Re: [PATCH 04/25] Centralize i386 linux target descriptions

[Simon Marchi <simon.marchi@polymtl.ca>]

On 2017-06-12 10:41, Yao Qi wrote:
> This patch moves all the tdesc_i386*_linux target descriptions to a
> function i386_linux_read_description, which returns the right target
> description according to xcr0.  This also remove the duplication in
> getting target descriptions in corefile and native target.
> 
> gdb:
> 
> 2017-04-27  Yao Qi  <yao.qi@linaro.org>
> 
> 	* i386-linux-tdep.c (i386_linux_read_description): New function.
> 	(i386_linux_core_read_description): Call
> 	i386_linux_read_description.
> 	* i386-linux-tdep.h (i386_linux_read_description): Declare.
> 	(tdesc_i386_linux, tdesc_i386_mmx_linux): Remove declarations.
> 	(tdesc_i386_avx_linux, tdesc_i386_mpx_linux): Likewise
> 	(tdesc_i386_avx_mpx_linux, tdesc_i386_avx_avx512_linux): Likewise.
> 	(tdesc_i386_avx_mpx_avx512_pku_linux): Likewise.
> 	* x86-linux-nat.c (x86_linux_read_description): Call
> 	i386_linux_read_description.
> ---
>  gdb/i386-linux-tdep.c | 34 ++++++++++++++++++++++------------
>  gdb/i386-linux-tdep.h | 10 ++--------
>  gdb/x86-linux-nat.c   | 24 ++++++++----------------
>  3 files changed, 32 insertions(+), 36 deletions(-)
> 
> diff --git a/gdb/i386-linux-tdep.c b/gdb/i386-linux-tdep.c
> index 1909d61..1bc1a6f 100644
> --- a/gdb/i386-linux-tdep.c
> +++ b/gdb/i386-linux-tdep.c
> @@ -678,16 +678,9 @@ i386_linux_core_read_xcr0 (bfd *abfd)
>    return xcr0;
>  }
> 
> -/* Get Linux/x86 target description from core dump.  */
> -
> -static const struct target_desc *
> -i386_linux_core_read_description (struct gdbarch *gdbarch,
> -				  struct target_ops *target,
> -				  bfd *abfd)
> +const struct target_desc *
> +i386_linux_read_description (uint64_t xcr0)

This would need a comment

   /* See i386-linux-tdep.h.  */

>  {
> -  /* Linux/i386.  */
> -  uint64_t xcr0 = i386_linux_core_read_xcr0 (abfd);
> -
>    switch ((xcr0 & X86_XSTATE_ALL_MASK))
>      {
>      case X86_XSTATE_AVX_MPX_AVX512_PKU_MASK:
> @@ -708,10 +701,27 @@ i386_linux_core_read_description (struct gdbarch 
> *gdbarch,
>        break;
>      }
> 
> +  return NULL;
> +}

I was going to say: I think it would make more sense if 
i386_linux_read_description always returned something non-NULL, but then 
I realized you need to handle the special case in 
i386_linux_core_read_description...

> +
> +/* Get Linux/x86 target description from core dump.  */
> +
> +static const struct target_desc *
> +i386_linux_core_read_description (struct gdbarch *gdbarch,
> +				  struct target_ops *target,
> +				  bfd *abfd)
> +{
> +  /* Linux/i386.  */
> +  uint64_t xcr0 = i386_linux_core_read_xcr0 (abfd);
> +  const struct target_desc * tdesc = i386_linux_read_description 
> (xcr0);

Extra space after *.

The logic seems OK to me.  It's not related to your patch, but I find a 
bit confusing that i386-linux has SSE, and in i386-mmx-linux, the mmx is 
there to denote the lack of sse.  I looks like Mark agrees with me :)

   https://sourceware.org/ml/gdb-patches/2010-04/msg00300.html

Simon