Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions quest/m0/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -32,5 +32,6 @@ edge, which embeds moq-relay; it moved there on 2026-09-28.
- [Publish the release line's drain() wrappers](/quest/m0/ffi-shutdown-backport.md) - Dart `moq` 0.1.1 reaches pub.dev from `release`
- [CI cost and latency](/quest/m0/ci/README.md) - cloud CI runs only in GitHub's merge queue, one tested entry at a time, expensive checks run nightly and gate `release`, caches move to R2, and runners are chosen last
- [More tests under load](/quest/m0/test-flakes-2/README.md) - load-only and queue failures, one quest per flake, fixed at the cause, since each one drops a queued PR
- [just check passes on macOS](/quest/m0/macos-check.md) - the version test, the doc re-export link test, and the carla bridge docs pass on a macOS host
- [Agent build output can't fill the desktop's disk](/quest/m0/agent-disk-budget.md) - finished agent checkouts reclaim their mbx targets automatically, so parallel agent builds stop filling the root disk
- [MoQ builds against rind 0.4 and its corrected adapters](/quest/m0/rind-adapters.md) - the calls rind 0.4 and its adapter patch releases change, adapted the day they publish
55 changes: 55 additions & 0 deletions quest/m0/macos-check.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,55 @@
# [S] just check passes on macOS

## Goal

`just check` on a clean `main` checkout passes on a macOS host inside the Nix
dev shell, as it already does on Linux. Three failures block it today, all

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Three failures may be one short: #8392 also records a macOS-only clippy failure, an unfulfilled expect(dead_code) at rs/moq-bench/src/host.rs:43. That expectation is gated on not(target_os = "linux") and moq-bench's features never touch host.rs, so just check --all, or any selection that includes moq-bench (a default workspace member), should report it on macOS. If the Done condition is a fully green macOS just check, does this list need a fourth item?

reproduced on `origin/main` by
[moq-dev/moq#8392](https://github.com/moq-dev/moq/pull/8392) and
[moq-dev/moq#8394](https://github.com/moq-dev/moq/pull/8394):

- `test gh`: `sh/version.test.py`'s
`test_cut_promotes_only_published_library_dependencies_and_preserves_exemptions`
(`'0.0.1' != '0.22.0'`) and `test_release_plz_keeps_manual_cut_despite_breaking_commits`,
with release-plz 0.3.149.
- `check js`: `doc/.vitepress/api/index.test.ts`, "a re-export from another
workspace package links to its owner"; the rendered link holds
`/private/var/folders/...`.
- `check rs`: `RUSTDOCFLAGS="-D warnings" cargo doc --no-deps -p moq-carla-bridge`
reports unresolved links to `Bridge`, `Source`, and `protocol`.

Non-goal: a macOS runner on the PR gate (rejected in
[macos.md](/quest/m0/nightly/macos.md)).

## Plan

Planned 2026-11-09 (maintainer, /quest-complete follow-up of #8392 and #8394).
Milestone: βœ… m0 (rejected: a later milestone).

Causes, read from source on Linux (no macOS host was available):

- carla bridge, **confirmed**: `demo/carla/bridge/src/lib.rs` puts
`#![cfg(target_os = "linux")]` after its `//!` docs, so on macOS the crate is
empty and the intra-doc links dangle. Either gate the links too or build the
crate (and its docs) only on Linux, e.g. by keeping it out of the macOS doc
selection.
- doc re-export link, **guess**: `tmpdir()` is `/var/folders/...` on macOS,
a symlink to `/private/var/...`. `index.ts` computes package paths with
`relative(root, dir)`; if one side is resolved through the symlink (by
TypeScript or the workspace scan) and the other is not, the link becomes a
`../../private/var/...` path. Confirm by checking which side resolves, then
realpath both sides in `index.ts` (production paths can be symlinked too),
not only in the test.
- `sh/version.test.py`, **guess**: the Linux dev shell provides release-plz
0.4.1 (`pkgs.release-plz` in `flake.nix`), while the failing host ran
0.3.149, so it likely picked up a non-Nix `release-plz` (e.g. from
`~/.cargo/bin`) ahead of the dev shell. The flake uses one nixpkgs for both
systems, so a different darwin build is less likely. If it is PATH, the
test should fail loud on a release-plz older than it supports instead of
asserting wrong versions; if the dev shell's darwin release-plz really is
0.3.149, align it.
Comment on lines +43 to +50

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The PATH mechanism cannot be what happens on macOS: flake.nix:177-179 includes pkgs.release-plz only outside Darwin ("Marked broken on Darwin in nixpkgs"), so a macOS dev shell ships no release-plz to shadow. What it finds instead is just install's unpinned cargo install ... release-plz (justfile:98), which leaves ~/.cargo/bin on PATH and is the documented non-Nix path. That makes "align the dev shell's darwin release-plz" an override of a package nixpkgs marks broken rather than a one-line fix; the fail-loud option is the one that matches the tree.

Technical details
# release-plz cause on macOS

## Affected sites
- `quest/m0/macos-check.md:43-50` β€” the `sh/version.test.py` cause and its two options.
- `flake.nix:177-179` β€” `pkgs.release-plz` sits inside `pkgs.lib.optionals (!pkgs.stdenv.hostPlatform.isDarwin)`, commented "Marked broken on Darwin in nixpkgs, but builds fine on Linux."
- `justfile:98` β€” `cargo install --locked cargo-shear cargo-sort cargo-semver-checks release-plz`, unpinned, into the user's cargo bin.
- `doc/setup/index.md:34-36` β€” `just install` is the documented setup path without Nix.

## Required outcome
- The plan states that a macOS dev shell provides no release-plz at all, so the version on a macOS host is whatever the host installed.
- The remaining option matches that state: `sh/version.test.py` (or `just install`) pins or requires the release-plz version the tests need and fails loud on an older one, instead of treating a Darwin dev-shell build as an alignment target.

## Open questions for the human (optional)
- Should the flake override release-plz for Darwin so the dev shell pins it there too, or is an unpinned host install acceptable? The answer decides whether the quest also edits `flake.nix`.


Verify each fix on a macOS host (a maintainer machine or a manual Platform
dispatch), and on Linux.

Public API: none. Wire: none.