size_t cast in libssp

size_t cast in libssp

[angie]

After some casual analysis, I noticed that libssp (in gets-chk.c) we 
were doing:

 if (slen >= (size_t) INT_MAX)
    return gets (s);

  if (slen <= 8192)
    buf = alloca (slen + 1);
  else
    buf = malloc (slen + 1);
  if (buf == NULL)
    return gets (s);

What's the reason behind casting with size_t for INT_MAX, it seems to 
be a design
error, there is no guarantee that INT_MAX is a valid size_t. If we 
want to limit it
to size_t, we'd instead want a macro finding the minimum of INT_MAX 
and SIZE_MAX,
thereby rendering the whole cast unnecessary and potentially badly 
thought out.

Please get back to me at this address regarding this so that we may 
discuss future changes

Thanks!

Re: size_t cast in libssp

[Andreas Schwab]

<angie@auth.se> writes:

> What's the reason behind casting with size_t for INT_MAX, it seems to
> be a design
> error, there is no guarantee that INT_MAX is a valid size_t. If we
> want to limit it
> to size_t, we'd instead want a macro finding the minimum of INT_MAX
> and SIZE_MAX,

If INT_MAX > SIZE_MAX then (size_t)INT_MAX == SIZE_MAX, ie. this is the
minimum either way.

Andreas.

-- 
Andreas Schwab, schwab@redhat.com
GPG Key fingerprint = D4E8 DBE3 3813 BB5D FA84  5EC7 45C6 250E 6F00 984E
"And now for something completely different."