[Swan-dev] atoi -- just say no

[D. Hugh Redelmeier]

| From: Matt Rogers <mrogers at redhat.com>

| On 05/17, D. Hugh Redelmeier wrote:
| > Through this process, I'm coming to think that a ttoul-like function that 
| > also does range-checking would be worthwhile.  It is just too easy to be 
| > lazy about range checks.
| 
| It could just be a wrapper for ttoul, and that could replace the manual checks
| in plutomain, etc..

Yes.

| Although I did notice:
| 
|                 case 'x':       /* --crlcheckinterval <time>*/
|                         ugh = ttoul(optarg, 0, 10, &u);
|                         if (ugh != NULL)
|                                 break;
|                         crl_check_interval = u;
|                         continue;
| 
| This is one that is not range checked. What would be a good ceiling for crl fetch intervals?

I don't know any, so I didn't put one in.

But we don't have a way of expressing the maximum value that can be
stored in a time_t (and hence a deltatime_t).  That is surely an upper
bound that we should be checking because it is going to be less than
ULONG_MAX.

I guess we should invent one.  Perhaps this will do:

#define TIME_T_MAX ((time_t) (1ul << (sizeof(time_t) * BITS_PER_BYTE - 1) - 1))

One imperfection is that some imaginary C implementation (Burroughs
5000 and successors?  CDC 6600 and successors?) might have bits in an
integral type that don't participate in the value -- that's legal in
C (or was when I last paid attention).