Re: [PATCH] dwarf2out: Emit DWARF 6 DW_AT_language_{name,version}

Hi Jakub,

On Tue, 2025-01-07 at 20:22 +0100, Jakub Jelinek wrote:
> DWARF has voted in yesterday https://dwarfstd.org/issues/241209.1.html ,
> which is basically just a guarantee that the DWARF 6 draft
> DW_AT_language_{name,version} attribute codes and content of
> https://dwarfstd.org/languages-v6.html can be used as an extension
> in DWARF 5 and won't be changed.
> 
> So, this patch is an alternative to the
> https://gcc.gnu.org/pipermail/gcc-patches/2024-November/669671.html
> patch, which had the major problem that it required changing all the
> DWARF consumers to be able to debug C17 or later or C++17 or later
> sources.

Note that most consumers (binutils, gdb, systemtap, valgrind, elfutils)
have already been updated to use the new DW_LANG constants. We can
easily backport that to the last stable releases before gcc 15 is
released.

> This patch uses still DWARF 5 DW_LANG_C11 or DW_LANG_C_plus_plus_14,
> the latest code in DWARF 5 proper, so all DWARF 5 capable consumers
> should be able to deal with that, but additionally emits the
> DWARF 6 attributes so that newer DWARF consumers can see it isn't
> just C++14 but say C++23 or C11 but C23.  Consumers which don't know
> those DWARF 6 attributes would just ignore them.  This is like any other
> -gno-strict-dwarf extension, except that normally we emit say DWARF 5
> codes where possible only after DWARF 5 is released, while in this case
> there is a guarantee it can be used before DWARF 6 is released.

Code looks ok to me, it is only for C and C++, would be nice to at
least get it for Fortran and Ada too.

> --- gcc/dwarf2out.cc.jj	2025-01-02 11:23:35.541251268 +0100
> +++ gcc/dwarf2out.cc	2025-01-07 10:09:16.866866563 +0100
> @@ -8755,6 +8755,14 @@ break_out_comdat_types (dw_die_ref die)
>          unit = new_die (DW_TAG_type_unit, NULL, NULL);
>          add_AT_unsigned (unit, DW_AT_language,
>                           get_AT_unsigned (comp_unit_die (), DW_AT_language));
> +	if (unsigned lname = get_AT_unsigned (comp_unit_die (),
> +					      DW_AT_language_name))
> +	  {
> +	    add_AT_unsigned (unit, DW_AT_language_name, lname);
> +	    add_AT_unsigned (unit, DW_AT_language_version,
> +			     get_AT_unsigned (comp_unit_die (),
> +					      DW_AT_language_version));
> +	  }

This relies on language_name and language_version always being set
together. Which is the case in the code at this time. Should we assert
that?

> @@ -25376,11 +25388,28 @@ gen_compile_unit_die (const char *filena
>  	    language = DW_LANG_C99;
>  
>  	  if (dwarf_version >= 5 /* || !dwarf_strict */)
> -	    if (strcmp (language_string, "GNU C11") == 0
> -		|| strcmp (language_string, "GNU C17") == 0
> -		|| strcmp (language_string, "GNU C23") == 0
> -		|| strcmp (language_string, "GNU C2Y") == 0)
> -	      language = DW_LANG_C11;
> +	    {
> +	      if (strcmp (language_string, "GNU C11") == 0)
> +		language = DW_LANG_C11;
> +	      else if (strcmp (language_string, "GNU C17") == 0)
> +		{
> +		  language = DW_LANG_C11;
> +		  lname = DW_LNAME_C;
> +		  lversion = 201710;
> +		}
> +	      else if (strcmp (language_string, "GNU C23") == 0)
> +		{
> +		  language = DW_LANG_C11;
> +		  lname = DW_LNAME_C;
> +		  lversion = 202311;
> +		}
> +	      else if (strcmp (language_string, "GNU C2Y") == 0)
> +		{
> +		  language = DW_LANG_C11;
> +		  lname = DW_LNAME_C;
> +		  lversion = 202500;
> +		}
> +	    }

The use of language_string (lang_hooks.name) is a little clumsy, but
already part of the current code. In the future it would be nice to
have separate language and version hooks for this.

Personally I would just go with the new DW_LANG_C17/C23 language code,
they will be supported by consumers when gcc 15 is released.

The 00 for the month in C2Y is clever. Does that match the
__STDC_VERSION__ defined?

> @@ -25392,12 +25421,30 @@ gen_compile_unit_die (const char *filena
>  	    language = DW_LANG_C_plus_plus_11;
>  	  else if (strcmp (language_string, "GNU C++14") == 0)
>  	    language = DW_LANG_C_plus_plus_14;
> -	  else if (strcmp (language_string, "GNU C++17") == 0
> -		   || strcmp (language_string, "GNU C++20") == 0
> -		   || strcmp (language_string, "GNU C++23") == 0
> -		   || strcmp (language_string, "GNU C++26") == 0)
> -	    /* For now.  */
> -	    language = DW_LANG_C_plus_plus_14;
> +	  else if (strcmp (language_string, "GNU C++17") == 0)
> +	    {
> +	      language = DW_LANG_C_plus_plus_14;
> +	      lname = DW_LNAME_C_plus_plus;
> +	      lversion = 201703;
> +	    }
> +	  else if (strcmp (language_string, "GNU C++20") == 0)
> +	    {
> +	      language = DW_LANG_C_plus_plus_14;
> +	      lname = DW_LNAME_C_plus_plus;
> +	      lversion = 202002;
> +	    }
> +	  else if (strcmp (language_string, "GNU C++23") == 0)
> +	    {
> +	      language = DW_LANG_C_plus_plus_14;
> +	      lname = DW_LNAME_C_plus_plus;
> +	      lversion = 202302;
> +	    }
> +	  else if (strcmp (language_string, "GNU C++26") == 0)
> +	    {
> +	      language = DW_LANG_C_plus_plus_14;
> +	      lname = DW_LNAME_C_plus_plus;
> +	      lversion = 202400;
> +	    }

Why 202400 and not 202600?

Cheers,

Mark