Compile the Windows half of the tree in CI - #36
Merged
Merged
Conversation
Everything under cfg(target_os = "windows") was invisible to CI: src/service.rs — the entire Windows Service entry point — plus the windows-service and tracing-appender dependencies and every Windows branch in the platform modules. test.yml runs ubuntu-only, so the only thing that ever compiled that code was release.yml, on a tag, after the decision to ship had already been made. The gap had teeth. windows-service 0.7 -> 0.8 (#25) could not be reviewed at all: ubuntu CI reported green without compiling a line of what the bump affects, and cross-compiling to x86_64-pc-windows-msvc fails locally in ring's build script with no MSVC toolchain available. For a binary that installs onto customers' own machines as a Windows Service, release time is the wrong moment to discover that. Adds a `windows` job on windows-latest running build + clippy. A full build rather than `check`, because linking is part of what can break when windows-service resolves Service Control Manager entry points — a check would type-check those and never link them. No `cargo test` on Windows: `--all-targets` already type-checks the test code, which is what catches a Windows-only compile regression. Running the suite there means triaging tests that assume POSIX paths and a present ffmpeg — worth doing, not worth coupling to this. Also drops number_prefix from the cargo-audit comment's list of unmaintained crates; indicatif 0.18 stopped depending on it in #35. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The gap
Everything under
cfg(target_os = "windows")is invisible to CI:src/service.rs— the entire Windows Service entry pointwindows-serviceandtracing-appenderdependenciestest.ymlruns ubuntu-only. The only thing that has ever compiled thatcode is
release.yml, on a tag — after the decision to ship was made.Why it matters now
windows-service0.7 → 0.8 (#25) cannot be reviewed under the currentsetup. Ubuntu CI reports green without compiling a single line of what the
bump affects. I tried to close the loop locally and
cargo check --target x86_64-pc-windows-msvcfails inring's build scriptwith no MSVC toolchain available (mingw isn't installed either).
This repo's own
dependabot.ymlsays Cargo bumps get reviewed because"CameraNode ships as a BINARY onto customers' own hardware… it sits on
machines we do not control." For a binary that registers with the Windows
Service Control Manager, release time is the wrong moment to find out.
The change
A
windowsjob onwindows-latest:cargo build --lockedcheck— linking is part of what breaks whenwindows-serviceresolves SCM entry points, andcheckwould type-check them and never link themcargo clippy --all-targets --locked-D warnings, matching the ubuntu jobNo system-dependency step: the MSVC toolchain on
windows-latestalreadycovers rusqlite's bundled SQLite and the zip crate's bzip2-sys.
No
cargo teston Windows, deliberately.--all-targetsalreadytype-checks the test code, which is what catches a compile regression.
Actually running the suite there means triaging tests that assume POSIX
paths and a present ffmpeg — worth doing, but not worth coupling to this
change.
Also
Drops
number_prefixfrom the cargo-audit comment's list of unmaintainedcrates.
indicatif0.18 stopped depending on it in #35, so the comment wasalready stale.
Unblocks #25.
🤖 Generated with Claude Code