NumericBox unsafe double conversion

[Zolnai Tamás]

2013/3/22 Stephan Bergmann <sbergman at redhat.com>

> On 03/22/2013 11:44 AM, Zolnai Tamás wrote:
>
>> Otherwise in ImplNumericGetValue() the problem of overflow/underflow can
>> be avoid with these lines
>> sal_Int64 nValue = aStr.toInt64()
>> if( OUString::valueOf(nValue) != aStr )
>> {
>>      if( bNegative )
>>          rValue = SAL_MIN_INT64;
>>      else
>>          rValue = SAL_MAX_INT64;
>>      return sal_True;
>> }
>>
>
> ...but only if aStr is known to be in canonical form, esp. neither
> contains minus zero nor contains excessive leading zero digits.  None of
> that is guaranteed for ImplNumericGetValue in its current form (and also
> consider nDecDigits > 0, where input rStr="0" and nDecDigits=2, say, would
> lead to aStr="000" and the above comparison erroneously failing).
>

Oh, that's true. Thanks. I extend the check to take notice of these two
things.
>From that matter, why is that the toInt64() function doesn't pay attention
on overflow/underflow? Maybe it would simpler to handle these "flows" in
that function.


Regards,
Tamás
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/libreoffice/attachments/20130322/6326c2e2/attachment.html>