On Fri, 14 Mar 2025, Florian Weimer wrote:
> * Richard Biener:
>
> > @@ -3869,6 +3867,18 @@ gen_command_line_string (cl_decoded_option *options,
> > case OPT_fchecking_:
> > /* Ignore these. */
> > continue;
> > + case OPT_D:
> > + case OPT_U:
> > + if (startswith (options[i].arg, "_FORTIFY_SOURCE")
> > + && (options[i].arg[sizeof ("_FORTIFY_SOURCE") - 1] == '\0'
> > + || (options[i].opt_index == OPT_D
> > + && options[i].arg[sizeof ("_FORTIFY_SOURCE") - 1] == '=')))
> > + {
> > + switches.safe_push (options[i].orig_option_with_args_text);
> > + len += strlen (options[i].orig_option_with_args_text) + 1;
> > + }
> > + /* Otherwise ignore these. */
> > + continue;
>
> Would it be possible to cover -Wp,-D,_FORTIFY_SOURCE… and
> -Wp,-U_FORTIFY_SOURCE as well? Some of us used -Wp,-D,_FORTIFY_SOURCE…
> for compatibility with the GCJ compiler driver, where
> -D_FORTIFY_SOURCE=2 meant something completely different. And now we
> are stuck with those variants because some scrits use sed to remove
> those flags, assuming the presence -Wp, prefix.
I've tested that it works with -Wp as well without any special further
handling:
> ./xgcc -B. -g t.c -c -Wp,-D_FORTIFY_SOURCE=2
> readelf -w t.o | grep DW_AT_producer
<d> DW_AT_producer : (indirect string, offset: 0): GNU C23 15.0.1
20250314 (experimental) -D_FORTIFY_SOURCE=2 -mtune=generic -march=x86-64
-g
Richard.
