This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH][BZ #16004] Check overlaps in memcpy_chk.
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: OndÅej BÃlka <neleai at seznam dot cz>
- Cc: <libc-alpha at sourceware dot org>
- Date: Mon, 21 Oct 2013 17:50:12 +0000
- Subject: Re: [PATCH][BZ #16004] Check overlaps in memcpy_chk.
- Authentication-results: sourceware.org; auth=none
- References: <20131005133423 dot GA8405 at domone> <Pine dot LNX dot 4 dot 64 dot 1310051925240 dot 9147 at digraph dot polyomino dot org dot uk> <20131021171406 dot GA27539 at domone dot podge>
On Mon, 21 Oct 2013, Ondrej Bilka wrote:
> On Sat, Oct 05, 2013 at 07:28:27PM +0000, Joseph S. Myers wrote:
> > There are tests for *_chk functions in debug/. If you add a new check you
> > should also add testcases for it to confirm that the error is caught at
> > runtime. This does of course require also adding the same check to the
> > generic version of the function, and in the case of memcpy_chk to the x86
> > (32-bit) versions as well.
> >
> It also requires to edit compile time check. I spend hour trying to
> write testcase and wondering where is error until I realized that it got
> optimized away at compile time.
I'm not clear what compile-time check you're referring to. If it's
something in GCC, it's certainly reasonable to build testcases with
-fno-builtin-<function>. But really, if the error can be detected at
compile time, GCC should be able to do so - and for examples that can't be
detected at compile time, you can always use "volatile" to ensure the
compiler has no information about the actual values of function arguments.
--
Joseph S. Myers
joseph@codesourcery.com