diff --git a/src/main.rs b/src/main.rs
index 1643de2..6d57d2f 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -107,7 +107,14 @@ fn uu_main() -> Result<(), Box<dyn Error>> {
         writeln!(out, "{table}")?;
     }
 
-    let intrp = Interpreter::new(ExecMode::Uu, cg, metadata);
+    // `--posix` / `--traditional` select ExecMode::Posix (RS first-char-only,
+    // FS="" whole-record field, stricter regex). Keep this mapping out of RS-only PRs.
+    let mode = if args.posix || args.traditional {
+        ExecMode::Posix
+    } else {
+        ExecMode::Uu
+    };
+    let intrp = Interpreter::new(mode, cg, metadata);
     AwkRt::new(intrp, bc, &args.read_queue, diagnostics).main_event_loop()?;
 
     Ok(())
diff --git a/tests/by-util/test_awk.rs b/tests/by-util/test_awk.rs
index 481b35c..f4a8b1f 100644
--- a/tests/by-util/test_awk.rs
+++ b/tests/by-util/test_awk.rs
@@ -602,6 +602,35 @@ fn ascii_char_fs_field_split() {
         .stdout_is("foo  bar baz\n");
 }
 
+/// gawk `rscompat`: under `--traditional`, multi-character `RS` uses only the
+/// first character (not a regexp). Requires `ExecMode::Posix` from `-c`.
+#[test]
+fn traditional_rs_uses_first_character_only() {
+    ucmd()
+        .args(&["-c", r#"BEGIN { RS = "bar" } { print $1, $2 }"#])
+        .pipe_in("0foobar0faz\n1foobar1faz")
+        .succeeds()
+        .stdout_is("0foo \nar0faz 1foo\nar1faz \n");
+}
+
+#[test]
+fn posix_rs_uses_first_character_only() {
+    ucmd()
+        .args(&["-P", r#"BEGIN { RS = "bar" } { print $1, $2 }"#])
+        .pipe_in("0foobar0faz\n1foobar1faz")
+        .succeeds()
+        .stdout_is("0foo \nar0faz 1foo\nar1faz \n");
+}
+
+/// POSIX `FS = ""`: the whole record is `$1` (not per-character fields).
+#[test]
+fn posix_empty_fs_is_whole_record_field() {
+    ucmd()
+        .args(&["-P", r#"BEGIN { FS = ""; $0 = "abc"; print NF, $1 }"#])
+        .succeeds()
+        .stdout_is("1 abc\n");
+}
+
 #[test]
 fn return_stmnt_sets_typedness_eagerly() {
     ucmd()