Stephen Hemminger <step...@networkplumber.org> writes:
> On Mon, 30 Nov 2020 23:59:39 +0100
> Petr Machata <m...@pmachata.org> wrote:
>
>> +char *sprint_size(__u32 sz, char *buf)
>> +{
>> + size_t len = SPRINT_BSIZE - 1;
>> + double tmp = sz;
>> +
>> + if (sz >= 1024*1024 && fabs(1024*1024*rint(tmp/(1024*1024)) - sz) <
>> 1024)
>> + snprintf(buf, len, "%gMb", rint(tmp/(1024*1024)));
>> + else if (sz >= 1024 && fabs(1024*rint(tmp/1024) - sz) < 16)
>> + snprintf(buf, len, "%gKb", rint(tmp/1024));
>> + else
>> + snprintf(buf, len, "%ub", sz);
>> +
>> + return buf;
>> +}
>
> Add some whitespace here and maybe some constants like mb and kb?
Sure.
> Also, instead of magic SPRINT_BSIZE, why not take a len param (and
> name it snprint_size)?
Because keeping the interface like this makes it possible to reuse the
macroized bits in q_cake. I feel like the three current users are
auditable enough that the implied length is not a big deal. And no new
users should pop up, as the comment at the function makes clear.
> Yes when you copy/paste code it is good time to get it back to current
> style standards.
