Skip to content

ci(ios): overlap device-free smoke setup as one parallel group - #4690

Merged
thymikee merged 3 commits into
mainfrom
ci/faster-ios-smoke-leg
Nov 18, 2026
Merged

thymikee merged 3 commits into
mainfrom
ci/faster-ios-smoke-leg

Conversation

@thymikee

@thymikee thymikee commented Nov 18, 2026 •

Copy link
Copy Markdown
Member

Summary

Three Smoke Tests (ios) steps ran in sequence although none needs a booted simulator: runner build (31.2 s median, 24.4-38.1), fixture restore (29.2 s, 22.7-38.7), simulator boot (5.2 s). They share no outputs and write disjoint paths, so they now run as one job-level parallel: group and the leg pays the slowest member. android.yml already does this.

A composite cannot host a parallel: group or read the caller's step context, so the group lives in apple.yml and its results enter smoke-ios-run as inputs. The host focus canary stays ahead of the group, since the boot can raise the Simulator app, and the phase requires its outcome. Readiness stays mandatory inside the phase.

Before (6 main runs) After (4 runs)
Leg window 208-238 s, median 224 186-216 s, median 194
boot + build + restore 65.6 s in sequence 31-39 s as one group

Window: Build the shared CLI end to iOS smoke tests end, both sides. Each run against its own serial equivalent cancels host speed: the overlap saves 35-46 s, all cache hits. Runs: 44303358376, 44306734275.

Validation

Head 8a2c399d8f, Apple green on every attempt. A hoisted step that fails surfaces only as a skipped phase, so the phase requires each one's outcome, the diagnostic trigger names them, and the summary prints all four. tsc -b, apple-ci-impact.test.ts (53) and trusted-fixture-artifact.test.mjs (16) pass.

🤖 Generated with Claude Code

The iOS smoke leg ran the simulator boot, the XCTest runner build, and the
fixture app restore one after another, although none of the three needs a
booted simulator. Move them to one job-level parallel group so the leg pays
the slowest member instead of all three.

A composite action cannot host a parallel group, and it cannot read the
caller's step context, so the group lives in apple.yml and the boot, build,
and restore results cross into smoke-ios-run as inputs. The host focus
canary stays ahead of the group because the boot can raise the Simulator
app, and readiness stays mandatory inside the phase so no step reaches a
simulator that is still starting.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@github-actions

github-actions Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

Size Report

Metric Base Current Diff
Installed (including dependencies) 6.40 MB 6.39 MB -6.7 kB
Package (unpacked) 6.40 MB 6.39 MB -6.7 kB
Package (download) 1.94 MB 1.94 MB -2.2 kB

Startup median (7 runs, lower is better):

Scenario Base Current Diff
CLI --version 25.3 ms 26.7 ms +1.4 ms
CLI --help 53.7 ms 63.2 ms +9.5 ms

@apex-sentinel-staging

This comment has been minimized.

@apex-sentinel-staging apex-sentinel-staging Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Apex Sentinel: 1 new finding on 6bcdc01. The verdict and the full list are in the summary comment.

Comment thread .github/workflows/apple.yml Resolved
thymikee and others added 2 commits November 18, 2026 13:37
The canary used to run inside the iOS phase, so its failure failed the phase.
It now runs ahead of the device-free parallel group, and `!cancelled()` does
not see a prior step's failure, so the phase would have run roughly 200 s of
XCTest, replays, and E2E against a host that cannot hand focus to Finder, and
reported a replay failure instead of the cause.

Require the canary's outcome in the phase condition, and name it in the host
diagnostic trigger and the phase summary, where a hoisted step that failed
otherwise shows up only as a skipped phase.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The host-diagnostic guard looks the step up by name, so declare the field the
workflow parser already returns.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@apex-sentinel-staging

Copy link
Copy Markdown

Apex Sentinel · ✅ Approve · review of 8a2c399

Since the review of 6bcdc01: 0 new · 0 still open · 1 no longer raised (1 marked fixed)

No open findings.

Reviewed with this repository's own Sentinel configuration (15c46788028a).

💬 Talking to Sentinel

In a finding's thread, reply with:

Command What it does
@apex-sentinel-staging note <context> Adds context; Sentinel re-checks the cited code
@apex-sentinel-staging pivot <why> Says the finding is wrong; Sentinel re-checks the cited code
@apex-sentinel-staging like Marks the finding as useful
@apex-sentinel-staging dislike <why> Marks it as not useful, so it is not raised again

On this conversation, comment @apex-sentinel-staging review to run a full review once again.

@thymikee

Copy link
Copy Markdown
Member Author

This PR is ready at 8a2c399. The iOS smoke job passed at this head, and that run went through the new parallel group, so the changed route was exercised. There are no conflicts.

Not blocking, and you can take or leave these: (1) the comment above the group in apple.yml and the PR body say boot, native build and fixture restore ran one after another, but since #3836 the boot already overlapped the build, so the new overlap is fixture restore with the runner build and the saving is about 25-30 s, not more; (2) the saving rests on one PR run (244 s against 234-252 s on four main runs), so a median of at least three head runs against the main median would show it; (3) smoke-ios-run could drop the fixture-source input by moving the cache-source report after the group, and could replace the fixture-producer step with one wait-for-artifact-seconds expression; (4) the group test in apple-ci-impact.test.ts could read steps through runnableSteps like device-lane-impact.test.ts does; (5) the hoisted setup steps are listed by id in several places and no longer have the phase 80-minute timeout, so is the 90-minute job timeout enough for you there?

No red run of the group exists yet, so sibling cancellation and outcome values are shown only by the workflow semantics. The canary thread (#4690 (comment)) is fixed at this head, so please resolve it.

@thymikee thymikee added the ready-for-human Valid work that needs human implementation, judgment, or maintainer merge label Nov 18, 2026
@thymikee
thymikee merged commit 7610947 into main Nov 18, 2026
20 checks passed
@thymikee
thymikee deleted the ci/faster-ios-smoke-leg branch November 18, 2026 13:08
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready-for-human Valid work that needs human implementation, judgment, or maintainer merge

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant