[Bug c/92612] [10 Regression] Linker error in 525.x264_r after r278509

[wilco at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92612

--- Comment #2 from Wilco <wilco at gcc dot gnu.org> ---
(In reply to Martin Liška from comment #1)
> Following patch fixes that:
> 
> diff --git a/benchspec/CPU/525.x264_r/src/ldecod_src/inc/configfile.h
> b/benchspec/CPU/525.x264_r/src/ldecod_src/inc/configfile.h
> index 12ed1cc8..450930e2 100644
> --- a/benchspec/CPU/525.x264_r/src/ldecod_src/inc/configfile.h
> +++ b/benchspec/CPU/525.x264_r/src/ldecod_src/inc/configfile.h
> @@ -18,9 +18,9 @@
>  //#define LEVEL_IDC       21
>  
>  
> -InputParameters cfgparams;
>  
>  #ifdef INCLUDED_BY_CONFIGFILE_C
> +InputParameters cfgparams;
>  // Mapping_Map Syntax:
>  // {NAMEinConfigFile,  &cfgparams.VariableName, Type, InitialValue,
> LimitType, MinLimit, MaxLimit, CharSize}
>  // Types : {0:int, 1:text, 2: double}
> @@ -59,6 +59,7 @@ extern Mapping Map[];
>  #endif
>  extern void JMDecHelpExit ();
>  extern void ParseCommand(InputParameters *p_Inp, int ac, char *av[]);
> +extern InputParameters cfgparams;
>  
>  #endif

(In reply to Martin Liška from comment #0)
> Since defaulting to -fno-common, one can't build the benchmark due to:

Yes some SPEC benchmarks rely on tentative definitions. Given it is a
portability issue, but the source can't be fixed, the easiest way is to add
-fcommon in the config file for the failing cases. This is similar to the other
workarounds like -fno-strict-aliasing.