Re: [PATCH] dsound: Don't use lrintf if it's not available.

[Matteo Bruni]

2017-10-29 21:59 GMT+01:00 Alex Henrie :
> Signed-off-by: Alex Henrie
>
> Fixes https://bugs.winehq.org/show_bug.cgi?id=43000
>
> lrintf is not part of C89, and we were already using HAVE_LRINTF in
> msvcrt.
>
> dlls/dsound/dsound_convert.c | 16 ++++++++++++++++
>  1 file changed, 16 insertions(+)
>
> diff --git a/dlls/dsound/dsound_convert.c b/dlls/dsound/dsound_convert.c
> index 6887bae558..ad1aba3737 100644
> --- a/dlls/dsound/dsound_convert.c
> +++ b/dlls/dsound/dsound_convert.c
> @@ -120,7 +120,11 @@ static inline unsigned char f_to_8(float value)
>          return 0;
>      if(value >= 1.f * 0x7f / 0x80)
>          return 0xFF;
> +#ifdef HAVE_LRINTF
>      return lrintf((value + 1.f) * 0x80);
> +#else
> - return (value + 1) * 0x80;
>
> +#endif
>  }

The fallback should probably add 0.5f before truncating. Not entirely
equivalent but probably closer to the original code.

BTW, we also use lrint() in d3dx9, so we might want to fix it there too.
Actually, wouldn't it be better to add generic lrint() / lrintf()
fallbacks in libs/port/ instead?

[PATCH] dsound: Don't use lrintf if it's not available.

[Alex Henrie]

2017-11-01 13:26 GMT+01:00 Matteo Bruni :
> 2017-10-29 21:59 GMT+01:00 Alex Henrie :
> > Signed-off-by: Alex Henrie
> >
> > Fixes https://bugs.winehq.org/show_bug.cgi?id=43000
> >
> > lrintf is not part of C89, and we were already using HAVE_LRINTF in
> > msvcrt.
> >
> > dlls/dsound/dsound_convert.c | 16 ++++++++++++++++
> >  1 file changed, 16 insertions(+)
> >
> > diff --git a/dlls/dsound/dsound_convert.c b/dlls/dsound/dsound_convert.c
> > index 6887bae558..ad1aba3737 100644
> > --- a/dlls/dsound/dsound_convert.c
> > +++ b/dlls/dsound/dsound_convert.c
> > @@ -120,7 +120,11 @@ static inline unsigned char f_to_8(float value)
> >          return 0;
> >      if(value >= 1.f * 0x7f / 0x80)
> >          return 0xFF;
> > +#ifdef HAVE_LRINTF
> >      return lrintf((value + 1.f) * 0x80);
> > +#else
> > - return (value + 1) * 0x80;
> >
> > +#endif
> >  }
>
> The fallback should probably add 0.5f before truncating. Not entirely
> equivalent but probably closer to the original code.
>
> BTW, we also use lrint() in d3dx9, so we might want to fix it there too.
> Actually, wouldn't it be better to add generic lrint() / lrintf()
> fallbacks in libs/port/ instead?

You're right, it would be better to move the lrint and lrintf fallback
implementations from msvcrt to libport. Should I also move the rint,
rintf, llrint, and llrintf fallbacks to libport?

-Alex