Re: [Qemu-devel] [PATCH V2] softfloat: Rename float*_is_nan() functions to float*_is_quiet_nan()

[Andreas Färber]

Am 17.12.2010 um 17:48 schrieb Peter Maydell:

> On 17 December 2010 16:19, Andreas Färber <address@hidden>  
> wrote:
> > > Can this patch be applied? (There are more ARM/softfloat patches
> > > in my queue which will break it again otherwise...)
> >
> > If we're engaging into refactoring the softfloat library, I still  
> > have a
> > large'ish patch lying around to fix signature mismatches between  
> > header and
> > sources wrt to integer arguments and migrating to POSIX integer types
> > (BeOS/Haiku system headers define int32 etc. differently). Browsing  
> > through
> > your patch it shouldn't conflict, but I guess it'll be best to have  
> > it go
> > through your queue to avoid troubles.
>
> I wasn't planning to put this patch into my "ARM fixes" queue to be
> pulled directly, because it's a bit wider in scope than fixing things
> for ARM targets. (Hence the "can this be applied?" request :-))
>
> On the types issue, at the moment softfloat uses "int32" etc for
> "a handy type holding at least 32 bits", and "bits32" for "exactly
> 32 bits". So I guess changing the 'bits' types to the POSIX int32_t
> and friends would be straightforward enough, but what does your
> patch do with the int32 types?

My patch does not touch the bits* types. I didn't notice any problem  
there.

I replaced int32 by int32_t, int64 by int64_t etc. No sane code puts  
more than 32 bits into an "int32" variable, and my guests on OSX/ppc64  
host still appeared to work. I don't have arm guests though so please  
check on your side.

There's int_least*_t in case we really need it:
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/stdint.h.html

Andreas