fix(docker): include experiments/ in builder so workspace manifest parses#235
Merged
Conversation
…rses The Dockerfile only COPYed `crates ./crates`, but the workspace `Cargo.toml` lists `experiments/v0.4-*` as members. Cargo refuses to parse the workspace manifest without their per-crate manifests present, so `cargo build --release --bin forkd-controller` died with 'No such file or directory' on `experiments/v0.4-uffd-wp-poc/Cargo.toml'. This broke `release.yml`'s `build + push Docker image` job from v0.5.0 through v0.5.2 — the tarball, PyPI, and npm publishes all succeeded, but the ghcr.io controller image was never pushed. Fix: COPY the experiments dir into the builder. They're not selected by `--bin forkd-controller` so they don't compile, but cargo can now parse the manifest. Adds ~100 KB of source to the builder stage; the final runtime image is unchanged. Verified on dev box: image builds clean (152 MB), forkd-controller 0.5.2 runs from the image. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>
2 tasks
WaylandYang
added a commit
that referenced
this pull request
Jun 11, 2026
…ed (#238) publish-pypi.yml was failing with HTTP 400 'File already exists' when a v* tag got force-pushed (e.g. when fixing release.yml's Docker job for v0.5.2 — see #235). The publish step has nothing to do, but the action defaults to skip-existing=false, so the no-op surfaces as a failure. Set skip-existing=true: PyPI immutability means re-publish is a logical no-op anyway, and treating it as a hard error means every tag-fix dance leaves a red X behind that's actually a non-event. Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
WaylandYang
added a commit
that referenced
this pull request
Jun 11, 2026
* fix(ci): pypi-publish skip-existing=true so re-pushed tags don't go red publish-pypi.yml was failing with HTTP 400 'File already exists' when a v* tag got force-pushed (e.g. when fixing release.yml's Docker job for v0.5.2 — see #235). The publish step has nothing to do, but the action defaults to skip-existing=false, so the no-op surfaces as a failure. Set skip-existing=true: PyPI immutability means re-publish is a logical no-op anyway, and treating it as a hard error means every tag-fix dance leaves a red X behind that's actually a non-event. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]> * feat(cli): forkd quickstart — zero-to-first-fork in one command Conversion-funnel work: stars are growing ~10x faster than downloads, and the most-cited reason is first-run friction (host setup, kernel download, snapshot bake, netns provisioning — four manual steps across two scripts and three verbs before the first fork). `sudo -E forkd quickstart` collapses that to one command: 1. doctor::preflight() — hard-gates on platform/KVM/firecracker; reports kernel/tap/docker as data rather than failures 2. Heals what's missing WITH CONSENT (or --yes): guest kernel download, tap creation, netns provisioning. The setup scripts are embedded via include_str! so a tarball install works without the repo checked out; an interactive prompt lists exactly what will be touched before anything runs. Nothing on the host changes without an explicit yes. 3. Snapshot: reuse existing > bake from python:3.12-slim via local Docker (preferred — uses *this host's* Firecracker, so the vmstate can never hit cross-FC-version restore errors) > hub pull fallback with a compat warning. 4. fork -n 10 --per-child-netns, then prints next steps. Idempotent: re-running skips completed setup and reuses the snapshot. README: new "One command" section at the top of Quick start; stale v0.3.4 tarball URLs bumped to v0.5.2. Co-Authored-By: Claude Fable 5 <[email protected]> --------- Co-authored-by: Claude Opus 4.7 (1M context) <[email protected]>
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
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.
Summary
The Dockerfile only copied
crates ./cratesinto the builder stage, but the workspaceCargo.tomllistsexperiments/v0.4-*as members. Cargo refuses to parse the workspace manifest without their per-crate manifests present, socargo build --release --bin forkd-controllerdied withNo such file or directoryonexperiments/v0.4-uffd-wp-poc/Cargo.toml.This broke
release.yml'sbuild + push Docker imagejob from v0.5.0 through v0.5.2 — the tarball, PyPI, and npm publishes all succeeded each release, but the ghcr.io controller image was never pushed.Fix
Add
COPY experiments ./experimentsto the builder stage. Experiments aren't selected by--bin forkd-controllerso they don't compile; cargo just needs them present to parse the workspace.Verification (dev box)
Test plan
🤖 Generated with Claude Code