chore(dev): kill Ruby and port release scripts to native vdev#25379
Open
chore(dev): kill Ruby and port release scripts to native vdev#25379
Conversation
github-actions
Bot
added
domain: ci
pront
added
the
no-changelog
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 26839667cb
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
thomasqueirozb
approved these changes
May 6, 2026
Contributor
thomasqueirozb
left a comment
thomasqueirozb
left a comment
There was a problem hiding this comment.
Glad to see all of this deleted 👍
pront
added a commit
that referenced
this pull request
May 6, 2026
Codex review on PR #25379 caught two regressions in the previous commits: P1: scripts/environment/Dockerfile still COPYed scripts/Gemfile and scripts/Gemfile.lock — both deleted in this branch. The dev container build would fail at the COPY step. Drop those lines. P2: scripts/check-events is still a Ruby script and is wrapped by `vdev check events` (vdev/src/commands/check/mod.rs:35), so `make check-events` requires a Ruby interpreter. The previous cleanup of bootstrap-macos.sh and bootstrap-ubuntu-24.04.sh was too aggressive: it stripped Ruby itself, not just the bundler / Gemfile-related bits. Restore `ruby` (Homebrew on macOS, apt on Ubuntu) and re-list it in vdev/README.md as a prerequisite, scoped specifically to `vdev check events`. scripts/check-events only uses the Ruby standard library (`require 'find'`), so the Gemfile and bundler stay deleted. The remaining Ruby footprint is now exactly: one runtime dependency for one scripted check, queued for a follow-up port. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 4a04f070ee
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
pront
added a commit
that referenced
this pull request
May 6, 2026
- rustfmt cleanup over the previous commit (pre-push hook).
- generate_cue: change `git log v{last}..` to `v{last}...`. Codex review
on PR #25379 noted that `--cherry-pick --right-only` is a no-op with
the asymmetric two-dot range — the Ruby script the port replaces used
three dots (`v#{last_version}...`) so it could omit commits already
released on a patch branch. With two dots, fixes from a back-released
patch could re-appear in the next minor release CUE.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2401331aff
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Ports the two remaining Ruby release scripts to native Rust and folds the resulting `vdev release` surface down to the bits that are actually called. - `scripts/generate-release-cue.rb` (467 lines) is replaced by a `pub(super) fn run` in `vdev/src/commands/release/generate_cue.rs`. The function walks `git log v<last>...HEAD` (three-dot symmetric difference, so `--cherry-pick --right-only` filters out commits already released on a patch branch — the Ruby form), parses the conventional-commit messages, processes `changelog.d/*.md` fragments into the release CUE schema, and runs `cue fmt`. It is called only from `vdev release prepare` (no standalone CLI subcommand). - `scripts/release-commit.rb`, `vdev release commit`, `vdev release push`, `vdev build release-cue`, and the corresponding Makefile targets (`make release`, `release-commit`, `release-push`) are removed: none of them were called by CI workflows, scripts, or issue templates, and the original `release-commit.rb` had been broken for some time (pointing at `docs/reference/releases`, a directory that no longer exists). Manual operator tagging happens via `git tag` directly per the release issue templates. - The `scripts/util/*.rb` library, `Gemfile`, `Gemfile.lock`, and the `Gemfile` COPY in `scripts/environment/Dockerfile` are deleted. `ruby` itself stays in the bootstrap scripts and `vdev/README.md` — `scripts/check-events` is the last remaining Ruby script and gets ported in a follow-up commit. - Docs, issue templates, the `changes.yml` workflow filter, and `website/README.md` are repointed at `cargo vdev release prepare`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the last Ruby script in the repo (515 lines) with a native
`vdev check events` that walks Rust source via `syn`'s AST instead of
regex-scraping. The validation rules from `docs/specs/instrumentation.md`
are preserved unchanged; behaviour against the current tree matches
the Ruby script (`0 error(s)` on master, identical reports on a
synthetic fixture exercising every rule branch).
Implementation notes:
- Structural scanning (struct fields, impl-block trait/event names)
uses syn visitors. Macro argument parsing (`counter!(...)`,
`trace!(...)`, etc.) uses small targeted regexes on the macro's
already-tokenised input. Use-counting (`emit!(EventName)` /
`register!(...)`) uses one regex pass over the raw source because
syn does not descend into the bodies of opaque macros like
`tokio::select!` where some emit sites live.
