Fix record field-reorder redefine (jolt-wf4) and builtin-shadowing local names (jolt-fjb1) (#124)

* Don't direct-link a var redefined earlier in the same unit (jolt-wf4)

defrecord/deftype expands to (do (def R (make-deftype-ctor ...)) (def ->R R) ...),
so the ->R alias references R within one compiled unit. Under direct-link a var
ref embeds the cell's root as a compile-time constant, but on a redefine R's old
root is still in place when that unit compiles — the (def R new) sibling hasn't
run yet — so ->R sealed to the stale pre-redef ctor. (defrecord R [x a])
(defrecord R [a x]) (:a (->R 10 20)) read the old [x a] layout and returned 20.

Track the vars a unit (re)defines and force their later in-unit references to the
live indirect deref. The cell is registered only after its own init is emitted,
so a recursive self-reference inside the init still direct-links (it runs after
the def completes); only sibling references after the def go indirect.

* Emit Janet's `in` as a value so a user local can't shadow it (jolt-fjb1)

The back end emits `in` to deref var cells ((in cell :root)) and index
shape-recs. It emitted the bare symbol, so a user local named `in` shadowed
Janet's builtin in the surrounding scope and the generated cell-deref called the
user's value as a function — "<table> called with 2 arguments, possibly expected
1". malli's explainer binds [value in acc], so m/explain hit this on every
schema (m/validate was unaffected — its path doesn't bind `in`).

Embed `in`'s function value at the emit sites (as jolt-call/core-get already
are); a value in head position can't be shadowed. Fixes m/explain on malli
(loaded with JOLT_FEATURES=clj so its .cljc reader-conditionals resolve).

---------

Co-authored-by: Yogthos <yogthos@gmail.com>

src/jolt/backend.janet

@@ -32,6 +32,14 @@
 # var-get/tuple-slice/...) and jolt runtime helpers resolve by name.
 (def jolt-runtime-env (curenv))
 
+# Janet's `in` accessor as an embedded VALUE (jolt-fjb1). The back end uses `in`
+# to index cells/shape-recs/arg tuples in code it EMITS. Emitting the bare symbol
+# `in` is unhygienic: a user local named `in` (malli's explainer binds `[value in
+# acc]`) shadows it in the surrounding scope, so the generated `(in cell :root)`
+# would call the user's value as a function — "<table> called with 2 arguments".
+# Embedding the function value (as with jolt-call/core-get) can't be shadowed.
+(def- jin in)
+
 (defn ctx-janet-env
   "Lazily create/cache a per-context Janet environment for compiled code: a child
   of the runtime env (so core fns resolve) that holds this context's user defs.
@@ -120,10 +128,21 @@
 # So we direct-link exactly the call-optimization case; everything else stays
 # indirect (live var deref → redefinable). Default user/REPL units: flag off,
 # so all user calls are indirect and redefinable with no annotation.
+# A var DEF'd earlier as a sibling in the current compilation unit can't be
+# direct-linked/const-linked: its embedded root is the value from BEFORE this
+# unit runs, but the unit's own (def …) rebinds it first — so a later reference
+# in the same unit (e.g. deftype's `(def ->R R)` alias after `(def R …)`) would
+# capture the stale pre-redef root (jolt-wf4). Self-reference inside a def's own
+# init is unaffected: the cell is registered only AFTER its init is emitted, so a
+# fn body's recursive call still direct-links (and runs after the def completes).
+(defn- unit-redefined? [ctx cell]
+  (let [ud (get (ctx :env) :unit-defs)] (and ud (get ud cell) true)))
+
 (defn- direct-var? [ctx cell]
   (and (get (ctx :env) :direct-linking?)
        (not (cell :dynamic))
        (not (let [m (cell :meta)] (and m (get m :redef))))
+       (not (unit-redefined? ctx cell))
        (let [r (cell :root)] (or (function? r) (cfunction? r)))))
 
 # Whole-program constant-linking (closed world): under JOLT_WHOLE_PROGRAM every
@@ -141,6 +160,7 @@
   (and (get (ctx :env) :whole-program?)
        (get (ctx :env) :direct-linking?)
        (not (cell :dynamic))
+       (not (unit-redefined? ctx cell))
        (not (nil? (cell :root)))))
 
 # Fresh Janet symbol for back-end-introduced bindings (arity dispatch). NOT
@@ -181,7 +201,7 @@
 (defn- emit-arity-invoke [ctx ar jargs]
   (def nfixed (length (vview (ar :params))))
   (def call @[(emit-arity-fn ctx ar)])
-  (for i 0 nfixed (array/push call ['in jargs i]))
+  (for i 0 nfixed (array/push call [jin jargs i]))
   # empty rest binds to NIL, not () — (f) with [& r] gives r = nil in Clojure
   (when (ar :rest)
     (array/push call ['if ['> ['length jargs] nfixed] ['tuple/slice jargs nfixed]]))
@@ -213,7 +233,7 @@
 # live-root call the :var emit uses, for the ns get/set helpers below.
 (defn- core-fn-call [ctx nm & args]
   (def cell (cell-for ctx "clojure.core" nm))
-  (tuple (tuple 'in (tuple 'quote cell) :root) ;args))
+  (tuple (tuple jin (tuple 'quote cell) :root) ;args))
 
