gh-108562: Revert enabling -fstrict-overflow for libmpdec (GH-116302)

gh-108562: partial reversion of pr114751

Reverts -fstrict-overflow for libmpdec

Author: skirpichev

Makefile.pre.in

@@ -1213,7 +1213,7 @@ PYTHON_HEADERS= \
 
 ##########################################################################
 # Build static libmpdec.a
-LIBMPDEC_CFLAGS=$(PY_STDMODULE_CFLAGS) @LIBMPDEC_CFLAGS@ $(CCSHARED)
+LIBMPDEC_CFLAGS=@LIBMPDEC_CFLAGS@ $(PY_STDMODULE_CFLAGS) $(CCSHARED)
 
 # "%.o: %c" is not portable
 Modules/_decimal/libmpdec/basearith.o: $(srcdir)/Modules/_decimal/libmpdec/basearith.c $(LIBMPDEC_HEADERS) $(PYTHON_HEADERS)

configure

@@ -3896,11 +3896,6 @@ AS_VAR_IF([with_system_libmpdec], [yes], [
   LIBMPDEC_LDFLAGS="-lm \$(LIBMPDEC_A)"
   LIBMPDEC_INTERNAL="\$(LIBMPDEC_HEADERS) \$(LIBMPDEC_A)"
 
-  dnl Enable strict-overflow for libmpdec, if available, see GH-108562
-  AS_VAR_IF([ac_cv_cc_supports_fstrict_overflow], [yes], [
-    AS_VAR_APPEND([LIBMPDEC_CFLAGS], [" -fstrict-overflow"])
-  ])
-
   dnl Disable forced inlining in debug builds, see GH-94847
   AS_VAR_IF([with_pydebug], [yes], [
     AS_VAR_APPEND([LIBMPDEC_CFLAGS], [" -DTEST_COVERAGE"])