RE: [RFA 6/6] Remove long_long_align_bit gdbarch attribute

[Anton Kolesov <Anton.Kolesov@synopsys.com>]

> First, arc-tdep.c claimed that long long alignment was 32 bits, but gcc's arc.h
> says:
> 
>     #define ADJUST_FIELD_ALIGN(FIELD, TYPE, COMPUTED) \
>     (TYPE_MODE (strip_array_types (TYPE)) == DFmode \
>      ? MIN ((COMPUTED), 32) : (COMPUTED))
> 
> Here, DFmode means "double".  So, I've implemented arc_type_align
> according to this.  I do not have a good way to test this.
> 
> diff --git a/gdb/arc-tdep.c b/gdb/arc-tdep.c index b0d51addd3..86cf522443
> 100644
> --- a/gdb/arc-tdep.c
> +++ b/gdb/arc-tdep.c
> @@ -1957,6 +1957,19 @@ arc_tdesc_init (struct gdbarch_info info, const
> struct target_desc **tdesc,
>    return TRUE;
>  }
> 
> +/* Implement the type_align gdbarch function.  */
> +
> +static ULONGEST
> +arc_type_align (struct gdbarch *gdbarch, struct type *type) {
> +  type = check_typedef (type);
> +
> +  if (TYPE_CODE (type) == TYPE_CODE_FLT && TYPE_LENGTH (type) == 8)
> +    return 4;
> +
> +  return TYPE_LENGTH (type);
> +}
> +

Hi Tom,

If I read this correctly, then it would set 4 byte alignment only to 8 byte
floating point type? That doesn't look right, because on ARC there is never
more than 4 byte alignment for any data type. GCC arc.h has:

    #define BIGGEST_ALIGNMENT 32

According to comments it used to be 64, but was changed to 32 at some
undefined point in the past.

I would have implemented this as:

static ULONGEST
arc_type_align (struct gdbarch *gdbarch, struct type *type) {
  type = check_typedef (type);
  return std::min<ULONGEST>(4, TYPE_LENGTH (type));
}

Anton