[teamleaderleo]

Summary

build-windows / build x86_64-pc-windows-gnu fails for every PR whose change routes it on. It runs cargo build -p cmux-tui --target x86_64-pc-windows-gnu, and that fails with could not compile cmux-automations (lib) due to 35 previous errors. The errors are std::os::unix, std::os::fd, and libc::localtime_r / gmtime_r / timegm / kill / pid_t (seen on #28678 and #28759, for example job 124460652801).

cmux-tui uses cmux-automations only from cli/automation.rs. That module is already #[cfg(unix)], as is its run_if_requested call in cli.rs. Only the dependency was declared for every target. This PR moves it to [target.'cfg(unix)'.dependencies], so a Windows build no longer compiles the crate. Nothing changes on macOS or Linux, and Cargo.lock is unchanged.

Base push runs don't run build-windows, so this red never shows on base. next-autoland then counts it as each PR's own red, which holds back otherwise-ready PRs.

Testing

Red: base, as build-windows on #28678 and #28759 at their heads (35 errors in crates\cmux-automations). Green: this PR's own build-windows job.

Observed / Not run

Observed: the Windows errors in those jobs' logs, all in cmux-automations, and that cmux-tui reaches the crate only through cfg(unix) code (git grep cmux_automations).
Not run: a local Windows cross-build (this VM has no mingw toolchain); this PR's CI build-windows job runs it.

Changelog

none

Agent session: agent session

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Fixes the Windows build of cmux-tui by making cmux-automations a unix-only dependency.

cmux-automations uses std::os::unix and libc unix calls, so compiling it for x86_64-pc-windows-gnu failed with 35 errors. cmux-tui only reaches the crate from cli/automation.rs, which is already #[cfg(unix)]. Moving the dependency to [target.'cfg(unix)'.dependencies] means Windows builds skip the crate entirely. macOS and Linux builds are unchanged, and Cargo.lock is untouched.

^{Written for commit 365416c. Summary will update on new commits.}

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review 592cf2d: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr29075 -f ref=lane/ao-win-automations -f platform=macos -f backend_mode=local

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.4119158082902 (5ada6804dc)

[teamleaderleo]

Review (codex gpt-6-astra high): SHIP at 592cf2d

Every cmux_automations reference in cmux-tui is in src/cli/automation.rs, whose module and dispatch are cfg(unix). Tests and build.rs have none, and no other crate in the Windows jobs depends on cmux-automations. On unix the dependency is unchanged. No blocking or follow-up findings.

[teamleaderleo]

Green: build-windows / build x86_64-pc-windows-gnu passes at 592cf2d (job 124479332637). On #28678 and #28759 it failed with 35 errors in cmux-automations. The only other red is release-notary, which also fails on base.

[teamleaderleo]

Review (codex gpt-6-astra high): SHIP at 4740ddd

Carried from the SHIP at 592cf2d. The new head is a conflict-free merge of feat-cmux-next, and the PR's diff against base is still the same one file, cmux-tui/crates/cmux-tui/Cargo.toml.

[austinywang]

Review (review_pr26612, independent correctness review): SHIP at 4740ddd

No correctness findings. The sole diff moves cmux-automations into cfg(unix) dependencies. Every direct use is in cli/automation.rs; both that module and its CLI dispatch are already gated by cfg(unix). Unix availability is preserved, and Windows no longer builds an unused Unix-only crate.

Executed at this exact head: Windows job completed its Build libghostty-vt + cmux-tui (Windows GNU) step successfully at 09:21:14 UTC; the acpmux Windows build and startup passed too. The Windows job's remaining runtime long-path checks were still running when reviewed; this receipt does not claim those passed. No local builds or tests were run.

This shared repair removes the inherited Windows compiler failure blocking #26612 and other cmux-tui PRs. The release-notary guard failure is also present on base.

[cmux-bot]

cmux-bot approves 4740ddd: an independent review passed this exact head (austinywang: Review (review_pr26612, independent correctness review): SHIP at 4740ddd). A push needs a new review; a later BLOCK at this head dismisses this approval.

[austinywang]

Review (review_pr26612, independent correctness review): SHIP at 365416c

Fresh review of the full diff and incoming base interactions found no correctness defects. The PR still changes only Cargo.toml, moving cmux-automations to cfg(unix) dependencies. Incoming Rust changes do not alter automation callers or their Unix guards; Unix availability is preserved and Windows skips the unused Unix-only crate.

I merged the current base through GitHub update-branch with the expected previous head, preserving both histories. Autoland had classified the previous release-guard failure as fixed on base but was waiting on a queued compile before updating the branch. This review independently checked the new head against base 07088897284b6637a4ddb5763e06ca1601d4aae2.

Observed: the previous-head full Windows job passed, including compilation, long paths, process exit, key handoff and ConPTY resize. Current-head Windows and cargo jobs are still running; no exact-new-head build or runtime pass is claimed. No local builds/tests ran.