- `## skip check-validity-events ##` is added to
`lib/vector-common/src/internal_event/component_events_dropped.rs`.
That impl block is the foundation type the `registered_event!` macro
abstracts over, so it has to implement `RegisterInternalEvent` by
hand. The Ruby script's regex never matched its multi-line `impl`
header and so silently let it through; the AST scanner sees it
correctly and the existing skip-marker mechanism makes the intent
explicit.
- `proc-macro2`, `quote`, and `syn` are declared as workspace deps with
default-features off; vdev opts into the syn `full`, `visit`,
`extra-traits`, `printing`, `parsing` features and proc-macro2's
`span-locations` locally. These features are already pulled in
transitively by Vector's existing dep graph (verified via
`cargo tree -p vector`), so the additions do not change Vector's
production binary.
- 32 unit tests are added: 14 covering the macro/log/metric arg
parsers (literal-first, named-field, variable, trailing-literal,
`%`-capture, message format checks, comma-arg nesting,
`CamelCase → snake_case`) and 18 driving every branch of
`validate_event` with synthetic `Event` structs.
- Differential test (run by hand against a worktree with a synthetic
fixture file): `ruby scripts/check-events` and `cargo vdev check
events` flagged the same eight events plus the same duplicate pair;
outputs differed only in sort order and a Ruby string-escape
artefact.
The bootstrap scripts (`scripts/environment/bootstrap-{macos,
ubuntu-24.04}.sh`), `vdev/README.md`, and `.ruby-version` lose their
Ruby references — there is no Ruby left in the repo to install.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
pront
force-pushed
the
pront/kill-ruby-scripts
branch
from
1f4b272 to
2f0b924
Compare
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2f0b924cae
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Two bugs caught in code review of `vdev/src/commands/check/events.rs`: 1. The native check globbed `src/**/*.rs` and `lib/**/*.rs` relative to the caller's CWD. The Ruby script wrapper used to chdir to the repo root, so this silently regressed: running `vdev check events` from anywhere outside the repo root produced `0 error(s)` without scanning anything. Fix: chdir to `paths::find_repo_root()` at the top of `Cli::exec`, matching the previous wrapper's behaviour. Running from a subdir now scans correctly; running from outside the repo errors out with a useful message. 2. `parse_log_args` claimed the first bare positional expression as the log message before scanning the rest of the args for a string literal. For real Vector code that uses `warn!(%error, "Failed to flush.")`, the literal is the actual message but the function never reached it — so the message-format check (capital + trailing period) silently never fired on those sites. Fix: walk all args once, classify each as a literal candidate, `message =` named-field candidate, bare-positional candidate, or plain parameter, then pick the message in that priority order. Adds a regression test covering the bare-then-literal pattern. Both bugs were flagged by Codex code review on PR #25379. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Three review findings on the check-events / release-cue ports: - `vdev check events` only ran its log-message format check via the syn AST visitor, which does not descend into the bodies of opaque outer macros like `tokio::select!` / `cfg_if!`. Log calls nested inside those macros were silently exempt. Replaces the AST format check with a raw-text pass that finds every `(trace|debug|info| warn|error)!(...)` invocation regardless of nesting. Adds two unit tests against the new pass — including one that asserts a violation inside a synthetic `tokio::select!` block is still reported. - `generate_cue::run` no longer relies on `--cherry-pick --right-only` alone to dedupe commits against earlier releases — that flag only filters patch-id-equivalent commits, so a conflict-resolved cherry- pick (different SHA, same PR) could re-appear in the next release CUE. Adds a `collect_released_identifiers` pass that pulls the SHA and PR-number sets out of every existing `website/cue/reference/releases/*.cue`, then filters the fetched commit set against both. Two unit tests cover the extractor. - `Commit::validate` now bails when the conventional-commit parser could not classify the subject (`type: None`). Previously such commits passed validation and landed in the published release CUE with a null type — a malformed PR title would silently ship. Also factors `parse_log_args` so it shares one `&str` implementation with the new raw-text pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pre-push hook reformatting on the format-check + dedup fixes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`cargo clippy -p vdev --all-targets` flagged two map_unwrap_or lints in the test-only `check()` helper. Identical pattern to fixes already applied in production code; just missed when test cases were added later. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Drop the pedantic clippy allows the file was carrying and refactor to satisfy them: rename Event fields whose `_event` suffix tripped struct_field_names, group the three skip flags into a SkipFlags substruct (struct_excessive_bools), and split validate_event and Cli::exec into smaller helpers (too_many_lines). Also remove a redundant `name_ends_with` wrapper that just called str::ends_with, and merge the parse_log_args TokenStream wrapper into its `&str` body. One callsite was already operating on stringified tokens, so it was doing a wasted parse-and-restringify round trip. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 2c841c3a94
