[committed] d: gdc driver ignores -static-libstdc++ when automatically linking libstdc++ library

Hi,

This patch adds handling of `-static-libstc++' to the gdc driver, so
that libstdc++ is appropriately linked if it is either needed or seen on
the command-line.

This is required for bootstrapping the self hosted D front-end, so will
also be backported to all supported releases.

Bootstrapped and regression tested, and committed to mainline.

Regards
Iain.

---
	PR d/102574

gcc/d/ChangeLog:

	* d-spec.cc (lang_specific_driver): Link libstdc++ statically if
	-static-libstdc++ was given on command-line.
---
 gcc/d/d-spec.cc | 43 ++++++++++++++++++++++++++++++++++---------
 1 file changed, 34 insertions(+), 9 deletions(-)

Hi Iain

> On 4 Oct 2021, at 17:00, Iain Buclaw via Gcc-patches <gcc-patches@gcc.gnu.org> wrote:
> 
> Hi,
> 
> This patch adds handling of `-static-libstc++' to the gdc driver, so
> that libstdc++ is appropriately linked if it is either needed or seen on
> the command-line.
> 
> This is required for bootstrapping the self hosted D front-end, so will
> also be backported to all supported releases.
> 
> Bootstrapped and regression tested, and committed to mainline.
> 
> Regards
> Iain.
> 
> ---
> 	PR d/102574
> 
> gcc/d/ChangeLog:
> 
> 	* d-spec.cc (lang_specific_driver): Link libstdc++ statically if
> 	-static-libstdc++ was given on command-line.
> ---
> gcc/d/d-spec.cc | 43 ++++++++++++++++++++++++++++++++++---------
> 1 file changed, 34 insertions(+), 9 deletions(-)
> 
> diff --git a/gcc/d/d-spec.cc b/gcc/d/d-spec.cc
> index 16ff1539e9f..5adc662c6f2 100644
> --- a/gcc/d/d-spec.cc
> +++ b/gcc/d/d-spec.cc
> @@ -83,6 +83,9 @@ lang_specific_driver (cl_decoded_option **in_decoded_options,
>   /* "-lstdc++" if it appears on the command line.  */
>   const cl_decoded_option *saw_libcxx = 0;
> 
> +  /* True if we saw `-static-libstdc++'.  */
> +  bool saw_static_libcxx = false;
> +
>   /* Whether we need the C++ STD library.  */
>   bool need_stdcxx = false;
> 
> @@ -248,6 +251,11 @@ lang_specific_driver (cl_decoded_option **in_decoded_options,
> 	  shared_libgcc = false;
> 	  break;
> 
> +	case OPT_static_libstdc__:
> +	  saw_static_libcxx = true;
> +	  args[i] |= SKIPOPT;
> +	  break;
> +
> 	case OPT_static_libphobos:
> 	  if (phobos_library != PHOBOS_NOLINK)
> 	    phobos_library = PHOBOS_STATIC;
> @@ -452,16 +460,33 @@ lang_specific_driver (cl_decoded_option **in_decoded_options,
> #endif
>     }
> 
> -  if (saw_libcxx)
> -    new_decoded_options[j++] = *saw_libcxx;
> -  else if (need_stdcxx)
> +  if (saw_libcxx || need_stdcxx)
>     {
> -      generate_option (OPT_l,
> -		       (saw_profile_flag
> -			? LIBSTDCXX_PROFILE
> -			: LIBSTDCXX),
> -		       1, CL_DRIVER, &new_decoded_options[j++]);
> -      added_libraries++;
> +#ifdef HAVE_LD_STATIC_DYNAMIC
> +      if (saw_static_libcxx && !static_link)
> +	{
> +	  generate_option (OPT_Wl_, LD_STATIC_OPTION, 1, CL_DRIVER,
> +			   &new_decoded_options[j++]);
> +	}

For targets that don’t support HAVE_LD_STATIC_DYNAMIC it would be useful
to push the option back out, so that they can use that to substitute a static version
of the library using %:replace-outfile(-lxxxxx libxxxxx+.a%s) [ see darwin.h for 
examples. ] .. I suppose we could figure out a follow-on patch and test that on
Darwin?

so
#else
 … code to push the -libstdc++ out

(yes, we do have this problem also with the g++ driver… I posted a patch eons
 ago .. but suspect it was never applied)

Iain

> +#endif
> +      if (saw_libcxx)
> +	new_decoded_options[j++] = *saw_libcxx;
> +      else if (need_stdcxx)
> +	{
> +	  generate_option (OPT_l,
> +			   (saw_profile_flag
> +			    ? LIBSTDCXX_PROFILE
> +			    : LIBSTDCXX),
> +			   1, CL_DRIVER, &new_decoded_options[j++]);
> +	  added_libraries++;
> +	}
> +#ifdef HAVE_LD_STATIC_DYNAMIC
> +      if (saw_static_libcxx && !static_link)
> +	{
> +	  generate_option (OPT_Wl_, LD_DYNAMIC_OPTION, 1, CL_DRIVER,
> +			   &new_decoded_options[j++]);
> +	}
> +#endif
>     }
> 
>   if (shared_libgcc && !static_link)
> -- 
> 2.30.2
>

> On 04/10/2021 18:48 Iain Sandoe <idsandoe@googlemail.com> wrote:
> 
> For targets that don’t support HAVE_LD_STATIC_DYNAMIC it would be useful
> to push the option back out, so that they can use that to substitute a static version
> of the library using %:replace-outfile(-lxxxxx libxxxxx+.a%s) [ see darwin.h for 
> examples. ] .. I suppose we could figure out a follow-on patch and test that on
> Darwin?
> 
> so
> #else
>  … code to push the -libstdc++ out
> 
> (yes, we do have this problem also with the g++ driver… I posted a patch eons
>  ago .. but suspect it was never applied)
> 

Sure, seems reasonable to me, will follow up.

I had a look at darwin.h, and it occurs to me that you probably will be wanting -static-libphobos added there as well, which means moving that option to common.opt.

Iain.

Hi Iain,

> On 5 Oct 2021, at 09:02, ibuclaw@gdcproject.org wrote:
> 
>> On 04/10/2021 18:48 Iain Sandoe <idsandoe@googlemail.com> wrote:
>> 
>> For targets that don’t support HAVE_LD_STATIC_DYNAMIC it would be useful
>> to push the option back out, so that they can use that to substitute a static version
>> of the library using %:replace-outfile(-lxxxxx libxxxxx+.a%s) [ see darwin.h for 
>> examples. ] .. I suppose we could figure out a follow-on patch and test that on
>> Darwin?
>> 
>> so
>> #else
>> … code to push the -libstdc++ out
>> 
>> (yes, we do have this problem also with the g++ driver… I posted a patch eons
>> ago .. but suspect it was never applied)
>> 
> 
> Sure, seems reasonable to me, will follow up.

as it happens the current state breaks Darwin bootstrap, there’s a “set but unused”.

fixed by https://gcc.gnu.org/pipermail/gcc-patches/2021-October/581016.html

if that fix is acceptible, then we can backport it to the branches too - which will also be broken; if not then we should shake down an alternate soon.

> I had a look at darwin.h, and it occurs to me that you probably will be wanting -static-libphobos added there as well, which means moving that option to common.opt.

could be; but the fix above at least provides independent control over the static-libstdc++ situation (which tends to be more problematic on Darwin, since libstdc++ (latest) is always co-installed with the system’s (very old) version.

Iain