Re: [PATCH 2/3] Add default implementation of fenv.h and all methods
"Howland, Craig D. - US via newlib" <[email protected]>
| Newsgroups | gmane.comp.lib.newlib |
|---|---|
| Message-ID | <[email protected]> |
> From: [email protected] <[email protected]> on behalf of Joel Sherrill <[email protected]> > Sent: Thursday, August 1, 2019 12:30 PM > To: Howland, Craig D. - US > Cc: [email protected] > Subject: Re: [PATCH 2/3] Add default implementation of fenv.h and all methods > > On Wed, Jul 31, 2019 at 6:58 PM Howland, Craig D. - US via newlib < > [email protected]> wrote: > > The RISC-V-specific things would probably be best removed from > > libc/include/sys/fenv.h, leaving the original one in machine (unlike the > > primary fenv.h, which will no longer need to be under machine, as you > > noted.) At a glance, using shorthand: > > -#define FE_TONEAREST_MM 0x00000004 > > -#define FE_RMODE_MASK 0x7 > > (Put another way, it should only have the bare-bones items defined in C > > and POSIX.) > > > > Thanks. I hadn't reviewed it against C/POSIX yet and should have. > > Do you think there should be comments about constants and types in the > default sys/fenv.h or just a blanket, this is the POSIX minimum that a port > has to define > > I honestly was worried the structure of this work or the regenerated stuff > would be awfully broken. I am thrilled to get meaty technical comments. > I did check the fenv.h and sys/fenv.h against POSIX. It certainly would be nice for sys/fenv.h to at a minimum point to POSIX for people to look up easily, and perhaps even better to also have a quick summary of what it wants, since it's intended as a template. Craig