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Replace the bespoke `camel_to_snake` regex helper in vdev's check-events with `convert_case::Casing::to_case`. Promote convert_case to a workspace dep so vdev and vector-config-common share the same version declaration. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The comment claimed declaring these at the workspace level prevented vdev's extra `syn` features from unifying into Vector's production build. That's not what workspace declarations do — `cargo` unifies features across the resolved dep graph regardless. The reason vdev's features don't reach the `vector` package is simply that vdev isn't a dep of `vector`. The comment was also stale: these crates are now direct deps of vector-config-macros, vector-config-common, and vector-common-macros, not just vdev. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The release-CUE generator rejected commit subjects with uppercase scope characters (e.g. `fix(ARC): ...`), even though `.github/workflows/semantic.yml` explicitly lists `ARC` as a valid scope. A PR title that passed semantic validation could therefore block `release prepare` later. Widen the scope character class to allow uppercase letters and add a regression test. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bec41f4905
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
split_comma_args only tracked `()[]{}` nesting, so a registered_event!
handle field with a generic type containing a comma would be split
mid-type. The existing src/internal_events/filter.rs field
`Registered<ComponentEventsDropped<'static, INTENTIONAL>> = register!(...)`
broke before the assignment, hiding the register! emission and causing
spurious missing-log/counter errors for FilterEventsDropped.
Track angle depth as a separate counter and only decrement `>` when
there is a matching `<`, so the `>` in `key => value` tag pairs (used
inside counter! args) stays unaffected.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
github-merge-queue
Bot
removed this pull request from the merge queue due to failed status checks
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
2 participants
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
Removes Ruby from the Vector release flow and trims the dead release-tooling that the Ruby scripts had been propping up. The release CUE generator is preserved as a native Rust library function called from
vdev release prepare; the standalone CLI surface around it that nobody actually used is deleted.One Ruby script remains:
scripts/check-events(a 515-line static analyzer used byvdev check events). Porting it is scoped as a follow-up so this PR stays focused on the release-flow cleanup. Bundler and theGemfileare gone for good —check-eventsonly uses the Ruby standard library — but the Ruby interpreter itself stays in the bootstrap scripts untilcheck-eventsis ported.Motivation
release-commit.rbpointed atdocs/reference/releases, a directory that has not existed in this repo for some time. The script could not have run successfully against the current tree, but nobody noticed — a strong signal it had been dead long enough to delete.cargo vdevalready exists as Vector's Rust-native dev CLI. Maintaining a parallel Ruby toolchain alongside it is duplicate cost with no upside. This PR continues the migration started in Rewritescripts/generate-component-docs.rbin Rust #22350 / feat(dev): rewrite scripts/generate-component-docs.rb in Rust (#22350) #24781 (which portedgenerate-component-docs.rb).vdev releasesubcommands and Makefile targets were revealed as orphaned. Removing them shrinks the operator-facing surface to just the commands that real callers (CI workflows, the issue templates) actually use.Vector configuration
N/A. Internal release tooling only.
How did you test this PR?
vdevcover bump-type computation, conventional-commit parsing, changelog-fragment parsing, commit validation, and CUE rendering.git worktreeof master withcargo vdev release prepare --version 0.56.0 --vrl-version 0.32.0 --dry-run. Produced 6 commits on the prep branch, generated0.56.0.cue, retired allchangelog.d/*.mdfragments viagit rm, andcue fmtaccepted the output. No pushes, no PR, no impact on the primary checkout.cargo clippy -p vdev --no-depsis clean (pedantic + warnings denied).Change Type
Is this a breaking change?
Does this PR include user facing changes?
no-changeloglabel to this PR.References
Follow-up to #22350 / #24781, which removed
generate-component-docs.rb.