<html>
<head>
<base href="http://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 --- - Miscompiled code and corrupted stack frame"
href="http://llvm.org/bugs/show_bug.cgi?id=15947">15947</a>
</td>
</tr>
<tr>
<th>Summary</th>
<td>Miscompiled code and corrupted stack frame
</td>
</tr>
<tr>
<th>Product</th>
<td>clang
</td>
</tr>
<tr>
<th>Version</th>
<td>unspecified
</td>
</tr>
<tr>
<th>Hardware</th>
<td>PC
</td>
</tr>
<tr>
<th>OS</th>
<td>All
</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>-New Bugs
</td>
</tr>
<tr>
<th>Assignee</th>
<td>unassignedclangbugs@nondot.org
</td>
</tr>
<tr>
<th>Reporter</th>
<td>llvmbugs.20.adept@0sg.net
</td>
</tr>
<tr>
<th>CC</th>
<td>llvmbugs@cs.uiuc.edu
</td>
</tr>
<tr>
<th>Classification</th>
<td>Unclassified
</td>
</tr></table>
<p>
<div>
<pre>Apple LLVM version 4.2 (clang-425.0.27) (based on LLVM 3.2svn)
Target: x86_64-apple-darwin12.3.0
Thread model: posix
Hallo LLVM team,
I have encountered a miscompilation which corrupts the stack frame. In can be
reduced to this code:
struct Vector4
{
// When the following lines are uncommented, the codegen is correct.
// Vector4() = default;
// Vector4( const Vector4 &other ) : x( other.x ), y( other.y ), z( other.z
), w( other.w ) {};
float x, y, z, w;
};
inline
Vector4 operator *( const Vector4 &lhs, float rhs )
{
Vector4 result;
// Should be legal. Since a Vector4 contains floats, a float * can alias as
a Vector4 *.
for( unsigned i = 0; i < 4; ++i )
((float *)&result)[i] = ((const float *)&lhs)[i] * rhs;
return result;
}
void testmain( float x )
{
extern Vector4 X;
extern Vector4 A;
X = A * x;
}
The disassembly shows that in the for-loop in the multiplication operator
destroys the value of %rbp that was saved onto the stack frame. Whats more, in
basic block 2 only half of the members are returned in the result.
# Assembly output for codegen.cc
# Generated at 2:32:41 vorm. on Freitag, Mai 10, 2013
# Using Release configuration, x86_64 architecture for test target of libmbl
project
.section __TEXT,__text,regular,pure_instructions
.private_extern __Z8testmainf
.globl __Z8testmainf
__Z8testmainf: ## @_Z8testmainf
.cfi_startproc
## BB#0:
pushq %rbp
Ltmp2:
.cfi_def_cfa_offset 16
Ltmp3:
.cfi_offset %rbp, -16
movq %rsp, %rbp
Ltmp4:
.cfi_def_cfa_register %rbp
xorl %eax, %eax
movq _A@GOTPCREL(%rip), %rcx
LBB0_1: ## =>This Inner Loop Header: Depth=1
movss (%rcx,%rax,4), %xmm1
mulss %xmm0, %xmm1
movss %xmm1, -8(%rbp,%rax,4)
incq %rax
cmpl $4, %eax
jne LBB0_1
## BB#2: ## %_ZmlRK7Vector4f.exit
movq -8(%rbp), %rax
movq _X@GOTPCREL(%rip), %rcx
movq %rax, (%rcx)
movq $0, 8(%rcx)
popq %rbp
ret
.cfi_endproc
The following is the disassembly when the two constructors are uncommented. The
function then works as expected.
# Assembly output for codegen.cc
# Generated at 2:37:34 vorm. on Freitag, Mai 10, 2013
# Using Release configuration, x86_64 architecture for test target of libmbl
project
.section __TEXT,__text,regular,pure_instructions
.private_extern __Z8testmainf
.globl __Z8testmainf
__Z8testmainf: ## @_Z8testmainf
.cfi_startproc
## BB#0:
pushq %rbp
Ltmp2:
.cfi_def_cfa_offset 16
Ltmp3:
.cfi_offset %rbp, -16
movq %rsp, %rbp
Ltmp4:
.cfi_def_cfa_register %rbp
xorl %eax, %eax
movq _A@GOTPCREL(%rip), %rcx
LBB0_1: ## =>This Inner Loop Header: Depth=1
movss (%rcx,%rax,4), %xmm1
mulss %xmm0, %xmm1
movss %xmm1, -16(%rbp,%rax,4)
incq %rax
cmpl $4, %eax
jne LBB0_1
## BB#2: ## %_ZmlRK7Vector4f.exit
movq -16(%rbp), %rax
movq -8(%rbp), %rcx
movq _X@GOTPCREL(%rip), %rdx
movq %rcx, 8(%rdx)
movq %rax, (%rdx)
popq %rbp
ret
.cfi_endproc</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>