Repository navigation
What shipped this week, who built it, and what's open (0.3.47–0.3.54) #157
adityankale190895
announced in
Announcements
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Starting this week, there's a regular post on what shipped, who built it, and where help would land best next. This is the first one. It covers CLI 0.3.47 to 0.3.54 and the Core work underneath them.
Who built what
Rust support — FredApps, in four stacked PRs:
cargo testcommands for the selected crates (feat(rust): synthesize conservative cargo test commands #93)On a synthetic workspace, the generated command runs 6 of the 9 Rust tests, and it caught all three seeded regressions. Being straight about the limits: glob is a single crate, so every revision there still ran the full suite. No public workspace has narrowed yet. #131 tracks that.
Config and lockfile safety, docs and cleanup — sarankumar1325, nineteen PRs:
jest.config,vitest.configor Babel config changes (fix(git-diff): fall back when a package's jest, vitest or Babel config changes #121, closing Config files in subdirectories (babel.config, jest.config) may get no fallback #117), and when onlybabel.configchanges at the root (fix(git-diff): run every test when only babel.config changes #108)vite.config-only changes (test(impact): fixture for a vite.config-only change in a Vitest project #109), and a test for a root Vitest config withisolate: false(test: a root Vitest config without isolation gets no selected command #149, closing Test: a root Vitest config with isolate: false gets no selected command #126)src/(refactor(file-references): use the shared lockfile list #118–refactor(git-diff): use the shared lockfile names in the lockfile-free narrowing #120, refactor(ci-inference): use the shared lockfile names in lockfileFacts #146–fix(research): use the shared LOCKFILE in the generic path baseline #148, refactor(analyzer): use the shared lockfile names for config-file discovery #151), every shared lockfile classified as a config change (fix(impact): classify every shared lockfile as a config change #152),bun.lockin CI inference (fix(ci-inference): report bun.lock in lockfileFacts #105), and every lockfile hashed in the economics context (fix(cache): hash every lockfile in the economics context #107)replacetargets (feat(go): model in-repository replace targets under root-module scope #110)\0instead of raw NUL bytes in two source files (fix: use the "\0" escape instead of raw NUL bytes in source files #145, closing Replace the literal NUL bytes in git-diff.ts and failure-baseline.ts with the \0 escape #125)Named fallback reasons — tyza66, first-time contributor: a change to
.nvmrc,.node-version,.npmrc,.mocharcand similar files now says which tool configuration triggered the full run instead of "Unknown changed file" (#144, closing #142), and so does atsconfig.base.jsonortsconfig.*.jsonchange (#155, closing #141).CLI help — aniKet0753: the help text, usage and tests now list the same commands (#130, closing #129).
0.3.49 closes two misses in the same family as #108 and #121. Before it, a change to a test setup file (
setupFiles,setupFilesAfterEnv,globalSetupand similar) or tovitest.workspace.tsselected no tests at all. Now both run the full suite and name the file (#139, #140). That work built directly on the config fixtures above.Also new since 0.3.49: an Azure Pipelines task that observes changes (0.3.53), and
verify --staged/--pathsto verify only what will be committed (0.3.54).Where help would land best next
Every good first issue from last week was picked up within a day, which is the best problem to have. One is open right now:
jest --changedSince,vitest related,nx affectedandturboMore are coming. Larger pieces, if you want to own an area:
turbo run(tracking issue) #115, Bazel adapter: hand selected targets tobazel test(tracking issue) #116)Contributors who keep coming back get first pick of new issues and a say in design. See the tiers in CONTRIBUTING.md.
Try it on your own repository
If you've contributed, you already know what DiffCI is meant to catch. The most useful thing you can do next takes five minutes: run it on a repository you work on, and tell us where it's wrong.
On a public repository, the GitHub App posts one neutral check per pull request. It never blocks a merge. Then file a community result: a wrong plan is worth more to us than a right one.
Thanks to everyone above. Next update with the next batch of releases.
All reactions