[workspace-split] P1.3: extract cc-types leaf crate (partial)

[yaohaowei0914]

Advances #70 (Phase 1 leaf extractions). Follow-up to #79.

Summary

Extracts the three pure-leaf files from src/types/ into a new cc-types
workspace crate:

• message.rs → crates/cc-types/src/message.rs
• state.rs → crates/cc-types/src/state.rs
• transitions.rs → crates/cc-types/src/transitions.rs

app_state.rs, tool.rs, and config.rs stay in the root crate. Per the
routine's analysis in #70 and the cc-types note at the top of #79, those
three are not true leaves:

• types::app_state::AppState references crate::teams::types,
  crate::ui::status_line, crate::config::settings, and the now-extracted
  cc_keybindings::KeybindingRegistry.
• types::tool::ToolUseContext references crate::ipc::agent_channel::AgentSender.
• types::config pulls ToolUseContext and Tools from tool.

Extracting these would require either moving teams / ui / config / ipc out
first (Phase 5+) or rewriting the struct layouts with trait objects / generics
— much larger than a leaf move. Out of scope for Phase 1.

Call-site compatibility

The root crate's types/mod.rs now re-exports the three moved modules:

pub use cc_types::{message, state, transitions};

pub mod app_state;
pub mod config;
pub mod tool;

All existing crate::types::message::* / crate::types::state::* /
crate::types::transitions::* paths — roughly 80 call sites across
api/, compact/, engine/, query/, session/, tools/ etc. —
keep resolving with zero edits.

Why a partial extraction instead of deferring the whole thing

#79 suggested deferring all of cc-types "until teams / ui / config / ipc
have migrated". That would cost the main goal of Phase 1 — incremental-build
wins from the stable message / state / transitions types (referenced from
~80 files, changed rarely). Splitting out the clean subset now captures most
of the benefit immediately; the dirty subset still rides along with the
remaining root crate until its dependencies move.

Diff shape

Change	Notes
crates/cc-types/Cargo.toml (new)	deps: serde, serde_json, uuid, chrono (all workspace = true)
crates/cc-types/src/lib.rs (new)	3 pub mod declarations, comment explaining the partial boundary
3 × git mv	rename-tracking preserved, no content changes to message/state/transitions
Cargo.toml	one new workspace dep entry: cc-types = { path = "crates/cc-types" }
crates/claude-code-rs/Cargo.toml	one new line: cc-types = { workspace = true }
crates/claude-code-rs/src/types/mod.rs	pub mod {message,state,transitions} → pub use cc_types::{message, state, transitions}; (plus the three remaining local modules)

Verification

• cargo build : ok, 2 warnings (pre-existing
  web::handlers::session_id dead_code, byte-identical on rust-lite@ab8a2fc).
• cargo build --release : ok, same 2 warnings.
• cargo test -p cc-types : 0 tests moved with the files; the crate compiles
  clean with zero warnings.
• cargo test --bin claude-code-rs --offline -- --test-threads=1 :
  1794 / 1794 pass — matches the baselines reported in [workspace-split] Phase 0 + 1 (partial): Workspace skeleton, leaf extractions, CI adaptations #79 for the
  cc-keybindings and cc-observability extractions.
• cargo test --offline -- --test-threads=1 across the workspace:
  1794 / 1794 in claude-code-rs, 51 / 51 in cc-keybindings,
  10 / 11 in cc-observability. The single failure is
  cc-observability::sink::tests::off_mode_skips_redaction — the same
  pre-existing redact_value bug [workspace-split] Phase 0 + 1 (partial): Workspace skeleton, leaf extractions, CI adaptations #79 documented, unchanged by this PR.
• Smoke: claude-code-rs --version prints claude-code-rs 0.1.0.

Phase 1 checklist status (#70)

• crates/cc-keybindings/ ([workspace-split] Phase 0 + 1 (partial): Workspace skeleton, leaf extractions, CI adaptations #79, 9fd7229)
• crates/cc-observability/ ([workspace-split] Phase 0 + 1 (partial): Workspace skeleton, leaf extractions, CI adaptations #79, ef45df3)
• crates/cc-types/ — partial, this PR
  • message.rs / state.rs / transitions.rs extracted
  • app_state.rs / tool.rs / config.rs deferred to Phase 5+
• src/types/{message,state,transitions}.rs removed from root crate
• src/types/{app_state,tool,config}.rs still present (by design)
• All tests pass (1794 / 1794 bin tests; known pre-existing flakes unchanged)

Test plan

• cargo build
• cargo build --release
• cargo test --bin claude-code-rs --offline -- --test-threads=1
• cargo test -p cc-types
• claude-code-rs --version smoke test
• CI across Linux / Windows (left to the workflow)

Known caveats / follow-ups (unchanged from #79)

• cc-observability::sink::tests::off_mode_skips_redaction still flakes —
  pre-existing, tracked in [workspace-split] Phase 0 + 1 (partial): Workspace skeleton, leaf extractions, CI adaptations #79's description. Not touched here.
• CI-env-fragile Windows tests from [workspace-split] Phase 0 + 1 (partial): Workspace skeleton, leaf extractions, CI adaptations #79's "Known failing CI tests"
  section remain open.
• The design doc's "leaf" claim for cc-types should be amended — this
  PR's lib.rs comment and commit message capture the revised reality,
  but the design doc file itself has not been updated here to keep the
  diff surface minimal.

Refs #70. Part of #68.

🤖 Generated with Claude Code