Re: gc-pbkdf2-sha1

[Paul Eggert]

Simon Josefsson <address@hidden> writes:

> Bruno Haible <address@hidden> writes:
>
>> Simon Josefsson wrote:
>>
>>> +  l = dkLen / hLen;
>>> +  if (dkLen % hLen)
>>> +    l++;
>>
>> An equivalent but faster code is:
>>
>>      l = ((dkLen - 1) / hLen) + 1;
>
> Perhaps for clarity we could change it into:
>
> #define CEIL_DIV(a,b) (((a) - 1) / (b)) + 1
>   l = CEIL_DIV (dkLen, hLen);
>
> What do you think?

That implementation of CEIL_DIV doesn't work in general, e.g., if
dkLen has the minimum value for its type.  So I wouldn't use it, at
least not without a comment saying that it's not safe in general.

Bruno's rewrite is safe because dkLen cannot possibly be zero here.