Re: [PATCH] msvcr120: Implement strtof and _strtof_l (try 2)

[Henri Verbeet]

On 13 August 2015 at 02:23, Bernhard Übelacker  wrote:
> +static inline BOOL almost_equal(float d1, float d2) {
> - if(d1-d2>-1e-30 && d1-d2<1e-30)
> -
>    return TRUE;
> - return FALSE;
>
> +}
> -

You almost never want to compare floating point values like that. See
e.g. compare_float() in dlls/ddraw/tests/ddraw7.c for a better (if
perhaps not quite perfect) way. E.g. for the FLT_MAX case this only
works because you happen to get an exact match. "1e-30" is a double
instead of a float. Of course this is probably copied from the
existing tests, so I suppose I can't complain too much.

[PATCH] msvcr120: Implement strtof and _strtof_l (try 2)

[Piotr Caban]

On 08/13/15 11:08, Henri Verbeet wrote:
> On 13 August 2015 at 02:23, Bernhard Übelacker  wrote:
> > +static inline BOOL almost_equal(float d1, float d2) {
> > - if(d1-d2>-1e-30 && d1-d2<1e-30)
> > -
> >    return TRUE;
> > - return FALSE;
> >
> > +}
> > -
>
> You almost never want to compare floating point values like that. See
> e.g. compare_float() in dlls/ddraw/tests/ddraw7.c for a better (if
> perhaps not quite perfect) way. E.g. for the FLT_MAX case this only
> works because you happen to get an exact match. "1e-30" is a double
> instead of a float. Of course this is probably copied from the
> existing tests, so I suppose I can't complain too much.

Please also remove win_skips when _errno, strtof or _strtof_l are not 
available. The functions are present in all versions of msvcr120. Sorry 
for not spotting it earlier.

Thanks,
Piotr