[Bug target/87986] New: Assembler errors w/ -masm=intel

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87986

            Bug ID: 87986
           Summary: Assembler errors w/ -masm=intel
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: assemble-failure
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: asolokha at gmx dot com
  Target Milestone: ---
            Target: x86_64-pc-linux-gnu

Let's consider the following cases:

% cat logpnyyb.c 
int bx;

void
ue (void)
{
  bx = 0;
}

% cat rluc6lhr.c 
void
k1 (void)
{
}

Compiling them for x86_64 w/ -masm=intel yields the following:

% gcc-9.0.0-alpha20181111 -pipe -masm=intel -c logpnyyb.c
{standard input}: Assembler messages:
{standard input}:15: Error: invalid use of register

% cat logpnyyb.s 
        .file   "logpnyyb.c"
        .intel_syntax noprefix
        .text
        .comm   bx,4,4
        .globl  ue
        .type   ue, @function
ue:
.LFB0:
        .cfi_startproc
        push    rbp
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        mov     rbp, rsp
        .cfi_def_cfa_register 6
        mov     DWORD PTR bx[rip], 0
        nop
        pop     rbp
        .cfi_def_cfa 7, 8
        ret
        .cfi_endproc
.LFE0:
        .size   ue, .-ue
        .ident  "GCC: (GNU) 9.0.0-alpha20181111 20181111 (experimental)"
        .section        .note.GNU-stack,"",@progbits

% gcc-9.0.0-alpha20181111 -pipe -masm=intel -c rluc6lhr.c
{standard input}: Assembler messages:
{standard input}: Error: .size expression for k1 does not evaluate to a
constant

% cat rluc6lhr.s 
        .file   "rluc6lhr.c"
        .intel_syntax noprefix
        .text
        .globl  k1
        .type   k1, @function
k1:
.LFB0:
        .cfi_startproc
        push    rbp
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        mov     rbp, rsp
        .cfi_def_cfa_register 6
        nop
        pop     rbp
        .cfi_def_cfa 7, 8
        ret
        .cfi_endproc
.LFE0:
        .size   k1, .-k1
        .ident  "GCC: (GNU) 9.0.0-alpha20181111 20181111 (experimental)"
        .section        .note.GNU-stack,"",@progbits

Top of trunk gcc is not actually necessary, it also reproduces at least on the
gcc 8 branch.

I have binutils 2.31.1 installed. I'm not sure which tool, gcc or gas, might be
at fault here, so I'll move this PR to the binutils Bugzilla if necessary.