F90-related question

[Janus Weil]

2013/12/10 Steve Kargl <sgk@troutmask.apl.washington.edu>:
> On Tue, Dec 10, 2013 at 05:31:10PM +0100, Janus Weil wrote:
>>
>> Sorry, I don't follow what you're trying to say. The behavior changed
>> between 4.5 and 4.6 due to the addition of REAL(16), fine. But how is
>> that related to the sentence I want to remove? The fact that
>> -fdefault-real-8 also promotes DOUBLE PRECISION to REAL(16) is still
>> in there.
>>
>
> Yes, of course, you're correct.  My original text from 9 years ago
> was
>
> 197              @cindex option, -fdefault-real-8
> 198              @cindex -fdefault-real-8, option
> 199              @item -fdefault-real-8
> 200              Set the default real type to an 8 byte wide type.
> 201              Do nothing if this is already the default.
>
> This looks like an attempt to explain what the code says.  Looking at
> the diff, I had
>
> +  /* Choose the default double kind.  If -fdefault-real and -fdefault-double
> +     are specified, we use kind=8, if it's available.  If -fdefault-real is
> +     specified without -fdefault-double, we use kind=16, if it's available.
> +     Otherwise we do not change anything.  */
> +  if (gfc_option.flag_default_double && !gfc_option.flag_default_real)
> +    fatal_error ("Use of -fdefault-double-8 requires -fdefault-real-8");
> +
> +  if (gfc_option.flag_default_real && gfc_option.flag_default_double && saw_r8)
> +    gfc_default_double_kind = 8;
> +  else if (gfc_option.flag_default_real && saw_r16)
>
> When Daniel added the additional text to the manual, we probably
> should have reaped that sentence.  So, yes, it can be removed.

Taking this as an ok for the last patch I posted, I just committed it
as r205871 (with a small correction over the posted version, where
some possible kinds for DOUBLE PRECISION were missing):

http://gcc.gnu.org/viewcvs/gcc?view=revision&revision=205871

Cheers,
Janus