Use Ruby 3.4.9 and Xcode 26.6 - #825
Merged
Merged
Conversation
Moves off Xcode 16.2 / iOS 18.2, which no longer has a CI VM image or a bundled iPhone SE (3rd generation) simulator. `IMAGE_ID` is now derived from `.xcode-version`, matching wpios and wcios, so future bumps touch a single file. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
Brings the repo in line with the rest of the mobile repos (AINFRA-2723). `BUNDLE_FORCE_RUBY_PLATFORM` and the `ruby`-only `PLATFORMS` section keep the lockfile from pinning gems to a specific Darwin version, so it survives the CI image's macOS bumps. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
The 4.0.0, 5.0.0 and 6.0.0 breaking changes don't reach this pipeline: it doesn't call `pr_changed_files`, and the rest are Windows-agent scripts. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
SwiftUI rendering changed between iOS 18 and iOS 26, and the reference device changed, so every image had to be recaptured. Recorded on macOS 26.5.2 with Xcode 26.6 (17F113) and the iOS 26.5 simulator, which is what the `xcode-26.6` CI VM runs — these tests are coupled to the host macOS, not just the Xcode and simulator versions, so references recorded on a mismatched host fail on CI no matter how many times they are re-recorded. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
Contributor
There was a problem hiding this comment.
Pull request overview
Updates the repository’s CI/tooling configuration to align with newer Ruby and Xcode/iOS versions, including corresponding CI image/toolkit pins and lockfile re-resolution.
Changes:
- Bump Ruby to
3.4.9and update Bundler/lockfile settings (including forcing the Ruby platform in Bundler). - Move CI/Xcode targeting to
26.6(and iOS26.5simulator usage in Fastlane), including deriving the Buildkite image from.xcode-version. - Update the DocC GitHub Actions workflow to run on the newer macOS runner and select the newer Xcode app bundle.
Reviewed changes
Copilot reviewed 6 out of 141 changed files in this pull request and generated 1 comment.
Show a summary per file
| File | Description |
|---|---|
| Gemfile.lock | Re-resolves lockfile for Ruby-platform gems and updates Bundler version metadata. |
| fastlane/Fastfile | Updates the simulator OS/device target used by Fastlane test lanes. |
| .xcode-version | Adds Xcode version pin used by CI configuration. |
| .ruby-version | Bumps the Ruby version used by dev/CI tooling. |
| .github/workflows/docc.yml | Moves DocC publishing workflow to the newer macOS runner and Xcode selection. |
| .bundle/config | Forces Bundler to resolve Ruby-platform gems (avoid OS-specific lock entries). |
| .buildkite/shared-pipeline-vars | Derives Buildkite image ID from .xcode-version and bumps CI toolkit pin. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+6
to
+9
| # The ~> modifier is not currently used, but we check for it just in case | ||
| XCODE_VERSION=$(sed -E 's/^~> ?//' .xcode-version) | ||
|
|
||
| export CI_TOOLKIT="automattic/a8c-ci-toolkit#3.9.1" | ||
| export IMAGE_ID="xcode-$XCODE_VERSION" |
Just to be on the very latest tooling
Temporary. The tests run with `record: .failed`, so a failing CI run rewrites the mismatching references in place — but on the previous VM image those writes were silently dropped, and the uploaded artifacts came back identical to the committed files. Re-testing that on the `xcode-26.6` image: if the artifacts differ from what is committed, the CI-recorded images are usable as references. Revert once the question is answered. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
This reverts commit 492cefb.
