On Apr 26, 2011, at 6:23 AM, Guy Harris wrote:
>
> On Apr 25, 2011, at 3:01 PM, Michael Tüxen wrote:
>
>> dup_list = (guint32 *)((char *)sack_header + 16 + (nr * sizeof(struct gaps)))
>
> Clang likes something that doesn't involve casting a "char *", which is not guaranteed to contain an address that's 4-byte aligned, to a "guint32 *", which is a pointer that's supposed to be 4-byte aligned...
>
> ...and that, on at least some architectures, *MUST* be 4-byte aligned in order for dereferences of it to work!
The fields in the packet are 4 byte aligned. But if the whole structure might not...
So should we copy them?
Best regards
Michael
>
> One such architecture is SPARC, which traps on unaligned references. I think at least some versions of ARM are other such architectures; I don't think it traps on unaligned accesses, but it might, for example, just act as if the lower 2 bits of the address are 0.
>
> Unless sack_header is *guaranteed* to be aligned on a 4-byte boundary, that code is unsafe.
> ___________________________________________________________________________
> Sent via: Wireshark-dev mailing list <wireshark-dev@xxxxxxxxxxxxx>
> Archives: http://www.wireshark.org/lists/wireshark-dev
> Unsubscribe: https://wireshark.org/mailman/options/wireshark-dev
> mailto:wireshark-dev-request@xxxxxxxxxxxxx?subject=unsubscribe
>