Re: [patch, fortran] PR37077 Implement Internal Unit I/O for character KIND=4, Part 1

[Jerry DeLisle <jvdelisle at verizon dot net>]

On 07/12/2010 05:27 AM, Tobias Burnus wrote:
> On 07/11/2010 03:27 PM, Jerry DeLisle wrote:
> > Please disregard rev e of the patch.  I introduced a breakage in it
> > inadvertently.  The attached is the corrected patch, rev f.
>
> Thanks for the patch! I have not reviewed your patch yet, but only
> glanced at it.
>
> Question: Why do you define "mem_alloc_r4" but never use it? (In
> mem_read4 you use "mem_alloc_r".)

I am not done with the READ portion.  I anticipate using it.  If not, it will go 
away in Part 2 of this implementation.

> Typo: s/internel/internal/
>
> +      if (dtp->common.unit) /* char4 internel unit.  */

OK

> @@ -58,7 +82,10 @@ write_default_char4 (st_parameter_dt *dtp, gfc_cha
>         p = write_block (dtp, k);
>         if (p == NULL)
>   	return;
> -      memset (p, ' ', k);
> +      if (unlikely (is_char4_unit (dtp)))
> +	memset4 (p, 0, ' ', k);
> +      else
> +	memset (p, ' ', k);
>       }
>
>
> Question: Why is this unlikely? I understand that wide-char I/O is
> unlikely, but for write_default_char4 writing to a wide-char internal
> unit does not seem to be much more unlikely than writing to a file unit.
> (Ditto a few lines later; on the other hand, using unlikely in "write_a"
> is fine.)

OK

> Finally, I miss tests for the test suit.
>
> Tobias,
> who will try to review the patch this evening or tomorrow.

First of several test cases attached.

Jerry

Attachment: char4_iunit_1.f03
Description: Text document