This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug c++/25979] New: incorrect codegen for conditional
- From: "hhinnant at apple dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 26 Jan 2006 17:29:37 -0000
- Subject: [Bug c++/25979] New: incorrect codegen for conditional
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
I'm not positive whether or not this is a duplicate of 25895. I figured I'd
better enter it just in case it wasn't. Test case:
#include <stdio.h>
struct A
{
A() : data1_(0), data2_(0) {}
A(int i, int j) : data1_(i), data2_(j) {}
A operator+(int);
friend A operator+(int, const A&);
~A() {}
//private:
int data1_;
int data2_;
};
extern bool x;
void display(const A& x)
{
printf("%d %d\n", x.data1_, x.data2_);
}
int main()
{
A a1(1,2);
a1 = (x ? a1 + 3 : 3 + a1);
display(a1);
}
bool x = false;
A
A::operator+(int i)
{
A a;
a = *this;
a.data2_ = i;
return a;
}
A
operator+(int i, const A& x)
{
A a;
a = x;
a.data1_ = i;
return a;
}
Output:
3 0
Expected output:
3 2
The gimple tree (-fdump-tree-gimple) is showing statements like:
operator+ (&a1, 3, &a1) [return slot addr];
which shoud instead be:
operator+ (temp, 3, &a1) [return slot addr];
...
a1 = temp;
The bad codegen is sensitive to the presence or absence of special member
functions. For example if you comment out ~A(), you get the expected output.
--
Summary: incorrect codegen for conditional
Product: gcc
Version: 4.0.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: hhinnant at apple dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25979