## Summary
- **Extract `Indexer` class** — owns all index state (ProjectIndex,
MergedIndex shards, OpenFileIndex) and query methods (definition,
references, call/type hierarchy, workspace symbol search)
- **Extract `Compiler` class** — owns document state, PCH/PCM cache,
compile argument resolution, header context, `ensure_compiled`, and
worker forwarding
- **MasterServer is now a pure LSP handler registration layer** (~700
lines, down from ~3200)
- **`MergedIndexShard`** wraps `index::MergedIndex` with a lazily-cached
PositionMapper; `OpenFileIndex` gains matching
`find_occurrence()`/`find_relations()` APIs — callers get pre-converted
LSP ranges directly
- **Indexer returns typed values** (`vector<Location>`,
`vector<CallHierarchyIncomingCall>`, etc.) instead of pre-serialized
JSON, fixing the references handler from JSON string surgery to simple
vector concatenation
- **Fix**: duplicate `workspace/symbol` loop in the original code
## Test plan
- [x] 465 unit tests pass
- [x] 113 integration tests pass
- [x] 2/2 smoke tests pass
- [x] `clang-format` applied
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Server-side C++ compilation orchestration (module & precompiled header
builds) with LSP-integrated document handling.
* **Improvements**
* Deterministic, persistent, dependency-aware caching to avoid redundant
rebuilds and speed up incremental work.
* Better cross-file indexing and navigation, improved diagnostics and
more reliable include/import-aware completions.
* **Tests**
* Unit tests updated to the unified worker query/build request shapes.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
- Add `Embed` and `HasEmbed` structs to `directive.h` for storing
`#embed` / `__has_embed` directive info
- Implement `EmbedDirective` and `HasEmbed` PPCallbacks in
`DirectiveCollector`, using the current `CompilationUnitRef` API
- Add unit tests for both directives (including non-existent file
handling)
Rebased onto current main, resolving conflicts from the `Compiler/` →
`compile/` restructuring and the `CompilationUnitRef` API migration.
Original PR by @Guo-Shiyu.
## Test plan
- [x] `pixi run unit-test` — 465 tests passed, 0 failures
Co-authored-by: ykiko <ykikoykikoykiko@gmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
- classify unresolved using declarations with the `dependentName`
semantic token modifier
- include the declaration modifier on definitions to match clangd's
semantic token behavior
- extend the semantic token modifier legend coverage in integration
tests
## Testing
- pixi run python -m pytest -s --log-cli-level=INFO
tests/integration/test_server.py -k 'semantic_token_modifier_legend or
capabilities' --executable=./build/bin/clice
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added many new semantic token modifiers (deprecated, deduced,
readonly, static, abstract, virtual, dependent-name,
constructor/destructor, user-defined, mutable-usage flags) and new scope
markers (function, class, file, global).
* Improved tagging for declarations, definitions and dependent names in
semantic tokens.
* **Tests**
* Added an integration test verifying the semantic token modifier legend
and its ordering.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
- **Remove xmake build system**: delete `xmake.lua`, `test-xmake.yml`
workflow, and all xmake-related pixi tasks
- **Migrate packaging to CMake**: `publish-clice.yml` now uses `cmake
-DCLICE_RELEASE=ON` instead of xmake pack
- **Clean up**: remove unused `tests/uv.lock`, `.xmake/` from
`.gitignore`, xmake references from docs
- **Benchmark CI**: change trigger from `pull_request` to
`workflow_dispatch` (manual only)
- **Simplify CLI parsing**: use `DecoKV(style =
KVStyle::JoinedOrSeparate)` in `clice.cc` and `unit_tests.cc`, replacing
verbose `DecoKVStyled` with manual `static_cast` bitmask; use comma
separators; explicit `names` only for underscore fields
## Test plan
- [x] `pixi run cmake-build RelWithDebInfo` compiles successfully
- [x] Verify `pixi run test` passes
- [x] Verify `pixi run package` produces correct archives via CMake
release build
- [x] Verify benchmark workflow can be triggered manually via `gh
workflow run benchmark`
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Chores**
* Consolidated CI workflows; some automated triggers converted to manual
invocation
* Standardized shell for workflow steps and removed legacy build
workflow
* Switched packaging/build tasks to a CMake/Ninja flow and updated
artifact paths
* Adjusted ignore rules to include previously-ignored build metadata
* **Documentation**
* Removed XMake-specific build and test instructions; docs now reflect
the CMake-based workflow
* **Style**
* Updated CLI option declaration style (no user-facing flag name
changes)
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
PCH and PCM artifacts are now cached to disk at
`.clice/cache/{pch/,pcm/}` with content-addressed filenames, so they
survive server restarts. Dependency metadata is persisted in
`cache.json` (using eventide serde) with a shared path table for
deduplication.
