[PATCH] gas: Don't error when .debug_line already exists, unless .loc was used

[Mark Wielaard]

On Fri, 2020-09-11 at 06:51 -0700, H.J. Lu wrote:
> On Fri, Sep 11, 2020 at 5:48 AM Nick Clifton via Binutils
> <binutils@sourceware.org> wrote:
> > > The tests are unfortunately arch specific because the line table
> > > is only
> > > generated when actual instructions have been emitted.
> > 
> > Are you sure about that ?  How about using the NOP instruction ?
> > 
> > I admit that some architectures do need special syntax for their
> > nop instructions, but there is already a test in gas testsuite
> > that handles this (see gas/testsuite/gas/all/gas.exp:org-1).  The
> > nop-selecting code there could probably be extracted into a library
> > function then used in lots of places...
> 
> 7.66 '.nops SIZE[, CONTROL]'
> ============================
> 
> This directive emits SIZE bytes filled with no-op instructions.  SIZE
> is
> absolute expression, which must be a positve value.  CONTROL controls
> how no-op instructions should be generated.  If the comma and CONTROL
> are omitted, CONTROL is assumed to be zero.

O nice. Somewhat embarrassingly I hadn't seen that because I was using
the online documentation: 
https://sourceware.org/binutils/docs/as/Pseudo-Ops.html

Which I now realize is not current. Oops. Can someone (I?) update that?

Sadly .nops does produce one or more architecture specific nop
instructions, but does not result in a call to dwarf2_emit_insn like
real instructions. So it looks like I still need to use the
gas.exp:org-1 trick.

Or is it a bug that .nops doesn't trigger a call to dwarf2_imit_insn?

Another issue is that at least NEWS claims:

* Add support for .nops directive.  It is currently supported only for
  x86 targets.

Is that still true? If it isn't supported generically, it isn't really
that useful.

Thanks,

Mark