[Bug c++/12200] New: Nonintuitive behavior with __attribute__((packed))

["pgonzalez at bluel dot com" <gcc-bugzilla at gcc dot gnu dot org>]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12200

           Summary: Nonintuitive behavior with __attribute__((packed))
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: pgonzalez at bluel dot com
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: arm-arm-elf

In that attached code sample, the __attribute__((packed))
modifier does more than just aligning the fields within the 
struct.  It changes the overall alignment of the variable 
itself, so that "e" (and hence "e.d") is no longer 32-bit
aligned.  The problem goes away if the "a=1" line is 
commented out, since this is what is offsetting it in memory.

I think this is a bug, since (a) I believe most programmers would
expect "e.d" to be aligned correctly, (b) the attribute is being
applied to a type (not a variable), and (c) I can't think of any
useful application for this behavior.

______________

extern const short a = 1;

struct TPacked {
  short b;
  short c;
  int d;
} __attribute__((packed));

extern const TPacked e = { 2, 3, 4 };