Re: [PATCH 1/4] Use nanf("") instead of nanf(NULL)

[Craig Howland <howland at LGSInnovations dot com>]

On 08/27/2018 06:10 PM, Joseph Myers wrote:
> On Mon, 27 Aug 2018, Craig Howland wrote:
>
> > Do we want to allow this recent GCC change to force a source change, when the
> > GCC change (added warning) is not a solid one?  The C standard does not
> > require that the argument to nan() be non-null, although invoking general,
> The argument must, as per DR#477 (resolution integrated in C17), be a
> string.
>
> http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2257.htm#dr_477
Good that that is being taken care of.  While this does not change the fact that 
the change is not operationally needed in that the Newlib nan() ignores the 
argument, it does say changing for the sake the source being compliant (with the 
upcoming revision/clarification) should be done--even if generated code 
changes.  With that said, the edits look good.

Craig