cvs commit: src/lib/libipsec pfkey_dump.c

[Jacques A. Vidrine]

On Thu, Apr 17, 2003 at 08:23:57AM +0000, Munechika Sumikawa wrote:
> nectar> These sizes seem arbitrary.  Previously, at 10 bytes, they were just a
> nectar> bit too small to hold the representation of the largest 32-bit
> nectar> unsigned integer.  Now at 20 bytes, they are just a bit too small to
> nectar> hold the representation of the largest 64-bit unsigned integer.  This
> nectar> seems sloppy.  (As does using a 128 byte buffer to hold the result of
> nectar> concatenating two 20 byte buffers.)
> 
> You are right.  I've changed to 128 bytes.  

Er, that wasn't quite my point :-)  My point was that the buffer should
be sized correctly (rather than fudged), such as
    char prefbuf[] = "/4294967296"; 
    /* OR */
    char prebuf[12];
or whatever you expect the maximum integer to be.  [Isn't this a
network prefix?  Surely 3 digits is sufficient?]
It just seems like a lot of wastage, but of course it isn't really
important.

> It can be used until we
> support 512bit CPU :-)

Actually 128 bytes is just big enough to hold a NUL-terminated string
representation of a 421-bit unsigned integer.  You need 156 bytes for
512 bits :-)

Also, these two 128 byte buffers are still being concatenated into a
single 128 byte buffer, so on average you have room for about 209 bits
:-)

Cheers,
-- 
Jacques A. Vidrine <nectar at celabo.org>          http://www.celabo.org/
NTT/Verio SME          .     FreeBSD UNIX     .       Heimdal Kerberos
jvidrine at verio.net     .  nectar at FreeBSD.org  .          nectar at kth.se