fix(gc): PERRY_GC_STACKMAP_TRACE must parse its value, not its presence

changelog.d/8164-stackmap-trace-knob-polarity.md

@@ -0,0 +1,12 @@
+`PERRY_GC_STACKMAP_TRACE` parses its value instead of its presence, restoring
+the `lint` gate. The knob arrived in #8131 reading `var_os(..).is_some()`, which
+inverts the spelling a reader is most likely to try: `PERRY_GC_STACKMAP_TRACE=0`
+still sets the variable, so it turned the trace ON. `check_gc_env_knobs` — a
+required-check audit that exists to keep every GC knob on the shared parser —
+failed on it, and a red `lint` on main blocks every open PR.
+
+It now uses `gc::env_flag_enabled`, the default-OFF parser (the same shape as
+`policy.rs`'s `PERRY_GC_TRACE`), which fails toward the knob's documented
+default so a typo leaves the instrument off rather than silently arming it.
+`=1`/`on`/`true` still enable it and unset still leaves it off; `=0`/`off`/
+`false`/`no` now disable it.

crates/perry-runtime/src/gc/roots/stack_maps.rs

@@ -1114,7 +1114,11 @@ mod unwind {
 
     fn walk_trace_enabled() -> bool {
         static ON: std::sync::OnceLock<bool> = std::sync::OnceLock::new();
-        *ON.get_or_init(|| std::env::var_os("PERRY_GC_STACKMAP_TRACE").is_some())
+        // `env_flag_enabled`, not `var_os(..).is_some()`: presence-testing makes
+        // `PERRY_GC_STACKMAP_TRACE=0` ENABLE the trace, which is the opposite of
+        // what every other GC knob does and what anyone typing `=0` means. The
+        // shared parser fails toward the knob's documented default (OFF here).
+        *ON.get_or_init(|| crate::gc::env_flag_enabled("PERRY_GC_STACKMAP_TRACE"))
     }
 
     unsafe extern "C" fn walk_frame<F: FnMut(ResolvedRoot)>(