Re: [Qemu-devel] [PATCH 1/4] unicode: New mod_utf8_codepoint()

[Laszlo Ersek <lersek@redhat.com>]

On 03/22/13 10:23, Markus Armbruster wrote:
> Laszlo Ersek <lersek@redhat.com> writes:
>> On 03/14/13 18:49, Markus Armbruster wrote:

>>> +{
>>> +    static int min_cp[5] = { 0x80, 0x800, 0x10000, 0x200000, 0x4000000 };
>>> +    const unsigned char *p;
>>> +    unsigned byte, mask, len, i;
>>> +    int cp;
>>> +
>>> +    if (n == 0) {
>>> +        *end = (char *)s;
>>> +        return 0;
>>> +    }
>>
>> This is a special case (we return the code point U+0000 after looking at
>> zero bytes); we can probably expect the caller to know about n==0.
> 
> We could make it an error instead.  What's your gut feeling?

(If the question still stands -- maybe it doesn't any more, considering
future handling of '\0':) I guess this function would be called in a
loop, with increasing "s" and decreasing "n" values. "end" can only be
checked after the first call. If you write a loop that checks "end" in
the controlling expression, then accepting n==0 without error is useful.
If you write a loop that checks "n" in the controlling expression, then
refusing n==0 is OK. I'd probably write the latter kind of loop (I like
pre-testing more), but I can't say in general :)

Laszlo