Re: PATCH ld octets vs bytes support

[Timothy Wall <twall at domesolutions dot com>]

Sticking with the LD documentation, BYTE stores exactly one octet.  It'd
be up to an individual target to determine what it means to store a
number of octets that amount to less than a byte.

On the c54x (ignoring the LD docs for a moment), BYTE would be equivalent
to SHORT, using those terms (storing one byte/two octets in each case).
A reasonable thing to do here might be to set the size to be XXX_SIZE or
octets_per_byte, whichever is greater.  The LD documentation would need
to be updated appropriately, but it keeps the LD tokens (BYTE, SHORT,
etc) consistent with target usage.

T.

Ian Lance Taylor wrote:

>    Date: Wed, 26 Jan 2000 08:34:04 -0500
>    From: Timothy Wall <twall@tiac.net>
>
>    *************** lang_do_assignments (s, output_section_s
>    *** 3118,3139 ****
>                if (value.valid_p == false)
>                  einfo (_("%F%P: invalid data statement\n"));
>              }
>    !      switch (s->data_statement.type)
>    !        {
>    !        case QUAD:
>    !        case SQUAD:
>    !          dot += QUAD_SIZE;
>    !          break;
>    !        case LONG:
>    !          dot += LONG_SIZE;
>    !          break;
>    !        case SHORT:
>    !          dot += SHORT_SIZE;
>    !          break;
>    !        case BYTE:
>    !          dot += BYTE_SIZE;
>    !          break;
>    !        }
>              break;
>
>            case lang_reloc_statement_enum:
>    --- 3139,3164 ----
>                if (value.valid_p == false)
>                  einfo (_("%F%P: invalid data statement\n"));
>              }
>    !           {
>    !             int size;
>    !             switch (s->data_statement.type)
>    !               {
>    !               case QUAD:
>    !               case SQUAD:
>    !                 size = QUAD_SIZE;
>    !                 break;
>    !               case LONG:
>    !                 size = LONG_SIZE;
>    !                 break;
>    !               case SHORT:
>    !                 size = SHORT_SIZE;
>    !                 break;
>    !               case BYTE:
>    !                 size = BYTE_SIZE;
>    !                 break;
>    !               }
>    !             dot += size / opb;
>    !           }
>              break;
>
>            case lang_reloc_statement_enum:
>
> I'm a little concerned about this.  It seems that BYTE isn't going to
> work if opb > 1.  I think we should at least give a warning in the
> case where an unsupported data generation operator is used.
>
> Ian