Re: [v2 4/7] msvcirt: Implement ostream::operator<< for floats

[Piotr Caban]

On 06/22/16 11:53, Iván Matellanes wrote:
> -
>    str_length += _scprintf(sprintf_fmt, prec, d); /* number representation */
> -
>    if (str_length > 24) {
> -
>        /* too long to output */
> -
>        ostream_writepad(ostr, "", "");

Could you please add a test for it? Also a comment saying that this is 
expected behavior would be useful.
> - float param_float[] = {1.0f, 0.0f, 4.25f, 3.999f, 12.0005f, 15.33582f, 15.0f, 15.22f, 21.123f, 0.1f,
> -
>    13.14159f, 0.00013f, 0.000013f, 1.0f / 0.0f, -1.0f / 0.0f, 0.0f / 0.0f};

Please do something similar as in dlls/msvcrt/tests/misc.c to define 
INFINITY/NAN in tests.

Thanks,
Piotr

[v2 4/7] msvcirt: Implement ostream::operator<< for floats

[Iván Matellanes]

On 22/06/16 12:25, Piotr Caban wrote:
> On 06/22/16 11:53, Iván Matellanes wrote:
> > -
> >    str_length += _scprintf(sprintf_fmt, prec, d); /* number 
> >
> > representation */
> > -
> >    if (str_length > 24) {
> > -
> >        /* too long to output */
> > -
> >        ostream_writepad(ostr, "", "");
>
> Could you please add a test for it? Also a comment saying that this is 
> expected behavior would be useful.

Thanks for the feedback.
This is the expected behavior on Vista+. XP/2k3 crashes though, so I 
ended up removing the related tests. What's the proper way to handle this?

Cheers,
Iván

[v2 4/7] msvcirt: Implement ostream::operator<< for floats

[Piotr Caban]

On 06/22/16 14:23, Iván Matellanes wrote:
> On 22/06/16 12:25, Piotr Caban wrote:
> > On 06/22/16 11:53, Iván Matellanes wrote:
> > > -
> > >    str_length += _scprintf(sprintf_fmt, prec, d); /* number
> > >
> > > representation */
> > > -
> > >    if (str_length > 24) {
> > > -
> > >        /* too long to output */
> > > -
> > >        ostream_writepad(ostr, "", "");
> >
> > Could you please add a test for it? Also a comment saying that this is
> > expected behavior would be useful.
>
> Thanks for the feedback.
> This is the expected behavior on Vista+. XP/2k3 crashes though, so I
> ended up removing the related tests. What's the proper way to handle this?

I think that you can just change the comment to make clear this is not 
incomplete implementation.

Thanks,
Piotr