[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20120625084132.GA9406@openwall.com>
Date: Mon, 25 Jun 2012 12:41:32 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: compiler warnings: make clean linux-x86-clang > /dev/null
On Mon, Jun 25, 2012 at 10:26:22AM +0200, Frank Dittrich wrote:
> But some definitely look suspicious:
>
>
> single.c:330:8: warning: array index of '1' indexes past the end of an
> array (that contains 1 elements)
> [-Warray-bounds]
> if (first->data[1]) {
> ^ ~
> In file included from single.c:15:
> In file included from ./loader.h:16:
> ./list.h:19:7: note: array 'data' declared here
> char data[1];
> ^
This is as intended. "struct list_entry" is allocated dynamically, with
just sufficient space to hold the actual "data". An alternative would
have been to have extra pointer indirection here, which would waste a
little bit of memory on the pointer and be a little bit slower.
That said, I welcome suggestions on how to avoid scaring the compiler
with what looks to it like an out-of-bounds access. A real concern is
that some compiler might think that since the behavior is undefined
anyway, it can optimize this out, or maybe trap it as a runtime fault.
IIRC, formally we're only allowed to take the address of the array
element just beyond the limit (e.g., &data[1] would be valid), but not
dereference it.
Alexander
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
