------- Comment #4 from pinskia at gmail dot com 2010-08-10 12:36 -------
Subject: Re: New: optimizer dereference
On Aug 10, 2010, at 1:00 AM, "attardi at di dot unipi dot it"
<gcc-bugzi...@gcc.gnu.org
> wrote:
> Code produced using -O2 handles dereferencing incorrectly.
> Here is a program that shows the bug:
>
> #include <iostream>
> #include <vector>
>
> class Derived : public std::vector<char *>
> {
> public:
> Derived() {}
> };
>
> void* foo(void* arg) {
> void* baseptr = 0;
> *(std::vector<char*> **)&baseptr = (Derived *)arg;
> return baseptr;
> }
Have it do
jlong a = 0;
memcpy(&a, &arg, sizeof(void*));
That is the most portable way of putting a pointer value in jlong.
>
> void* foo2(void* arg) {
> void* baseptr = 0;
> *(std::vector<char*> **)&baseptr = *(Derived **)&arg;
> return baseptr;
> }
>
> int main()
> {
> Derived* d = new Derived();
> void* upcast = foo(d);
> void* upcast2 = foo2(d);
> std::cerr << d << ", " << upcast << ", " << upcast2 << std::endl;
> }
>
> If you compile without -O2 the two values upcast and upcast2 are the
> same, as they should be.
> Using -O2 the results are different.
>
> The bug occurrs with both:
>
> g++ (GCC) 4.1.2 20070925 (Red Hat 4.1.2-27)
> g++ (Ubuntu 4.4.1-4ubuntu9) 4.4.1
>
> This kind of code is generated by SWIG for Java, so I can't change
> the source.
>
> Curiously the bug disappears if you put a print statement before the
> assignment to baseptr.
>
>
> --
> Summary: optimizer dereference
> Product: gcc
> Version: 4.4.1
> Status: UNCONFIRMED
> Severity: critical
> Priority: P3
> Component: c++
> AssignedTo: unassigned at gcc dot gnu dot org
> ReportedBy: attardi at di dot unipi dot it
> GCC build triplet: x86_64-unknown-linux-gnu
> GCC host triplet: x86_64-unknown-linux-gnu
> GCC target triplet: x86_64-unknown-linux-gnu
>
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45246
>
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45246