### Key changes
- **protocol.h**: `output_path` field on `BuildPCHParams` /
`BuildPCMParams` so master specifies where workers write
- **stateless_worker.cpp**: Atomic write via `.tmp` + `fs::rename`;
`CompilationUnit` destroyed before rename to flush the file to disk;
fallback to temp file when `output_path` is empty (unit tests)
- **master_server.h**: `PCMState` struct, `pcm_states` map,
`load_cache()` / `save_cache()` / `cleanup_cache()` methods
- **master_server.cpp**: Cache lifecycle — load from `cache.json` on
startup, save after each PCH/PCM build and on exit; deterministic path
computation (`xxh3` preamble hash for PCH, module name + source path
hash for PCM); stale files (>7 days) cleaned on startup; `cache.json`
uses shared path table to avoid redundant storage of header paths across
entries
- **filesystem.h**: `fs::rename()` helper; `ThreadSafeFS` broadened to
match `.pch` extension instead of `preamble-` prefix
- **tests**: 11 new integration tests covering PCH/PCM persistence,
cross-session reuse, staleness detection, shared preamble dedup, and
restart survival; unit tests updated with `output_path`
### Naming scheme
- **PCH**: `.clice/cache/pch/<016x(xxh3(preamble))>.pch`
- **PCM**:
`.clice/cache/pcm/<module_name>-<016x(xxh3(source_path))>.pcm`
## Test plan
- [x] Unit tests — 448 passed
- [x] Integration tests — 92 passed (including 11 new persistent cache
tests)
- [x] Smoke tests — 1 passed
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
Fix two data races in the stateful worker that caused spurious
"redefinition" errors during rapid edits, and remove a didChange
workaround that is no longer needed after clice-io/eventide#95.
### stateful_worker.cpp
**Compile handler**: move `params` → `doc` field copy **after**
`strand.lock()`. Previously the copy happened before the lock, so a
concurrent Compile request waiting on the strand could overwrite
`doc.text` while `et::queue` was reading it on the thread pool:
```
T1: Compile A → doc.text = text_A → lock → et::queue reads doc.text
T2: Compile B → doc.text = text_B → waits for strand (overwrites!)
T3: et::queue sees text_B instead of text_A → PCH/text mismatch
```
**DocumentUpdate handler**: only mark `dirty`, stop modifying
`doc.text`/`doc.version`. The event loop notification can fire while
`et::queue` work is running on the thread pool — writing `doc.text` from
one thread while reading it from another is a data race.
### master_server.cpp
Remove the `{0,0}-{0,0}` range workaround for whole-document
`didChange`. eventide's variant deserialization now correctly rejects
`TextDocumentContentChangePartial` when the `range` field is absent
(clice-io/eventide#95), so `TextDocumentContentChangeWholeDocument` is
matched as intended.
### protocol.h
Remove `text` field from `DocumentUpdateParams` — the worker no longer
needs it since DocumentUpdate only sets the dirty flag.
### Integration tests (+312 lines)
Extend test_staleness.py from 5 to 14 tests covering document lifecycle:
- `didChange` body edit → recompilation with updated diagnostics
- `didChange` preamble edit → PCH rebuild + clean recompilation
- `didClose` + reopen → compiles fresh from disk
- `didClose` → hover returns None
- `didSave` header → dependent file recompiles
- `didSave` module → CompileGraph dependents invalidated
## Test plan
- [x] 422 unit tests pass (426 on CI with extra test suites)
- [x] 14 integration tests pass locally
- [x] Depends on clice-io/eventide#95 (merged)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Smaller document-update notifications sent to background workers (only
path and version).
* **Bug Fixes**
* Reduced races and unnecessary work between update and compile flows.
