[Bug gas/31400] New: The third expression of .align

[mengqinggang at loongson dot cn]

https://sourceware.org/bugzilla/show_bug.cgi?id=31400

            Bug ID: 31400
           Summary: The third expression of .align
           Product: binutils
           Version: 2.42
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: gas
          Assignee: unassigned at sourceware dot org
          Reporter: mengqinggang at loongson dot cn
  Target Milestone: ---

$ cat align.s
  1 ret
  2 .align 16, , 1
  3 ret
  4 .align 16, , -1
  5 ret
  6 .align 16, , -4294967295
  7 ret

$ as align.s
$ objdump -d a.out

On X86:
0000000000000000 <.text>:
   0:   c3                      ret
   1:   c3                      ret
   2:   66 66 2e 0f 1f 84 00    data16 cs nopw 0x0(%rax,%rax,1)
   9:   00 00 00 00
   d:   0f 1f 00                nopl   (%rax)
  10:   c3                      ret
  11:   c3                      ret

AArch64 has similar output:
0000000000000000 <.text>:
   0:   d65f03c0        ret
   4:   d65f03c0        ret
   8:   d503201f        nop
   c:   d503201f        nop
  10:   d65f03c0        ret
  14:   d65f03c0        ret

If the third parameter of .align is negative, should it be ignored?
But now there are different behaviors.


For llvm: if the third parameter is negative, it will cause an error.

# llvm-mc align.s -filetype=obj
align.s:4:14: error: alignment directive can never be satisfied in this many
bytes, ignoring maximum bytes expression
.align 16, , -1
             ^
align.s:6:14: error: alignment directive can never be satisfied in this many
bytes, ignoring maximum bytes expression
.align 16, , -4294967295
             ^

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