Views that rasterize an image don't render identically across macOS builds even with Xcode, the simulator runtime, and the device all matched, and the CI VM images move through point releases on their own schedule. Confirmed on this branch: with everything else matched, 30 assertions still mismatched on CI, all of them image-bearing. Recording on CI is not an escape either — the artifacts a failing run uploads are bit-identical to the committed references, so the simulator sandbox is dropping the writes. Text-only suites match exactly and stay on `.image`. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
Temporary. `record: .failed` replaces the diffing message with a generic "does not match reference", so the logs can't say whether the images differ in size — which no precision setting can fix — or by how much they differ perceptually. `.never` lets the real message through. Revert once it is read. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
Temporary. The diffing message is lines 2-3 of the failure and xcbeautify only prints line 1, so the real reason never reaches the log. The bundle carries it, along with the reference/failure/difference images the library attaches on failure — a different write path from the `__Snapshots__` files the simulator sandbox drops. Revert once the bundle is read. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
This reverts commit c01a170.
The earlier attempt set `perceptualPrecision` alone, which cannot pass: `precision` stays at 1.0 by default and demands every pixel match, so the failure count did not move by a single test. Reading the diffs out of the CI xcresult showed what actually differs — the `job • pronouns • location` line, laid out with tighter separator spacing on the CI host. That, not the presence of an image, is what every failing suite has in common. Thresholds now come from the measured worst case per view rather than a guess, and stay tight enough that losing the line would still fail. --- Generated with the help of Claude Code, https://claude.ai/code Co-Authored-By: Claude Code Opus 4.8 <noreply@anthropic.com>
|
| App Name | Gravatar Prototype Build | |
| Build Number | 2540 | |
| Version | PR #825 | |
| Bundle ID | com.automattic.gravatar-sdk-demo-uikit.prototype-build | |
| Commit | 844de0c | |
| Installation URL | 7tf6gequ095b8 |
mokagio
commented
Jul 21, 2026
Co-authored-by: Gio Lodi <giovanni.lodi42@gmail.com>
AliSoftware
approved these changes
Jul 21, 2026
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
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.
While working on #824 , I spun off to upgrade the tools here.
AI-generated details
Description
Moves the repo's CI tooling onto Ruby 3.4.9 and Xcode 26.6 / iOS 26.5, in line with wpios. Fixes AINFRA-2723.
The Ruby and Xcode bumps ride together because they share the CI config, and the lockfile re-resolve is only meaningful against the image that will run it. The
a8c-ci-toolkitpin goes from 3.9.1 to 6.1.1 in the same pass — none of the 4.x/5.x/6.x breaking changes reach this pipeline (nopr_changed_filescall; the rest are Windows-agent scripts).Supersedes #803, which stalled on snapshot failures.
Gotchas
The
job • pronouns • locationline lays out with slightly tighter spacing around its separators on the CI host than on a developer Mac — same string, same font, shifted a pixel or two. Every view that renders that line fails a pixel-exact comparison on CI; views without it match exactly. HenceSnapshotting+HostTolerance.swift, whose thresholds come from the diffs measured in a CI run, not from taste. The file documents the reasoning; the short version is that the allowance stays below the share of the frame that line occupies, so a regression removing it still fails.Two things that look like escape hatches are not. References cannot be recorded on CI: the run logs "a new snapshot was automatically recorded", but the artifacts come back bit-for-bit identical to what is committed — the simulator sandbox drops the writes. And re-recording locally does not converge, because the coupling is to the host macOS build, which the CI VM images move through on their own schedule.
BUNDLE_FORCE_RUBY_PLATFORMand theruby-onlyPLATFORMSsection are new, matching wpios — they stop the lockfile pinning gems to a Darwin version the CI image will eventually move off.Testing Steps
bundle exec fastlane testlocally on Xcode 26.6 (355 tests, 0 failures).Deploy Gravatar DocCworkflow still publishes on merge totrunk— it moved to themacos-26runner.Notes for reviewers
The two
Probe:commits and their reverts are the diagnostic runs that produced the numbers above. Happy to squash them out if you would rather the history not carry them, but they are why the thresholds are what they are.One of them is worth knowing about independently: wrapping
validate_swift_packagein|| EXIT_CODE=$?suppressed the toolkit function's internalset -e, so a run with 30 failing tests exited 0 and reported a green check. Reverted in667224d0.