Sent from my iPhone
On Oct 2, 2008, at 2:09 AM, "adam at os dot inf dot tu-dresden dot de"
<[EMAIL PROTECTED]> wrote:
I've been experiencing some (for me) rather unexpected behavior with
some C++
code. The following is a very reduced test case:
This comes down to non pod vs pod and tail filling.
class A
{ private: void *one;
} __attribute__((aligned(16)));
class C {};
class L { public: class M : A { C _c; }; M m; };
L x;
It's about the size of L::M. I.e. if this file is compiled, L::M has
a sizeof
of 16. If the first private is changed to public, the size of L::M
changes to
32 (-Dprivate=public). Is this expected behavior? When removing the
alignment
contraints, the size of L::M is 8 in both cases.
I'm seeing this behavior with all gcc versions tested (3.3, 3.4,
4.1, 4.2, 4.3,
head) as well as with different target architectures (i386, x86_64,
arm).
This is the script use to show the difference:
#! /bin/sh
echo 'print sizeof(L::M)' > x.cmd
echo -n "orig: "
gcc -Wall -g -c x.cc
gdb -x x.cmd -batch x.o
echo -n "with define: "
gcc -Dprivate=public -Wall -g -c x.cc
gdb -x x.cmd -batch x.o
Thanks
--
Summary: Different class sizes with public/private and
alignments
Product: gcc
Version: 4.3.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: adam at os dot inf dot tu-dresden dot de
GCC build triplet: i486-linux-gnu
GCC host triplet: i486-linux-gnu
GCC target triplet: i486-linux-gnu
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37715
