Re: gcc 11.1.0: printf("%.43f\n", 0x1.52f8a8e32e982p-140): printed value is incorrectly rounded

[Corinna Vinschen <vinschen@redhat.com>]

On Nov 30 19:51, Takashi Yano wrote:
> On Mon, 29 Nov 2021 16:55:32 +0100
> Corinna Vinschen wrote:
> > On Nov 29 23:24, Takashi Yano wrote:
> > > [...]
> > > > > > > I have tried to import gdtoa into newlib from OpenBSD.
> > > > > > > [...]
> > [...struct __reent stuff...]
> Thanks for checking and advice. I have done these.
> 
> > As for the allocations, how big are those?  If they are comparable with
> > the allocation we now perform in _ldtoa_r, it might not be worth to keep
> > both functions.  Some users of smaller targets might also complain that
> 
> I saw max 16404 byte malloc().
> 
> > using printf now always pulls in both variants of ldtoa, thus raising
> > code size unnecessarily.  It might be better to keep the calls separate
> > and only use one or the other, per target or per code size constraints,
> > perhaps as a configure option.
> 
> I have added --enable-newlib-use-gdtoa option, which defaults
> to 'yes', into newlib/configure.ac. Is this the right thing?

That patch looks good to me, at least as far as Cygwin is concerned.

This isn't essential, but it might make sense to rename __ldtoa to
_ldtoa_r to avoid an extra function call, which could be time consuming
and add stack pressure.  I. e., in gdtoa-ldtoa.c

  #ifdef _USE_GDTOA
  // all code in gdtoa-ldtoa.c
  #endif

and in ldtoa.c:

  #ifndef _USE_GDTOA
  // all code in ldtoa.c
  #endif

AFAICS, __ldtoa could easily be changed to take the long double argument
by value because it's used in only two places, one of which just checks
the value anyway.  But, as I said, not essential.  We can keep in mind
for the time being.


Could some people with other targets than Cygwin give this patch a try?
RTEMS, anybody?


Thanks,
Corinna