gettext: use relibc's error() instead of gnulib's replacement

gettext failed to link against the GCC 16 toolchain with

  ld: libc.a(...rcgu.o):(.bss.error_message_count+0x0): multiple
      definition of `error_message_count'; libgrt.a(libgrt_a-error.o):
      first defined here

Root cause is a cross-compile artefact, not a compiler change. gnulib's
check for a working error() is a RUN test, so cross-compiling reports
'checking for working error function... guessing no' and turns on
GNULIB_REPLACE_ERROR. Two things then go wrong at once: consumers are
redirected to rpl_error, and gnulib's own error.c -- listed in
am__objects_5 with no GL_COND_OBJ_ERROR guard -- still defines the
unrenamed error_* globals, which collide with relibc's.

relibc genuinely implements error(), error_at_line() and all three
globals (verified with nm). Two coordinated fixes:

- gl_cv_func_working_error=yes in the recipe. Stating the answer a run
  test cannot reach while cross-compiling is the standard autoconf
  mechanism, and is exactly what the neighbouring ac_cv_/gt_cv_ entries
  already do for the same reason. Consumers now call relibc's error()
  and no rpl_error reference remains.

- Guard gnulib's error.c on HAVE_ERROR, which configure already defines
  as 1 -- upstream gnulib later made this module conditional on the
  system lacking error(); this snapshot computes the right answer and
  does not act on it. Without this the definitions still collide, since
  relibc is Rust and one codegen unit carries several symbols, so
  referencing any of them drags the whole object in. glibc escapes this
  only by giving each its own object file.

gettext cooks clean. Patch generated by diff, applies at --fuzz=0.