* Prevented notifications from overwriting in-memory document text,
improving state consistency.
* Safer concurrent handling to avoid mid-request eviction of active
documents.
* **Tests**
* Added integration tests for staleness, dependency propagation, and LSP
lifecycle.
* Updated unit tests to match revised update behavior.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
Rewrites the template resolver to eliminate
`SubstType`/`CodeSynthesisContexts` dependency, fixing widespread
crashes on real-world C++ code.
### What changed
**Architecture**: replaced double-TreeTransform (PseudoInstantiator +
SubstType) with single-layer design:
- **`SubstituteOnly`** — new lightweight TreeTransform for Phase 2
(typedef expansion + parameter substitution). Does NOT override
`TransformDependentNameType`, breaking the typedef ↔ lookup infinite
cycle.
- **`PseudoInstantiator`** — retains heuristic lookup (the unique value
clang doesn't provide), delegates substitution to `SubstituteOnly`.
**Deleted**:
- `DesugarOnly` class
- `instantiate()` method and all `CodeSynthesisContexts` / `SubstType`
usage
- `state()` / `rewind()` stack management
- `std::abort()` on valid NNS kinds
- `#ifndef NDEBUG` debug flag + `std::print` logging
**Added**:
- `SubstituteOnly` class with depth guard
- `InstantiationStack::findArgument()` — depth/index based parameter
lookup
- CTD→TST resolution for `DependentTemplateSpecializationType` (enables
`__alloc_traits::rebind<T>::other` resolution)
- `active_resolutions` (DNT cycle detection) + `active_ctd_lookups` (CTD
cycle detection via RAII guard)
- Stack frame pollution fix: pop lookup frames before further
`TransformType`
- Pack argument support (single-element forwarding)
- Null safety on all Transform return paths
- Structured `LOG_DEBUG` trace logging with indentation
- `--log-level` / `--test-filter` CLI options for unit test runner
- Bounds checks in `hole()`, `ResugarOnly`, `visitTemplateDeclContexts`
**Tests**: 20 → 36 passing test cases (+5 documented TODOs for known
limitations). New coverage: recursive base classes, multiple
inheritance, typedef chains, CRTP, `remove_reference` partial specs,
`std::map`, `std::basic_string`, pack forwarding.
### Stress test result
```
CDB: llvm-project build (4669 C++ files)
Types resolved: 3,690,190
Types unchanged: 75,907,298
Crashes: 0
```
Before this PR, the same test produced ~52% crash rate (413 crashes in
800 files).
### Known limitations (documented as TODOs)
- NTTP partial specialization matching (`enable_if<true, X>`, `A<X, 0>`)
- Template template parameter deduction
- Non-dependent qualifier nested class templates
(`Outer<int>::Inner<X>`)
- Multi-element pack expansion
- `CXXDependentScopeMemberExpr` lookup (unimplemented)
- Operator-name lookup in dependent contexts
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
## Release Notes
* **New Features**
* Enhanced template resolution with improved cycle detection to prevent
infinite loops
* Better type substitution handling for complex dependent types
* **Bug Fixes**
* Fixed edge cases in template specialization resolution
* Improved null-safety in type transformations
* Enhanced handling of standard library template traits
* **Tests**
* Expanded test coverage for recursive and complex template patterns
* Added validation for standard library type resolution
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
Replace the push-based compilation model with a pull-based (lazy) model
where compilation is driven entirely by feature requests.
