On Thu, 20 Jan 2005 18:21:09 -0800, Seth Arnold <sarnold@private> wrote:
> On Fri, Jan 21, 2005 at 09:08:17AM +0800, Rogelio Serrano wrote:
> > Is there a way to have these kind of protection without hiding bugs.
> > Maybe the audit framework can help?
>
> "Hiding bugs" is _not_ what a non-executable stack or data segment does;
> if a program relies on executable stack or data segment for executing,
> the process will die a miserable death. And loudly, at that.
>
> If the program allows this behaviour accidently (say, bounds checking
> error such as the type that StackGuard can protect against) then the
> flaw will in fact be much _easier_ to spot with such a tool. (StackGuard
> has found a _lot_ of bugs in software that went unnoticed because the
> consequences weren't very dire, including a hilarious off-by-one array
> access in glibc's test suite.)
>
> No, in my experience, tools like stackguard, Solar's non-executable
> stack, and similar, do a great job _finding_ buggy code that would
> otherwise survive in the wild unnoticed for years...
>
Consider me a convert. That does make sense.
--
Blood is thicker than water... and much tastier
John Davidorff Pell
This archive was generated by hypermail 2.1.3 : Thu Jan 20 2005 - 18:31:17 PST