fix: find_call_macro null-deref + macro_call regression tests + task_wrap_up skill

[borisbat]

Summary

Three threads from the same dasImgui Phase 0a session, landing together because they're tightly coupled (the bugfix unblocks one of the macro patterns the new tests pin):

• Bugfix. findModuleCallMacro deref'd the result of Module::findCall without null-checking. For unregistered call_macro names this crashed with EXCEPTION_ACCESS_VIOLATION. Add the null check; find_call_macro(mod, name) is now a safe soft-probe that returns null on miss. (src/builtin/module_builtin_ast_adapters.cpp)
• Regression coverage. Three new tests under tests/macro_call/:
  • test_find_call_macro_null — pins the bugfix above.
  • test_dyn_register_call_macro — pins the "function_macro registers a per-decl call_macro instance from inside apply()" pattern (what dasImgui's [widget] annotation relies on).
  • test_call_macro_emit_gate — pins the "call_macro gates add_variable on a find_variable miss" pattern (the auto-emit / idempotent module-global trick).
• Task wrap-up skill + MOUSE FIRST hard rule.
  • New skills/task_wrap_up.md — blind-mouse curation pass for any major task wrap, not just PRs (review the query log, surface the un-asked, cost calculus).
  • skills/make_pr.md §5.5 collapsed to a one-line pointer at the new skill.
  • CLAUDE.md gains a row in the skill table and a "MOUSE FIRST (hard rule)" section at the bottom: mouse__ask is the first tool on any pattern/why/how question, before Explore/Grep/Read.
  • 9 new mouse-data/docs/ entries seeded during the same Phase 0a session: reload firing-order, C++-resource preservation across reload, ImGui backend-data-survives-ctx-roundtrip, plus six macro / daspkg patterns.

The bugfix was already compiled into the local daslang.exe at session start, so the regression tests (4/4) passed against the fix. CI will rebuild from source and validate.

Test plan

• daslang.exe dastest/dastest.das -- --test tests/macro_call — all 4 tests pass (1 existing + 3 new) in 31 ms, no GC COMPILE LEAK / GC APP LEAK.
• MCP format_file on all 5 new .das files — already formatted.
• MCP lint on all 5 new .das files — 0 issues, 0 errors.
• CI validates AOT against the C++ change (skipped locally — change is a defensive null-check + early return, AOT-compiled callers shouldn't shift).
• CI validates Sphinx + extended_checks (no doc/RST changes; should be no-op).

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR hardens the AST C++ binding for find_call_macro to avoid a crash on missing names, and adds targeted macro-call regression tests to lock in key call-macro patterns (missing lookup, dynamic registration, and emit-or-reuse gating). It also updates the internal “skills” docs and mouse cache guidance to formalize a task wrap-up curation step and the “MOUSE FIRST” rule.

Changes:

• Fix findModuleCallMacro to null-check Module::findCall before dereferencing, making find_call_macro a safe “soft probe” on misses.
• Add tests/macro_call/* regression tests + helpers to pin missing lookup behavior, dynamic call_macro registration from apply(), and find_variable-gated add_variable emission.
• Add skills/task_wrap_up.md, update skills/make_pr.md, and extend CLAUDE.md + mouse docs to reinforce blind-mouse workflow.

Reviewed changes

Copilot reviewed 20 out of 20 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
src/builtin/module_builtin_ast_adapters.cpp	Adds a null-check to prevent find_call_macro crashes when the name is not registered.
tests/macro_call/test_find_call_macro_null.das	New regression test ensuring missing find_call_macro lookups return null (no crash).
tests/macro_call/_find_call_macro_helper.das	Helper call_macro that probes find_call_macro at compile time and folds to a bool.
tests/macro_call/test_dyn_register_call_macro.das	New regression test for dynamic per-decl call_macro registration via a function annotation.
tests/macro_call/_dyn_register_helper.das	Defines the [dyn_register_42] function_macro that registers a per-function call_macro instance.
tests/macro_call/_dyn_register_decl.das	Separate module consuming [dyn_register_42], providing an answer function for the test.
tests/macro_call/test_call_macro_emit_gate.das	New regression tests for “emit if absent, reuse if present” module-global gating.
tests/macro_call/_call_macro_emit_gate_helper.das	Helper call_macro implementing find_variable + conditional add_variable + ExprVar folding.
skills/task_wrap_up.md	New skill doc for blind-mouse wrap-up curation after meaningful work.
skills/make_pr.md	Points PR preflight mouse step to the new task wrap-up skill and updates quick reference.
CLAUDE.md	Adds the new skill to the table and documents the “MOUSE FIRST” hard rule.
mouse-data/docs/why-does-find-call-macro-crash-with-exception-access-violation-when-called-with-a-name-that-isn-t-registered.md	New mouse entry documenting the root cause and fix for the crash.
mouse-data/docs/how-do-i-write-a-regression-test-for-a-daslang-c-binding-bug-that-needs-a-module-pointer.md	New mouse entry describing the “use a call_macro as a bridge” testing pattern.
mouse-data/docs/how-do-i-dynamically-register-a-call-macro-instance-from-inside-another-macro-s-apply-or-visit.md	New mouse entry documenting dynamic call_macro instance registration pattern + gotchas.
mouse-data/docs/why-can-t-i-use-a-function-macro-annotation-in-the-same-file-that-declares-it.md	New mouse entry documenting annotation visibility constraints (same-file usage gotcha).
mouse-data/docs/my-call-macro-errors-with-can-t-locate-variable-ident-before-it-gets-a-chance-to-fire-how-do-i-keep-the-typer-from-pre-resolving.md	New mouse entry documenting canVisitArgument to avoid pre-resolution failures.
mouse-data/docs/my-daspkg-package-s-register-native-path-module-isn-t-found-what-s-the-require-path-supposed-to-look-like.md	New mouse entry explaining dynamic-module require path rules (group/mod).
mouse-data/docs/what-s-the-exact-firing-order-of-before_reload-shutdown-after_reload-init-during-a-daslang-live-reload.md	New mouse entry documenting daslang-live reload lifecycle ordering.
mouse-data/docs/how-do-i-preserve-a-c-owned-resource-across-a-daslang-live-reload.md	New mouse entry documenting the persistent-store pointer roundtrip pattern.
mouse-data/docs/does-imgui-s-backend-state-survive-daslang-live-reload-if-i-only-roundtrip-the-imguicontext-pointer.md	New mouse entry documenting ImGui backend persistence across reload via ctx-pointer roundtrip.

Comments suppressed due to low confidence (2)

mouse-data/docs/what-s-the-exact-firing-order-of-before_reload-shutdown-after_reload-init-during-a-daslang-live-reload.md:34

• This document has a duplicated ## Questions section at the end (second heading + repeated question). Removing the duplicate will avoid accidental drift between copies.

- In what context (old or new) do [before_reload] and [after_reload] run?

## Questions
- What's the exact firing order of [before_reload] / shutdown() / [after_reload] / init() during a daslang-live reload, and is is_reload() true throughout?

mouse-data/docs/how-do-i-preserve-a-c-owned-resource-across-a-daslang-live-reload.md:95

• The ## Questions section is duplicated at the end of the file (second heading + repeated first question). Removing the duplicate keeps the entry clean and prevents divergence.

- Why use a state-check (`if (live_x != null) return`) on init instead of `if (!is_reload())`?

## Questions
- How do I preserve a C++-owned resource (GLFW window, ImGui ctx, audio handle) across a daslang-live reload so the daslang-side state survives?

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 20 out of 20 changed files in this pull request and generated no new comments.