### Server core (`master_server.cpp/h`)
- **Remove** `schedule_build()`, `run_build_drain()`, debounce timers,
and `DocumentState` flags (`build_running`, `build_requested`,
`drain_scheduled`)
- **Remove** `debounce_ms` config field
- `didOpen`/`didChange` only update `DocumentState` and mark `ast_dirty`
— no compilation triggered
- `didSave` marks dependent docs dirty via `CompileGraph::update()`,
invalidates PCH hashes, marks **all** open documents `ast_dirty` (header
saves), and queues background indexing
- **Implement** `ensure_compiled(path_id)` — the pull-based entry point
called by `forward_stateful()`/`forward_stateless()` before every
feature request:
1. Fast-path if `!ast_dirty`
2. Compile C++20 module deps via `compile_graph->compile_deps()`
3. Build/reuse PCH via `ensure_pch()` (only attach on success)
4. Send `CompileParams` to stateful worker
5. Publish diagnostics, clear dirty, schedule indexing
6. Generation mismatch → return `false`, keep dirty for retry
- `forward_stateless()` now also calls `compile_graph->compile_deps()`
before stateless requests (completion/signatureHelp)
- Move module-implementation-unit implicit dependency handling into
`resolve_fn` (was duplicated in `run_build_drain` and `ensure_compiled`)
### CompileGraph (`compile_graph.cpp/h`)
- **Add** `compile_deps(path_id)` — compiles all transitive module
dependencies but NOT the file itself (used for plain .cpp files that
`import` modules)
- Unify `compile`/`compile_deps` via `compile_impl(path_id, ancestors,
dispatch_self)` parameter
- `compile_deps` compiles dependencies concurrently via `when_all`
- Extract `finish()` lambda to deduplicate `compiling=false;
completion->set()` cleanup across all exit paths
- Use `std::ranges::remove` instead of legacy `std::remove`
### Test infrastructure (`conftest.py`)
- `open_and_wait()` now sends a hover request to trigger
`ensure_compiled()` (pull-based model requires a feature request to
compile)
- Fix URI handling: send percent-encoded URI on the wire, normalize for
internal lookups, store diagnostics under both raw and normalized URI
keys
- Add `_normalize_uri()` helper using `urllib.parse.unquote`
### Integration tests
- Update all tests for pull-based model: no more waiting on `didOpen`
diagnostics
- `_wait_for_index()` sends hover to trigger compilation before polling
`workspace/symbol`
- `test_hover_save_close` simplified — hover directly triggers
compilation
- `test_save_recompile` and `test_pch_*` wait for fresh diagnostics
after hover-triggered recompilation
### Unit tests (`compile_graph_tests.cpp`)
- Extract `compiled`/`graph` as TEST_SUITE members with
`std::optional<CompileGraph>`
- Extract `execute(callback)` helper to deduplicate event_loop
boilerplate
- Add 8 new `compile_deps` tests: no-deps, single dep, chain, diamond,
failure, plain-cpp, concurrent dedup, resolve-once
- Remove redundant `inline` on file-scope helpers
## Test plan
- [x] Unit tests: 426 passed, 5 skipped
- [x] Smoke tests: 1/1 passed
- [x] Integration tests: 69 passed, 0 failed, no hangs
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
## Summary
Implement the complete index system for cross-file LSP features. This
adds persistent two-tier indexing (ProjectIndex + per-file MergedIndex
shards), background indexing triggered on idle, and index-based query
handlers for major LSP requests.
### Index Data Layer (`src/index/`)
- **TUIndex**: Add binary serialization/deserialization via FlatBuffers,
enabling IPC between stateless worker and master server
- **ProjectIndex**: Add symbol name/kind storage, `PathPool` path
normalization (backslash -> forward slash), and binary persistence
- **MergedIndex**: Add `content` field to store file content for
reliable offset<->position mapping; add `removed` bitmap for garbage
collection of deleted entries; filter removed IDs in `lookup()` queries
- **schema.fbs**: Add TUIndex tables, `Symbol.name` field,
`MergedIndex.removed` bitmap and `MergedIndex.content` string
### Server (`src/server/`)
- **Background indexing**: Idle-triggered coroutine dequeues files from
CDB, dispatches `IndexParams` to stateless workers, merges returned
`TUIndex` into ProjectIndex/MergedIndex, and persists to `.clice/index/`
- **Index persistence**: `save_index()` / `load_index()` for startup
restoration; only rewrites shards flagged `need_rewrite()`
- **LSP handlers**:
