Fix for PR 16299: Do not assume that bad relocations always have an associated symbol.
This commit is contained in:
parent
9d1b068967
commit
644a26542c
|
@ -1,3 +1,8 @@
|
|||
Wed Oct 7 14:09:14 1998 Nick Clifton <nickc@cygnus.com>
|
||||
|
||||
* config/tc-d30v.c (md_apply_fix3): Do not assume that bad
|
||||
relocations are always associated with a symbol.
|
||||
|
||||
Tue Oct 6 09:31:15 1998 Catherine Moore <clm@cygnus.com>
|
||||
|
||||
* tc-sparc.h (TC_FORCE_RELOCATION): Define.
|
||||
|
|
|
@ -759,13 +759,14 @@ write_2_short (opcode1, insn1, opcode2, insn2, exec_type, fx)
|
|||
|
||||
if (exec_type == EXEC_SEQ
|
||||
&& (opcode1->op->flags_used & (FLAG_JMP | FLAG_JSR))
|
||||
&& ((opcode1->op->flags_used & FLAG_DELAY) == 0)
|
||||
&& ((opcode1->ecc == ECC_AL) || ! Optimizing))
|
||||
{
|
||||
/* Unconditional branches kill instructions in the right bin.
|
||||
Conditional branches don't always but if we are not
|
||||
optimizing, then we want to produce an error about such
|
||||
constructs. For the purposes of this test, subroutine
|
||||
calls are considered to be branches. */
|
||||
/* Unconditional, non-delayed branches kill instructions in
|
||||
the right bin. Conditional branches don't always but if
|
||||
we are not optimizing, then we have been asked to produce
|
||||
an error about such constructs. For the purposes of this
|
||||
test, subroutine calls are considered to be branches. */
|
||||
write_1_short (opcode1, insn1, fx->next, false);
|
||||
return 1;
|
||||
}
|
||||
|
@ -799,10 +800,11 @@ write_2_short (opcode1, insn1, opcode2, insn2, exec_type, fx)
|
|||
}
|
||||
}
|
||||
else if (opcode1->op->flags_used & (FLAG_JMP | FLAG_JSR)
|
||||
&& ((opcode1->op->flags_used & FLAG_DELAY) == 0)
|
||||
&& ((opcode1->ecc == ECC_AL) || ! Optimizing))
|
||||
{
|
||||
/* We must emit branch type instruction on its own with
|
||||
nothing in the right container. */
|
||||
/* We must emit (non-delayed) branch type instructions
|
||||
on their own with nothing in the right container. */
|
||||
write_1_short (opcode1, insn1, fx->next, false);
|
||||
return 1;
|
||||
}
|
||||
|
@ -1739,12 +1741,11 @@ md_apply_fix3 (fixp, valuep, seg)
|
|||
fixp->fx_done = 1;
|
||||
}
|
||||
else if (fixp->fx_pcrel)
|
||||
{
|
||||
value = * valuep;
|
||||
}
|
||||
else
|
||||
{
|
||||
value = fixp->fx_offset;
|
||||
|
||||
if (fixp->fx_subsy != (symbolS *) NULL)
|
||||
{
|
||||
if (S_GET_SEGMENT (fixp->fx_subsy) == absolute_section)
|
||||
|
@ -1765,23 +1766,30 @@ md_apply_fix3 (fixp, valuep, seg)
|
|||
|
||||
switch (fixp->fx_r_type)
|
||||
{
|
||||
case BFD_RELOC_8:
|
||||
/* Caused by a bad .byte directive. */
|
||||
as_fatal (_("line %d: unable to place address of symbol '%s' into a byte"),
|
||||
fixp->fx_line, S_GET_NAME (fixp->fx_addsy));
|
||||
break;
|
||||
case BFD_RELOC_8: /* Caused by a bad .byte directive. */
|
||||
/* Drop trhough. */
|
||||
|
||||
case BFD_RELOC_16:
|
||||
/* Caused by a bad .short directive. */
|
||||
as_fatal (_("line %d: unable to place address of symbol '%s' into a short"),
|
||||
fixp->fx_line, S_GET_NAME (fixp->fx_addsy));
|
||||
break;
|
||||
case BFD_RELOC_16: /* Caused by a bad .short directive. */
|
||||
/* Drop through. */
|
||||
|
||||
case BFD_RELOC_64:
|
||||
/* Caused by a bad .quad directive. */
|
||||
as_fatal (_("line %d: unable to place address of symbol '%s' into a .quad"),
|
||||
fixp->fx_line, S_GET_NAME (fixp->fx_addsy));
|
||||
case BFD_RELOC_64: /* Caused by a bad .quad directive. */
|
||||
{
|
||||
char * size;
|
||||
|
||||
size = (fixp->fx_r_type == BFD_RELOC_8) ? _("byte")
|
||||
: (fixp->fx_r_type == BFD_RELOC_16) ? _("short")
|
||||
: _("quad");
|
||||
|
||||
if (fixp->fx_addsy == NULL)
|
||||
as_bad (_("line %d: unable to place address into a %s"),
|
||||
fixp->fx_line, size);
|
||||
else
|
||||
as_bad (_("line %d: unable to place address of symbol '%s' into a %s"),
|
||||
fixp->fx_line,
|
||||
S_GET_NAME (fixp->fx_addsy),
|
||||
size);
|
||||
break;
|
||||
}
|
||||
|
||||
case BFD_RELOC_D30V_6:
|
||||
check_size (value, 6, fixp->fx_file, fixp->fx_line);
|
||||
|
|
Loading…
Reference in New Issue