[Bug gas/20068] New: GAS complains about ‘Error: floating point number invalid‘ when producing a literal pool.

[chuazl at comp dot nus.edu.sg]

https://sourceware.org/bugzilla/show_bug.cgi?id=20068

            Bug ID: 20068
           Summary: GAS complains about ‘Error: floating point number
                    invalid‘ when producing a literal pool.
           Product: binutils
           Version: 2.27 (HEAD)
            Status: NEW
          Severity: normal
          Priority: P2
         Component: gas
          Assignee: unassigned at sourceware dot org
          Reporter: chuazl at comp dot nus.edu.sg
  Target Milestone: ---

Created attachment 9248
  --> https://sourceware.org/bugzilla/attachment.cgi?id=9248&action=edit
Change the type of padding to O_constant

Assembling the following ARM assembly will result in gas complaining:
LDR R0, =0x12345678
FLDD D9, =0xfff000000fff
.LTORG
‘Error: floating point number invalid‘

Adding an additional 4 bytes into the literal pool will cause GAS to assemble
without any errors.
LDR R0, =0x12345678
LDR R0, =0x87654321
FLDD D9, =0xfff000000fff
.LTORG

Debugging GAS reveals that the error is generated from emit_expr_with_reloc(),
more specifically, it fails this test : 
if (op == O_big && exp->X_add_number <= 0)

Digging further, this expression is generated due to alignment for a 8 byte
long literal being added to the pool, more specifically in add_to_lit_pool().
If the pool is not 8 byte aligned and a 8 byte literal is added to the pool, a
4 byte pad will be generated with the following code.

<...>
else if (pool_size & 0x7)
{
          <....> 
          pool->literals[entry] = inst.reloc.exp;
          pool->literals[entry].X_add_number = 0;
          pool->literals[entry++].X_md = (PADDING_SLOT << 8) | 4;
          pool->next_free_entry += 1;
          pool_size += 4;
}

Since the current exp is of type O_big, the padding expression will be of type
O_big with X_add_number = 0 failing the above test.

Workaround patch:
Change the type of the padding to O_constant. Again, not too sure about the
correctness.

-- 
You are receiving this mail because:
You are on the CC list for the bug.