- `textDocument/definition` -- index-first lookup with stateful worker
fallback
- `textDocument/references` -- cross-file reference query via index
- `callHierarchy/prepare`, `incomingCalls`, `outgoingCalls` --
Caller/Callee relation traversal
- `typeHierarchy/prepare`, `supertypes`, `subtypes` -- Base/Derived
relation traversal
- `workspace/symbol` -- case-insensitive substring search over
ProjectIndex symbols
- **Stateless worker**: Add `Index` request handler that builds
`TUIndex` from compiled AST and returns serialized data
- **Config**: Add `enable_indexing` (default true) and `idle_timeout_ms`
(default 3000ms)
### Fixes and Cross-platform
- **ElaboratedType handling** in `decl_of()` for correct Base/Derived
relation emission
- **Windows path normalization** in `PathPool::intern()` and
`ProjectIndex::from()` (backslash -> forward slash)
- **`.gitattributes`**: Force LF in `tests/data/**` to prevent CRLF
byte-offset mismatches on Windows CI
- **Test fixture**: Clean `.clice/` before each test for hermetic index
state
### Tests
- **370-line** `index_query_tests.cpp`: unit tests for occurrence
lookup, relation queries, content retrieval, removed bitmap filtering
- **282-line** `test_index.py`: E2E integration tests for
GoToDefinition, FindReferences, CallHierarchy
(prepare/incoming/outgoing), TypeHierarchy
(prepare/supertypes/subtypes), WorkspaceSymbol
- Updated existing MergedIndex and ProjectIndex tests for new schema
fields
## Test plan
- [x] 414 C++ unit tests pass (including new IndexQuery, MergedIndex,
ProjectIndex tests)
- [x] 69 Python integration tests pass (including 10 new index feature
tests)
- [x] CI green on Linux, macOS, Windows
- [ ] Manual smoke test with VSCode extension
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
- Add `ensure_pch()` helper to MasterServer that builds/reuses
precompiled headers via stateless workers, with preamble hash-based
staleness detection (xxh3_64bits)
- Fix `BuildPCHParams` to carry `preamble_bound` so the stateless worker
truncates content at the preamble boundary (fixes redefinition errors
when PCH included full file)
- Wire PCH into both `run_build_drain` (stateful compile path) and
`forward_stateless` (completion/signatureHelp path)
- Add PCH state cleanup on `didClose` and hash invalidation on `didSave`
## Test plan
- [x] 398 unit tests pass (including 6 new PCH tests: PreambleHash x3,
PCHWorker x2, BuildPCHRequest assertion)
- [x] 5 new integration tests pass (`test_pch.py`: diagnostics on open,
body edit recompile, no-include file, hover with PCH, completion with
PCH)
- [x] 21 existing integration tests pass unchanged
- [x] Build succeeds with 0 errors
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Precompiled header (PCH) caching to speed compilations and reduce edit
latency
* Automatic attachment of cached PCH to compile requests, improving
hover and completion responsiveness
* Module-aware completions expanded to include available module
artifacts from other files
* **Bug Fixes**
* PCH cache cleared on file close; saving now triggers broader PCH
invalidation to prevent stale PCH use
* **Tests**
* Added unit and integration tests exercising PCH build, reuse, and
editor interactions
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
- **CMake-based CDB generation for module tests**: Replace hand-written
compile_commands.json with CMakeLists.txt (CMake 3.28 `FILE_SET
CXX_MODULES`) in all 26 `tests/data/modules/*/` directories. CDB is
generated on-the-fly via `cmake -G Ninja` during test setup.
- **`@pytest.mark.workspace()` decorator**: Introduce a marker + fixture
pattern so tests declare their workspace via decorator and receive a
resolved `workspace` path. The fixture auto-generates CDB when a
CMakeLists.txt is present.
- **`CliceClient` helper methods**: Add `initialize()`, `open()`,
`wait_diagnostics()`, and `open_and_wait()` to reduce boilerplate across
all test files.
- **Use `asyncio_mode = "auto"`**: Switch from `@pytest_asyncio.fixture`
+ `@pytest.mark.asyncio` to `@pytest.fixture` + auto mode for proper
Pylance type inference on fixtures.
- **Test cleanup**: Remove redundant section separators and docstrings,
delete `tests/pyproject.toml` (config moved to `pytest.ini`).
- **Format task**: Add `.cppm` to `format-cpp` glob pattern.
- **CI fix**: Disable `CMAKE_CXX_SCAN_FOR_MODULES` and prefer pixi
clang++ to fix macOS CI where CMake rejects module scanning.
