[schickling-assistant]

Problem

Relative cwd values such as . were copied into run metadata. After the daemon vanishes, worktree-reclaim tools cannot classify those records against an absolute worktree path.

Change

• Validate the caller's original cwd spelling first, preserving existing existence/access checks and error text.
• Resolve relative paths lexically with std::path::absolute, then freeze that cwd in the config before metadata publication or child creation, in both ordinary and operation-mode launches.
• Leave absolute cwd strings unchanged, including symlink paths; no canonicalization.
• Add a real-daemon regression for . and an absolute symlink path, plus validation-error coverage and README documentation.
• No changes to status computation, GC, or existing records. This repo has no changelog; the behavior is documented in README.

Verification

Checks ran on Linux through the repository's nix develop shell with heavy-work admission. Final head: c2d8ac40d312b9043e2266448c3019540e452111.

Final-head raw results:

CHECK RESULT: cargo clippy --workspace --all-targets: exit 0
test launch_stores_an_absolute_cwd_without_resolving_symlinks ... ok
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 30 filtered out; finished in 0.04s
CHECK RESULT: cargo test -p pty --test daemon_lifecycle -- --exact --test-threads=1 launch_stores_an_absolute_cwd_without_resolving_symlinks: exit 0
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 57 filtered out; finished in 0.00s
CHECK RESULT: cargo test -p pty --bin pty daemon::env::tests:: -- --test-threads=1: exit 0

Formatting (repo Nix shell, rustc 1.97.0 / rustfmt 1.9.0):

CHECK RESULT: cargo fmt --all -- --check: exit 1
CHECK RESULT: scoped rustfmt --check: exit 1

The scoped command was rustfmt --check --edition 2024 --config skip_children=true on the four changed Rust files. Every reported formatting hunk is in existing, unchanged code; none is in this change's added/modified lines. Baseline formatting was not rewritten. CI treats formatting and clippy as reporting-only.

The complete affected-crate run on 30492ead7ce4180e7037be5bacab3b369db0c81d reported:

CHECK RESULT: cargo clippy --workspace --all-targets: exit 0
CHECK RESULT: cargo test -p pty --no-fail-fast -- --test-threads=1: exit 101

It failed four binaries. To distinguish regressions from baseline failures, each was also run on origin/main (d50ad1b3c839e49c8115e1d9cc93048e6252e090) on the same host and Nix shell, with --test-threads=1:

Binary	Branch run	origin/main comparison	Classification
daemon_geometry	3 passed, 3 failed	3 passed, 3 failed; same WINCH-trap failures and listener EINVAL	Fails on main too (pre-existing)
daemon_lifecycle	28 passed, 3 failed	28 passed, 2 failed; same shutdown/backstop failures	Two fail on main too (pre-existing). Third was new-test fixture cleanup, fixed; final regression passes above.
daemon_roles	14 passed, 2 failed	14 passed, 2 failed; same redraw/READY failures and listener EINVAL	Fails on main too (pre-existing)
interactive	6 passed, 1 failed	6 passed, 1 failed; same deleted-picker missing-file failure	Fails on main too (pre-existing)

Raw baseline results:

CHECK RESULT: MAIN cargo test -p pty --test daemon_geometry -- --test-threads=1: exit 101
CHECK RESULT: MAIN cargo test -p pty --test daemon_lifecycle -- --test-threads=1: exit 101
CHECK RESULT: MAIN cargo test -p pty --test daemon_roles -- --test-threads=1: exit 101
CHECK RESULT: MAIN cargo test -p pty --test interactive -- --test-threads=1: exit 101

Earlier implementation check on efb192cf47bb84c5780c861480c297cbe43ab3a9 exposed a missing mutable worker binding, fixed in the next signed commit:

CHECK RESULT: cargo clippy --workspace --all-targets: exit 101
CHECK RESULT: cargo test -p pty --no-fail-fast -- --test-threads=1: exit 101

NOT RUN: complete affected-crate suite again after the test-only fixture correction; full workspace tests; release build; macOS checks; record migration/reclaim (out of scope).

Complexity and concerns

No new dependency or abstraction; a shared launch-time helper serves both launch modes. Existing formatting drift and the reproduced main-branch test failures remain visible above.

Friction

The regression initially reused a registry root that the daemon test harness deletes on drop; each case now owns a separate root. No production behavior was changed to accommodate the fixture.

Scope

No existing run records were modified. No migration or reclaim operation is included.

Posted on behalf of @schickling

field	value
agent_identity	unknown
session	unknown
agent_persona	unknown
agent_supervisor	unavailable
agent_tool	unknown
agent_tool_version	unknown
agent_runtime	unknown
agent_model	unknown
worktree	unknown