feat(054-fix-16-test-regressions-caused-by-item-051-bytecod): Fix 16 test regressions caused by item 051 bytecode format change. wf_exec:new/1 now expects {Bytecode, MetadataMap} tuple but many callers still pass plain [opcode()] lists. The fix: make wf_exec:new/1 accept both formats — if passed a plain list, wrap it as {List, #{}} with empty metadata. This ensures backward compatibility. All 16 failures are function_clause errors in wf_exec:new/1 called from wf_case_runner and other test modules. After fixing, rebar3 eunit must show 0 failures. Also implement real loop condition evaluation: currently the loop_check opcode handler always continues without evaluating the loop policy. For {count, N} loops, track iteration count per loop scope and exit when count is reached. ACCEPTANCE TEST: Add to test/wf_acceptance_tests.erl a loop_count_test that compiles wf_term:loop({count, 3}, wf_term:task(counter, #{function => fun(Ctx) -> N = maps:get(n, Ctx, 0), {ok, Ctx#{n => N + 1}} end})). Execute it. Assert the result is {done, _} not {yield, _}. Assert final context has n => 3.

Author: jmanhype

.wreckit/items/051-implement-real-task-function-dispatch-in-wfexec-cu/item.json

@@ -2,12 +2,14 @@
   "schema_version": 1,
   "id": "051-implement-real-task-function-dispatch-in-wfexec-cu",
   "title": "Implement real task function dispatch in wf_exec. Currently lookup_task_function/2 is a hardcoded mock that ignores the task name and returns fun(Ctx) -> {ok, maps:put(task_result, ok, Ctx)} end. Fix this so the executor actually calls the task function from the wf_term metadata. The compiler must propagate task metadata (the function field from task_metadata map) into the bytecode or a side-table accessible by the executor. The executor must call the real function when it encounters a task_exec opcode. ACCEPTANCE TEST: Create an escript that compiles wf_term:seq(wf_term:task(a, #{function => fun(Ctx) -> {ok, Ctx#{a_ran => true}} end}), wf_term:task(b, #{function => fun(Ctx) -> {ok, Ctx#{b_ran => true}} end})), executes it, and asserts the final context contains both a_ran and b_ran keys set to true. The test must fail with the current mock and pass after the fix. Do NOT use mocks or stubs. Write a real eunit test in test/wf_acceptance_tests.erl that verifies this.",
-  "state": "critique",
+  "state": "done",
   "overview": "",
-  "branch": null,
+  "branch": "wreckit/051-implement-real-task-function-dispatch-in-wfexec-cu",
   "pr_url": null,
   "pr_number": null,
   "last_error": null,
   "created_at": "2026-02-11T23:59:37.615Z",
-  "updated_at": "2026-02-12T00:09:25.850Z"
+  "updated_at": "2026-02-12T00:09:28.330Z",
+  "rollback_sha": "738a54d22a9ce4c23e8b054cafcdba3a69242725",
+  "completed_at": "2026-02-12T00:09:28.330Z"
 }

.wreckit/items/054-fix-16-test-regressions-caused-by-item-051-bytecod/item.json

@@ -0,0 +1,13 @@
+{
+  "schema_version": 1,
+  "id": "054-fix-16-test-regressions-caused-by-item-051-bytecod",
+  "title": "Fix 16 test regressions caused by item 051 bytecode format change. wf_exec:new/1 now expects {Bytecode, MetadataMap} tuple but many callers still pass plain [opcode()] lists. The fix: make wf_exec:new/1 accept both formats — if passed a plain list, wrap it as {List, #{}} with empty metadata. This ensures backward compatibility. All 16 failures are function_clause errors in wf_exec:new/1 called from wf_case_runner and other test modules. After fixing, rebar3 eunit must show 0 failures. Also implement real loop condition evaluation: currently the loop_check opcode handler always continues without evaluating the loop policy. For {count, N} loops, track iteration count per loop scope and exit when count is reached. ACCEPTANCE TEST: Add to test/wf_acceptance_tests.erl a loop_count_test that compiles wf_term:loop({count, 3}, wf_term:task(counter, #{function => fun(Ctx) -> N = maps:get(n, Ctx, 0), {ok, Ctx#{n => N + 1}} end})). Execute it. Assert the result is {done, _} not {yield, _}. Assert final context has n => 3.",
+  "state": "critique",
+  "overview": "",
+  "branch": null,
+  "pr_url": null,
+  "pr_number": null,
+  "last_error": null,
+  "created_at": "2026-02-12T00:10:54.767Z",
+  "updated_at": "2026-02-12T00:18:49.490Z"
+}