## Test plan
- [x] All 26 module test directories have CMakeLists.txt with FILE_SET
CXX_MODULES
- [x] generate_cdb() produces valid compile_commands.json with module
flags
- [x] Integration tests pass locally
- [ ] CI passes on all platforms (Linux, macOS, Windows)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Tests**
* Unified fixtures and client workflow: new init/open/wait helpers,
workspace marker support, bounded diagnostics waiting, CMake-based
compilation-database generation, and directory-backed temp-file
workflows; enabled asyncio test mode.
* **Chores**
* Added many C++20 module test projects and test data; removed prior
test pyproject in favor of pytest config; updated formatter to include
.cppm files.
* **Style**
* Reformatted many module/source implementations to consistent
multi-line function bodies.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
- **Use `Tester` as fixture base** for all test suites that need
compilation, replacing `TesterFixture` and removing redundant
`tester.clear()` calls (eventide zest now creates fresh instances per
TEST_CASE)
- **Remove local `Tester` variables** in `compilation_tests`,
`template_resolver_tests`, `selection_tests` — use inherited fixture
members directly
- **Normalize helper naming**: `expect_xxx` → `EXPECT_XXX`,
`go_to_definition` → `GO_TO_DEFINITION` for consistency
- **Extract shared `test/cdb_helper.h`**: deduplicate `CDBEntry`,
`json_escape`, `build_cdb_json` from `dependency_graph_tests` and
`compile_graph_integration_tests`
- **Add new test files/cases**: `project_index_tests.cpp`, expanded
`tu_index_tests`, `merged_index_tests`, `compilation_tests`
## Test plan
- [x] All existing unit tests pass
- [x] New index tests (TUIndex, MergedIndex, ProjectIndex) pass
- [x] Compilation tests (PCH, PCM, stop) pass
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Tests**
* Standardized test fixtures and helper naming, moved suites to a shared
fixture, and unified in-memory VFS and compile flows.
* Added broad new coverage: indexing, project indexing, compilation/PCH,
diagnostics, semantic features, and many targeted unit cases.
* Introduced a small compile-database helper and improved driver-style
test compilation paths.
* **Chores**
* Consolidated and reorganized test utilities and tester APIs for easier
maintenance and reuse.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
Initial integration of `CompileGraph` (#375) into `MasterServer`,
enabling basic end-to-end C++20 module support: on-demand PCM building,
dependency-ordered compilation, cascade invalidation on save, and
diagnostic integration.
This is a **first-pass implementation** — the core pipeline works, but
there are known areas for follow-up:
- PCM files go to system temp dir instead of `.clice/cache/`; no disk
cleanup on invalidation
- `run_build_drain` scans imports itself rather than delegating fully to
CompileGraph
- No incremental/partial rebuild (full PCM rebuild on any change)
- Cycle detection is tested at unit level but integration-level coverage
is minimal
## Changes
### Module dependency compilation (`master_server.cpp`)
Before sending a file to the stateful worker, `run_build_drain` now:
1. Scans imports via `scan_precise()` to discover module dependencies
2. Compiles each dep through `compile_graph->compile()`, which
recursively builds transitive PCMs
3. Handles implementation units — `module M;` implicitly needs the
interface PCM
4. Passes all built PCMs to the stateful worker, excluding the file's
own PCM
5. Skips compile on dep failure and resets `build_running` /
`drain_scheduled`
6. Re-lookups iterators after `co_await` to avoid use-after-invalidation
### Cascade invalidation (`didSave` / `didClose`)
- `didSave`: calls `compile_graph->update()` to mark transitive
dependents dirty, removes stale PCM paths, schedules rebuilds for open
dirtied files
- `didClose`: cancels in-flight compilations for the closed file
### Other fixes in this PR
- Debounce timers switched to `shared_ptr` to prevent use-after-free
when `didClose` destroys the timer mid-wait
- `fill_compile_args` returns `bool`; callers handle empty CDB
gracefully
- Adapt all `PositionMapper` call sites to the new `optional` return API
from eventide
## Test plan
- [x] 25 C++ unit tests for CompileGraph (cycles, partial failure,
cancel, update, empty graph)
- [x] 24 C++ integration tests with real clang PCM compilation
- [x] 3 worker-level module tests (BuildPCM, PCM-dependent compile,
multi-module)
- [x] 26 Python LSP integration tests (single module through circular
deps, hover, error diagnostics)
- [x] 371 unit tests + 54 integration tests pass
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
Add `CompileGraph`, a pull-based async scheduler for C++20 module
compilation. When a file is compiled that imports modules, the graph
automatically resolves, builds, and caches PCM dependencies in the
correct order before the main compile proceeds.
