On 04/03/2013 09:06 AM, Daniel P. Berrange wrote:
From: "Daniel P. Berrange" <berrange@redhat.com>
To avoid
virportallocatortest.c: In function 'bind': virportallocatortest.c:34:33: warning: cast increases required alignment of target type [-Wcast-align] struct sockaddr_in *saddr = (struct sockaddr_in *)addr; ^
Signed-off-by: Daniel P. Berrange <berrange@redhat.com> --- tests/virportallocatortest.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/tests/virportallocatortest.c b/tests/virportallocatortest.c index 9931e11..1935602 100644 --- a/tests/virportallocatortest.c +++ b/tests/virportallocatortest.c @@ -31,12 +31,14 @@ int bind(int sockfd ATTRIBUTE_UNUSED, const struct sockaddr *addr, socklen_t addrlen ATTRIBUTE_UNUSED) { - struct sockaddr_in *saddr = (struct sockaddr_in *)addr; + struct sockaddr_in saddr;
Hmm. POSIX requires that apps use sockaddr_storage to guarantee alignment before downcasting to struct sockaddr_in or further down to struct sockaddr. But since we are providing a replacement bind(), and POSIX doesn't give us any way to specify that 'struct sockaddr' in our signature should have been aligned as struct sockaddr_storage by any sane caller, we're stuck playing realignment games, or else using a pragma.
- if (saddr->sin_port == htons(5900) || - saddr->sin_port == htons(5904) || - saddr->sin_port == htons(5905) || - saddr->sin_port == htons(5906)) { + memcpy(&saddr, addr, sizeof(saddr));
Since this is just test code, performance doesn't matter, and using memcpy() for the realignment game makes sense. ACK. -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org