td>
</tr>
<tr>
<th>Severity</th>
<td>normal
</td>
</tr>
<tr>
<th>Priority</th>
<td>P
</td>
</tr>
<tr>
<th>Component</th>
<td>Common Code Generator Code
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>thanm@google.com
</td>
</tr>
<tr>
<th>CC</th>
<td>llvm-bugs@lists.llvm.org
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>Created <span class=""><a href="attachment.cgi?id=15875" name="attach_15875" title="test case">attachment 15875</a> <a href="attachment.cgi?id=15875&action=edit" title="test case">[details]</a></span>
test case
The attached test case and makefile illustrate a limitation in the current
clang/llvm implementation of the "-fdata-sections" command line option. This
option tells the compiler to place each named data item into a separate ELF
section in the resulting object file; doing this has benefits in that it helps
enable linker GC of unused sections and (if desired) linker reordering of
sections for performance/size reasons.
In the test case ("ex.cc") note the following two variables:
char foo::ID = 0;
char baz::ID = 0;
When you compile with -fdata-sections, these two vars are still placed into the
same ".bss" section, which effectively defeats the option.
To illustrate why this is a problem, download the testcase + makefile and run
"make doit", then diff the two files
ex.so.od.clang.txt ex.so.od.gcc.txt
For gcc (which doesn't make the same mistake), the linker is able to
garbage-collect the storage for "_ZN12_GLOBAL__N_13foo2IDE", whereas in the
LLVM case this dead storage can't be reclaimed (since the var in question is in
the same section with another variable that is not dead).
NB: there maybe other cases in which -fdata-sections also doesn't work (I
haven't done an exhaustive search). It does seem to work ok for C-style
variable declarations as far as I can tell.</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>