fix: replayConst with native_decide (#8157)

This PR fixes an incompatibility of `replayConst` as used by e.g.
`aesop` with `native_decide`-using tactics such as `bv_decide`

src/Lean/Environment.lean

@@ -2081,6 +2081,16 @@ where
   replayKernel (exts : Array (EnvExtension EnvExtensionState)) (consts : List AsyncConst)
       (kenv : Kernel.Environment) : Except Kernel.Exception Kernel.Environment := do
     let mut kenv := kenv
+    -- replay extensions first in case kernel checking needs them (`IR.declMapExt`)
+    for ext in exts do
+      if let some replay := ext.replay? then
+        kenv := { kenv with
+          -- safety: like in `modifyState`, but that one takes an elab env instead of a kernel env
+          extensions := unsafe (ext.modifyStateImpl kenv.extensions <|
+            replay
+              (ext.getStateImpl oldEnv.toKernelEnv.extensions)
+              (ext.getStateImpl newEnv.toKernelEnv.extensions)
+              (consts.map (·.constInfo.name))) }
     for c in consts do
       if skipExisting && (kenv.find? c.constInfo.name).isSome then
         continue
@@ -2102,15 +2112,6 @@ where
       -- realized kernel additions cannot be interrupted - which would be bad anyway as they can be
       -- reused between snapshots
       kenv ← ofExcept <| kenv.addDeclCore 0 decl none
-    for ext in exts do
-      if let some replay := ext.replay? then
-        kenv := { kenv with
-          -- safety: like in `modifyState`, but that one takes an elab env instead of a kernel env
-          extensions := unsafe (ext.modifyStateImpl kenv.extensions <|
-            replay
-              (ext.getStateImpl oldEnv.toKernelEnv.extensions)
-              (ext.getStateImpl newEnv.toKernelEnv.extensions)
-              (consts.map (·.constInfo.name))) }
     return kenv
 
 /-- Like `evalConst`, but first check that `constName` indeed is a declaration of type `typeName`.

tests/lean/run/replayConst.lean

@@ -0,0 +1,15 @@
+import Lean.Elab.Tactic.Basic
+import Std.Tactic.BVDecide
+
+/-! `replayConst` should be able to replay constants using `native_decide`. -/
+
+open Lean Lean.Meta Lean.Elab.Tactic in
+elab "replay" ts:tacticSeq : tactic => do
+  let initEnv ← getEnv
+  evalTactic ts
+  let finalState ← saveState
+  setEnv $ ← initEnv.replayConsts initEnv finalState.term.meta.core.env
+
+theorem test
+    (x y : BitVec 128) (h : y = x) : x = y := by
+  replay bv_decide