 (defn- emit-try [ctx node]
   (def core
@@ -408,11 +428,11 @@
   # NOT gated on compile-time shapes — core is baked without the flag but still
   # receives user shape-recs, so this must hold in baked core too. (jolt-t34)
   (defn get-or-shape [getexpr]
-    (if sidx ['in m sidx]
+    (if sidx [jin m sidx]
       (let [pos (jsym)]
         ['if ['and ['tuple? m] ['struct? ['get m 0]]]
           ['let [pos ['get [['get m 0] :idx] k]]
-            ['if ['nil? pos] (tuple core-get m k nil) ['in m ['+ 1 pos]]]]
+            ['if ['nil? pos] (tuple core-get m k nil) [jin m ['+ 1 pos]]]]
           getexpr])))
   (if (nil? d-expr)
     (let [fast (get-or-shape ['get m k])]
@@ -620,9 +640,9 @@
                # majority) pay two native table ops + a branch instead of a
                # function call.
                (let [qcell (tuple 'quote cell)]
-                 ['if ['in qcell :dynamic]
+                 ['if [jin qcell :dynamic]
                    (tuple var-get qcell)
-                   ['in qcell :root]]))))
+                   [jin qcell :root]]))))
       # (var x): the var object itself (not its value) — the embedded cell, by
       # reference. binding keys its thread-binding frame on this exact cell.
       :the-var (tuple 'quote (cell-for ctx (node :ns) (node :name)))
@@ -633,10 +653,15 @@
       :try (emit-try ctx node)
       :throw ['error (emit ctx (node :expr))]
       :def (let [cell (cell-for ctx (node :ns) (node :name))
-                 meta (node :meta)]
+                 meta (node :meta)
+                 setter (if (and meta (not (empty? meta))) (var-setter-meta cell meta) (var-setter cell))]
              (inline-stash! ctx cell node)
-             (tuple (if (and meta (not (empty? meta))) (var-setter-meta cell meta) (var-setter cell))
-                    (emit ctx (node :init))))
+             # Emit the init BEFORE marking the cell unit-defined, so a recursive
+             # self-reference in the init still direct-links; a LATER sibling
+             # reference in this unit then sees it as redefined (jolt-wf4).
+             (let [init-form (emit ctx (node :init))]
+               (when-let [ud (get (ctx :env) :unit-defs)] (put ud cell true))
+               (tuple setter init-form)))
       :let (emit-let ctx node)
       :fn (emit-fn ctx node)
       :invoke (emit-invoke ctx node)
@@ -649,6 +674,10 @@
 (defn emit-ir
   "IR node -> Janet form (public entry for the back end)."
   [ctx node]
+  # Fresh per-unit set of vars (re)defined in THIS top-level form, so a sibling
+  # reference to one can't direct-link to its pre-redef root (jolt-wf4). Scoped
+  # to one emit pass; each compile-and-eval / re-emit call is its own unit.
+  (when (table? (get ctx :env)) (put (ctx :env) :unit-defs @{}))
   (emit ctx node))
 
 # --- pipeline wiring (the self-hosted compile path) ---

test/integration/direct-linking-test.janet

@@ -58,6 +58,19 @@
 (let [ctx (init {:compile? true :aot-core? false})]
   (check "aot-core off still correct" (eval-string ctx "(last [1 2 3])") 3))
 
+# 6. Redefining a record with REORDERED fields must rebind the ctor (jolt-wf4).
+#    deftype expands to (do (def R (make-deftype-ctor ...)) (def ->R R) ...): the
+#    `->R` alias references R in the SAME compiled unit. Direct-link embeds a var's
+#    root as a compile-time constant, but R's redefined root hasn't RUN yet when
+#    that unit compiles — so ->R must NOT seal to R's stale (pre-redef) ctor.
+(let [ctx (init {:compile? true :direct-linking? true})]
+  (eval-string ctx "(defrecord R [x a])")
+  (eval-string ctx "(defrecord R [a x])")
+  (check "record field-reorder redefine: :a reads the new layout"
+         (eval-string ctx "(:a (->R 10 20))") 10)
+  (check "record field-reorder redefine: :x reads the new layout"
+         (eval-string ctx "(:x (->R 10 20))") 20))
+
 (if (pos? failures)
   (do (printf "direct-linking: %d failure(s)" failures) (os/exit 1))
   (print "direct-linking: all matrix cases passed"))

test/spec/functions-spec.janet

@@ -11,7 +11,13 @@
   ["variadic with fixed" "[1 [2 3]]" "(do (defn f [a & xs] [a xs]) (f 1 2 3))"]
   ["closure captures"   "8"      "(do (defn adder [n] (fn [x] (+ x n))) ((adder 5) 3))"]
   ["recursion"          "120"    "(do (defn fact [n] (if (< n 2) 1 (* n (fact (dec n))))) (fact 5))"]
-  ["named fn self-ref"  "120"    "((fn fact [n] (if (< n 2) 1 (* n (fact (dec n))))) 5)"])
+  ["named fn self-ref"  "120"    "((fn fact [n] (if (< n 2) 1 (* n (fact (dec n))))) 5)"]
+  # A param whose name collides with a Janet builtin the back end emits in
+  # generated code (here `in`, used to deref vars / index shape-recs) must not
+  # shadow it — malli's explainer binds `[value in acc]` (jolt-fjb1).
+  ["param named `in`"   "1"       "((fn [in] (first in)) [1 2 3])"]
+  ["param `in` via core fn" "3"   "((fn [in] (count in)) [1 2 3])"]
+  ["local `in` in let body" "2"   "(let [in [2 3]] (first in))"])
 
 (defspec "functions / application"
   ["apply"              "6"      "(apply + [1 2 3])"]