[m-rev.] write_float_special.exp4

[Julien Fischer]

On Thu, 21 Aug 2014, Paul Bone wrote:

> On Thu, Aug 21, 2014 at 12:53:56AM +1000, Julien Fischer wrote:
>>
>> Hi Zoltan,
>>
>> On Wed, 20 Aug 2014, Zoltan Somogyi wrote:
>>
>>> On my laptop, the output of the hardcoded/write_float_special test has long been:
>>>
>>> Inf: inf
>>> -Inf: -inf
>>> NaN: -nan
>>>
>>> The expected output does not have the - sign in front of the nan.
>>>
>>> I don't particularly care about the exact details of how we format non-number
>>> numbers, but having a test case always failing is annoying.
>>>
>>> I think either the MR_is_nan test should be moved earlier in MR_sprintf_float,
>>> to before the - is printed, or the output above should be added as tests/
>>> hardcoded/write_float_special.exp4, with corresponding .exp[56] for infinity/
>>> Infinity. What do others think?
>>
>> The former: the profileration of expected outputs for the above (and
>> related) test case(s) is already annoying.
>>
>
> I second this.
>
> It also has the potential to reduce problems when reading in Mercury terms
> with io.read, parsers won't need to understand -nan.  (Unless there's some
> other reason why they should.)

Such as them not understanding nan (as a floating point value) ;-)

Cheers,
Julien.