Re: [V4 00/21] Fortran dynamic array support

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Thu, 25 Aug 2016 19:06:26 +0200, Jan Kratochvil wrote:
> I have attached a fix.

That fix has a regression for normal (64-bit) inferiors.

I see the source handles negative stide specially.  Particularly the comment
here does not explain the code it comments:

      /* Adjust the data location with the value of byte stride if set, which
         can describe the separation between successive elements along the
         dimension.  */
      if (TYPE_BYTE_STRIDE (range_type) < 0)
        value += (TYPE_HIGH_BOUND (range_type) - TYPE_LOW_BOUND (range_type))
                  * TYPE_BYTE_STRIDE (range_type);
  else if (elt_stride < 0)
    {
      int offs = (elt_offs + 1) * elt_stride;

      elt_offs = TYPE_LENGTH (array_type) + offs;
    }

But I do not negative stride values described in DWARF-4 anyhow.  When it is
used?  With this interpretation of stride the gfortran random(?) stride value
for arrays with single row would be really wrong.  But I do not see
a definition of negative stride in DWARF.


Thanks,
Jan