Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
33 changes: 23 additions & 10 deletions .github/workflows/ci.yml
Original file line number Diff line number Diff line change
Expand Up @@ -58,13 +58,14 @@ concurrency:
jobs:
# Three-layer CI (#3295) plan: decide the matrix scope BEFORE paying for
# legs. One cheap ubuntu job picks one of three scopes:
# core — every PR (push) and master push: 6 test legs (ubuntu 22 full 3-shard
# suite, ubuntu 24 single-leg whole suite, windows 22 2-shard) + 2
# core — every PR (push) and master push: 9 test legs (all three combos at
# /3 shards: ubuntu 22, ubuntu 24, windows 22 — #4351 rebalance) + 2
# static legs (ubuntu 24, windows 22 — #3406 item 8: the windows
# static leg keeps typecheck/build + the kernel/pi-subagents
# suites covered on windows for EVERY PR, not only ci:full/nightly).
# The blocking layer; branch protection requires exactly
# these 6 test leg names.
# The blocking layer. master's required-checks list still carries
# the legacy 12 FULL-matrix leg names; core produces 9 of them (the
# win node-24 legs appear only in FULL scope), enforce_admins off.
# full — `ci:full` label on a PR, a release PR (#3688: bot-authored
# one-click fallback PRs, *_release-v* head branches, or a
# "[PR] release…"/"release vN…" title — no label needed),
Expand Down Expand Up @@ -110,13 +111,24 @@ jobs:
FULL_INPUT: ${{ inputs.full }}
run: |
set -euo pipefail
# #4351: the suite grew ~399 -> ~860 files in three weeks while the
# core shape stayed frozen at its ~650-file design point (#3295), so
# its heaviest legs became the wall-clock floor: the ubuntu-24
# single-leg whole suite measured ~21 min and the windows 2-shard
# halves ~13-17 min (up to ~35 min tails on flake re-runs). All three
# combos now shard /3: partitionAll stays total for any i/N, so
# union of shards == every file — zero coverage loss; wall drops to
# the slowest shard (~7-12 min). The win node-24 legs stay FULL-only.
CORE_TEST='{"include":[
{"os":"ubuntu-latest","node-version":22,"shard":1,"shardTotal":3},
{"os":"ubuntu-latest","node-version":22,"shard":2,"shardTotal":3},
{"os":"ubuntu-latest","node-version":22,"shard":3,"shardTotal":3},
{"os":"ubuntu-latest","node-version":24,"shard":1,"shardTotal":1},
{"os":"windows-latest","node-version":22,"shard":1,"shardTotal":2},
{"os":"windows-latest","node-version":22,"shard":2,"shardTotal":2}]}'
{"os":"ubuntu-latest","node-version":24,"shard":1,"shardTotal":3},
{"os":"ubuntu-latest","node-version":24,"shard":2,"shardTotal":3},
{"os":"ubuntu-latest","node-version":24,"shard":3,"shardTotal":3},
{"os":"windows-latest","node-version":22,"shard":1,"shardTotal":3},
{"os":"windows-latest","node-version":22,"shard":2,"shardTotal":3},
{"os":"windows-latest","node-version":22,"shard":3,"shardTotal":3}]}'
CORE_STATIC='{"include":[{"os":"ubuntu-latest","node-version":24},{"os":"windows-latest","node-version":22}]}'
FULL_TEST='{"include":[
{"os":"ubuntu-latest","node-version":22,"shard":1,"shardTotal":3},
Expand Down Expand Up @@ -212,9 +224,10 @@ jobs:
# across more legs: each runs a balanced slice (scripts/run-ci-tests.mjs partitions
# deterministically + totally, so union of shards == allFiles, nothing skipped,
# same tests/assertions/env, c=4 within each shard). Wall drops ~3x (~20 -> ~7min)
# with zero coverage loss. To go faster, add entries to the FULL_TEST include
# list in the plan job above (each entry carries its own shardTotal, so the
# legacy "bump the /3" coupling is gone — any denominator works).
# with zero coverage loss. To go faster, add entries to the include list of
# the scope you want faster in the plan job above (each entry carries its own
# shardTotal, so the legacy "bump the /3" coupling is gone — any denominator
# works; #4351 used exactly this knob on CORE_TEST).
strategy:
fail-fast: false
matrix: ${{ fromJSON(needs.plan.outputs.testMatrix) }}
Expand Down
Loading