## Design
### Data model
Each compilation unit (`CompileUnit`) tracks:
- `dependencies` / `dependents` — forward and reverse dependency edges
- `dirty` / `compiling` — current state flags
- `generation` — monotonic counter incremented by `update()`, used for
ABA-safe stale detection
- `source` + `completion` — cancellation token source and completion
event for cooperative async
### `compile(path_id)` — pull-based compilation
Lazily resolves dependencies (via `resolve_fn`) on first access, then
recursively compiles all transitive deps before dispatching the unit
itself:
- **Concurrent**: sibling deps compiled in parallel via `when_all`
- **Dedup**: diamond dependencies (A->B->D, A->C->D) — the second branch
waits on the first via `completion.wait()` instead of re-compiling
- **Cycle detection**: per-branch `ancestors` set (passed by value)
catches direct cycles; `has_wait_cycle()` BFS catches cross-branch
cycles (e.g. `1->{2,3}, 2->3, 3->2`) that would deadlock at
`completion.wait()`
- **Cancellation**: all `co_await` wrapped with `with_token()`, so
`update()` can cancel in-flight compilations immediately
- **Generation check**: captures generation counter before `co_await`;
if `update()` bumped it during dispatch, the result is discarded (unit
stays dirty)
### `update(path_id)` — cascade invalidation
BFS along `dependents` edges to mark the entire reverse-transitive
closure as dirty. For the source node, clears `resolved` and dependency
edges so they are re-scanned on next compile. Cancels any in-flight
compilations via `source->cancel()`.
## Test plan
22 unit tests covering:
- [x] No deps, single dep, chain, diamond (compile ordering + dedup)
- [x] Update invalidation, cascade through chains and diamonds
- [x] Re-resolution after update (deps can change)
- [x] Stale back-edge cleanup
- [x] Direct cycle detection (A->B->A)
- [x] Cross-branch cycle detection (when_all deadlock case)
- [x] Self-loop
- [x] Dispatch failure propagation
- [x] cancel_all + recompile
- [x] Update during in-flight compile (cancellation + generation check)
- [x] CI green on Linux, macOS, Windows
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
## Summary
- Add `scan_module_decl()` — a lightweight preprocessor-based fallback
that resolves module declarations inside `#if`/`#ifdef` conditionals.
When `scan()` detects `need_preprocess=true`, this function runs clang's
preprocessor to evaluate conditions and extract the actual module name.
It stops lexing as soon as the module declaration is found, making it
much cheaper than `scan_precise()`.
- Integrate the fallback into `scan_dependency_graph()` for wave 0
source files, so conditional module declarations (e.g. `#ifdef
USE_MODULES / export module M; / #endif`) are correctly registered in
the dependency graph.
- Add comprehensive test cases covering all C++20 module declaration
forms from cppreference, including `scan_module_decl()` tests for
conditional resolution and `scan_precise()` tests for module import
semantics.
## Test plan
- [x] All 310 unit tests pass (0 failures, 9 skipped)
- [x] `scan()` tests cover: primary interface, implementation, dotted
names, partitions, GMF, conditional module declarations, private module
fragment
- [x] `scan_module_decl()` tests cover: basic, conditional with `-D`,
conditional with `#if` expression, GMF with conditional, implementation
unit, dotted name, partition, no-module file
- [x] `scan_precise()` tests cover: named import, multiple imports,
dotted import, partition import, export-import, export-import partition,
implementation import, GMF with import, mixed includes/imports,
no-module file
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Improved detection of module declarations hidden by conditional
compilation via a lightweight fallback scan. Resolved module vs.
interface classification is cached to avoid repeated work and is used
consistently in dependency mapping.
* Better handling and classification of module imports, partitions, and
global-fragment includes when building module relationships.
* **Tests**
* Added comprehensive unit tests covering module declaration extraction,
fallback resolution under preprocessor guards, imports, partitions,
includes, and macro-driven cases.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