local/patches/gettext/02-gnulib-error-system-provided.patch

@@ -0,0 +1,105 @@
+gnulib: do not define error() when the C library provides it.
+
+Companion to `gl_cv_func_working_error=yes` in the recipe. gnulib's probe
+for a working error() is a RUN test, so under cross-compilation it reports
+
+  checking for working error function... guessing no
+
+and sets GNULIB_REPLACE_ERROR. The cache variable stops that, so consumers
+call relibc's error() directly instead of rpl_error.
+
+That alone is not sufficient: gnulib-lib/Makefile lists error.o in
+am__objects_5 with no GL_COND_OBJ_ERROR guard, so error.c is compiled
+regardless and defines error_message_count, error_one_per_line and
+error_print_progname unconditionally. Those collide with relibc's, which
+carries all three plus error() and error_at_line():
+
+  ld: libc.a(...rcgu.o):(.bss.error_message_count+0x0): multiple definition
+      of `error_message_count'; libgrt.a(libgrt_a-error.o): first defined here
+
+relibc is Rust, so a single codegen unit carries several symbols and
+referencing any one of them drags in the whole object, definitions included
+-- glibc avoids this only because each lives in its own object file.
+
+Upstream gnulib later made this module conditional on the system lacking
+error(). configure already computes the right answer here (config.h defines
+HAVE_ERROR 1); it simply is not acted upon. Honour it and compile nothing
+when the C library supplies error().
+
+Redox's relibc provides error_at_line() alongside error(); a libc offering
+only one of the two would need this guard narrowed.
+
+--- a/gettext-runtime/gnulib-lib/error.c
++++ b/gettext-runtime/gnulib-lib/error.c
+@@ -22,6 +22,29 @@
+ # define _GL_NO_INLINE_ERROR
+ #endif
+ 
++/* Red Bear: relibc supplies error(), error_at_line() and the three error_*
++   globals (verified with nm on libc.a). This gnulib snapshot compiles
++   error.c unconditionally -- am__objects_5 in gnulib-lib/Makefile has no
++   GL_COND_OBJ_ERROR guard -- so its strong .bss definitions collide with
++   relibc's during static linking:
++
++     ld: libc.a(...rcgu.o):(.bss.error_message_count+0x0): multiple
++         definition of `error_message_count'; libgrt.a(libgrt_a-error.o):
++         first defined here
++
++   (relibc is Rust, so one codegen unit carries several symbols; referencing
++   any of them drags the whole object in, definitions included.)
++
++   Upstream gnulib later made this module conditional on the system lacking
++   error(). configure here already reaches the right conclusion -- config.h
++   defines HAVE_ERROR 1 -- it just is not acted on. Honour it: when the C
++   library provides error(), compile nothing. Redox's relibc provides
++   error_at_line() alongside error(); a libc offering only the former would
++   need this guard narrowed.  */
++#if defined HAVE_ERROR && HAVE_ERROR
++typedef int _gl_error_c_skipped_system_provides_error;
++#else
++
+ #include <error.h>
+ 
+ #include <stdarg.h>
+@@ -408,3 +431,4 @@
+ weak_alias (__error, error)
+ weak_alias (__error_at_line, error_at_line)
+ #endif
++#endif /* Red Bear: system provides error() */
+--- a/gettext-tools/gnulib-lib/error.c
++++ b/gettext-tools/gnulib-lib/error.c
+@@ -22,6 +22,29 @@
+ # define _GL_NO_INLINE_ERROR
+ #endif
+ 
++/* Red Bear: relibc supplies error(), error_at_line() and the three error_*
++   globals (verified with nm on libc.a). This gnulib snapshot compiles
++   error.c unconditionally -- am__objects_5 in gnulib-lib/Makefile has no
++   GL_COND_OBJ_ERROR guard -- so its strong .bss definitions collide with
++   relibc's during static linking:
++
++     ld: libc.a(...rcgu.o):(.bss.error_message_count+0x0): multiple
++         definition of `error_message_count'; libgrt.a(libgrt_a-error.o):
++         first defined here
++
++   (relibc is Rust, so one codegen unit carries several symbols; referencing
++   any of them drags the whole object in, definitions included.)
++
++   Upstream gnulib later made this module conditional on the system lacking
++   error(). configure here already reaches the right conclusion -- config.h
++   defines HAVE_ERROR 1 -- it just is not acted on. Honour it: when the C
++   library provides error(), compile nothing. Redox's relibc provides
++   error_at_line() alongside error(); a libc offering only the former would
++   need this guard narrowed.  */
++#if defined HAVE_ERROR && HAVE_ERROR
++typedef int _gl_error_c_skipped_system_provides_error;
++#else
++
+ #include <error.h>
+ 
+ #include <stdarg.h>
+@@ -408,3 +431,4 @@
+ weak_alias (__error, error)
+ weak_alias (__error_at_line, error_at_line)
+ #endif
++#endif /* Red Bear: system provides error() */

recipes/tools/gettext/02-gnulib-error-system-provided.patch

@@ -0,0 +1 @@
+../../../local/patches/gettext/02-gnulib-error-system-provided.patch

recipes/tools/gettext/recipe.toml

@@ -9,7 +9,8 @@ tar = "https://ftp.gnu.org/gnu/gettext/gettext-0.22.5.tar.gz"
 blake3 = "cb3f3a34da7ce1a92746df81f5b78c5d53841973a24eb80ab76537263d380ec0"
 patches = [
     "redox.patch",
-    "01-external-gettext.patch"
+    "01-external-gettext.patch",
+    "02-gnulib-error-system-provided.patch"
 ]
 script = """
 DYNAMIC_INIT
@@ -56,6 +57,16 @@ COOKBOOK_CONFIGURE_FLAGS+=(
     gt_cv_locale_ja=false
     gt_cv_locale_tr_utf8=false
     gt_cv_locale_zh_CN=false
+    # relibc implements error(), error_at_line() and the error_* globals.
+    # gnulib's probe for them is a RUN test, so cross-compiling makes it
+    # report "checking for working error function... guessing no" and set
+    # GNULIB_REPLACE_ERROR. Consumers then call rpl_error while gnulib's
+    # own error.c still defines the unrenamed globals, which both collides
+    # with relibc (multiple definition of `error_message_count') and leaves
+    # rpl_error undefined. Stating the answer the run test cannot reach is
+    # the standard autoconf mechanism, and is what the ac_cv_/gt_cv_ entries
+    # above already do for the same reason.
+    gl_cv_func_working_error=yes
 )
 cookbook_configure