starred/orca
1
0
Fork 0
mirror of https://github.com/stablyai/orca.git synced 2026-10-30 16:47:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

Revert iteration checks and custom Vitest selection (#34834)

Browse Source 
* revert: remove iteration checks and custom Vitest selection

* chore: use upstream affected-test selection

---------

Co-authored-by: OrcaWin <293788423+OrcaWin@users.noreply.github.com>
This commit is contained in:
OrcaWin
2026-10-28 00:11:04 -07:00
committed by GitHub
co-authored by OrcaWin
parent 1983946e25
commit 7cdfebdc1b
19 changed files with 204 additions and 2473 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.github/CONTRIBUTING.md
+3 -78
View File
@@ -93,51 +93,6 @@ pnpm dev:fresh [--keep] [--seed-from[=<profile dir>]] # temporary profile, rem
### Targeting Local Tests
During iteration, use `pnpm check:iteration --maxWorkers=2` to select checks for staged,
unstaged, and untracked edits against HEAD. Before handoff, run
`pnpm check:changed --maxWorkers=2` for the whole branch against its merge base with
`origin/main` (override with `--base=<ref>`). `check:changed --scope=worktree` is the
explicit form of `check:iteration`. A clean worktree has nothing to check in iteration
mode, even if the branch has committed changes. Both commands check the current
working-tree bytes; they do not validate a different version stored only in the index.
Both gates print each step's command, deadline, start, and completion, emit a heartbeat
every 15 seconds, and stream Vitest progress. Each step has a five-minute deadline;
pass `--stepTimeoutMs=<ms>` to increase it for a larger check. A timeout fails the gate
and requests termination of the step's process group/tree. Process shutdown can take
additional time. `--maxWorkers` also caps concurrent check tasks, while typecheck and
Vitest retain their own bounded worker pools. The deadline applies to each step,
not to the total command, and does not apply to direct `pnpm test` commands.
Passing results are cached separately for branch and worktree scope; a timeout, failure,
or cancellation never records a pass. Starting a real recheck invalidates the earlier
pass for that scope, including with `--force`. The cache includes runtime and effective
test-selection context. Inputs are rehashed before saving a pass; an edit detected
during checks fails the gate and requires a retry. Keep edits paused during checks:
this is not an immutable filesystem snapshot.
For gates with unit tests, reuse also requires matching dependency-record evidence.
New incomplete records arriving during checks prevent a pass; a successful run can
cache its own passing record updates immediately. Unreadable or oversized evidence
disables caching without discarding otherwise successful checks.
Stream the command's output to observe progress. If you redirect it to a log, tail that
log while the same command runs. A tool's output-yield interval can be short, but its
execution deadline must allow the selected steps and process shutdown to finish; poll
the existing process/session instead of launching a replacement on an observation timeout.
For example, start `pnpm check:iteration --maxWorkers=2` with a short tool-output yield,
then read that same running session until it exits and preserve its exit status.
Uncatchable termination of the gate, or descendants that escape its POSIX process group,
can require separate cleanup; these are not proof of a completed or passing check.
Closed output consumers cancel active checks. Slow consumers can omit live progress
with a notice; a final output drain that cannot finish fails the command.
Iteration checks use the CI unit inventory, leaving shell, Electron, and cross-version
tests to their dedicated CI jobs. Ordinary `pnpm test` and the branch handoff gate retain
their broader local discovery, and explicit test commands can still run those suites.
Both gates use the installed incremental dependency recording and its cold-cache static
fallback described below. Deleting a related source runs the full inventory for that
gate's scope, because a cold graph cannot reliably resolve a removed import. Neither
gate replaces CI's full coverage of global inputs.
Use a test file or directory when you know the relevant suites. For dependency-based
feedback, Vitest can select tests affected by your edits:
@@ -152,39 +107,9 @@ also includes committed changes since the branch's merge base. Update that ref w
you need a newer baseline. Both accept the same path filters and options as `pnpm test`.
When Git reports no changed files at all, the launcher exits before starting Vitest.
Local runs enable Vitest's [dependency recording](https://vitest.dev/config/experimental#experimental-recorddependencies).
Ordinary test runs populate the cache; later changed-file runs read it instead of
transforming every test's import graph. Records are separate for the Bun and Node
projects and live in each checkout's `node_modules/.vite`. A test without a record
runs when at most 100 such tests are in scope, which records them. Beyond that, as in
a new worktree, unrecorded tests are selected with Vitest's static import graph
instead of running all of them. Keep Vite's local test cache between runs. Use
`--experimental.recordDependencies=false` to use the static import graph for every test.
The cold fallback still scans test import graphs; it limits test execution, not all
transformation work.
Recorded source closures, configuration, and runtime context are checked for freshness;
stale records use the same cold-cache fallback. Failed or partial suites retain the
union of known dependencies and run again until they complete successfully. Sources
modified during a run cannot receive a fresh record for that run.
This is local feedback, not a complete regression check. Vitest records loaded
imports, including executed dynamic imports, but cannot infer arbitrary filesystem
reads, subprocess inputs, or externalized module dependencies. Explicitly run tests
for changed fixtures, generated files, CLI entry points, and other indirect inputs.
Configuration, lockfile, patch, and workspace-package changes passed directly to
Vitest's affected-test command rerun every test within the requested scope. The check
gates instead select colocated tests for these global inputs and rely on CI for full
coverage. CI does not enable recording;
the existing CI selection policy is unchanged. The static graph has the same blind spots
and also misses dynamic imports with computed specifiers, so the first changed-file run
in a new worktree is a little less complete than later ones.
`pnpm build:packages` (which `pnpm test` runs first) skips compiling workspace packages
when their sources, build script, and compiler are unchanged and `dist` is intact.
Vitest also supports [`related --run <source-files>`](https://vitest.dev/guide/cli#vitest-related)
and dependency-aware watch mode. `pnpm test --changed` uses Git to provide the source
inputs automatically and exits after one run.
Affected tests use Vitest's static import graph. Run suites explicitly for computed
dynamic imports, fixture files, subprocess inputs, or other dependencies outside that
graph. Local affected tests do not replace CI.
To reduce main-process rebuild CPU and memory when you do not need its source maps,
set `ORCA_DEV_MAIN_SOURCEMAPS=0` before `pnpm dev`. In Bash/zsh: