ci: restore agent-device iOS 26 pointer validation - #101
Draft
thiagobrez wants to merge 15 commits into
Draft
Conversation
…ort (#84) Adds a paste-ready upstream issue draft, an evidence staging script, and a handoff README for the human filer. The second report listed on #84 (the dyld/libcurl runner-artifact failure) is dropped as unconfirmed: it occurred once and the evidence did not support the cache-key theory from #56. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…-device report (#84) A dispatchable workflow erases and cold-boots the iOS 26.5 simulator on hosted macos-26 runners, drives the same gesture drag as the device-contract job, and records whether the app observed it, plus probes that separate a warm-up loss from an app-instance or a synthesis-path failure. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…ld-touch repro (#84) Reworks the reproduction to classify each synthesized gesture as lost, late, prompt, or errored by what the app observed, retry transient scenario setup so fewer iterations are wasted, and emit the delivery-latency distribution plus the XCTest-coordinate-tap discriminator. Documents run 34025552105: the first gesture was lost or late in 3 of 7 measured iterations (p90 latency a full 15 s timeout) while the XCTest tap landed sub-second every time. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Run 34027523634 measured all 12 iterations (0 wasted after the setup retry) and reproduced the lost/late synthesized gesture independently, p90 delivery latency at the full wait timeout, XCTest tap landing in 11 of 12. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
1 similar comment
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
Owner
Author
|
No changeset needed |
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
1 participant
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.
Restore Agent Device for all five iOS 26 pointer scenarios, preserving semantic destinations and public-outcome assertions. Detox retains Android, accessibility/lifecycle and native iOS multi-selection contracts.
Pin published
agent-device@0.21.0, which includes the merged alert-routing fix. Setup opens the scenario link once, accepts its confirmation when present, and verifies the initial outcome. TypedALERT_NOT_FOUNDis allowed; runner errors fail setup. Manual dispatch can build an immutable upstream SHA; local runs acceptAGENT_DEVICE_BIN.Fix the Scenario Lab’s stale initial outcome when resetting a pristine scenario or reopening the same link. A reset now creates a fresh scenario instance, so the parent receives its initial order again. No waits or retries are added.
The exact candidate remains the sole main device gate. Existing whole-job retry limits, runner-cache purge, render backstop, four-engine parity and the latest Pods cache fix are preserved.
Validation:
6d1b8a5: iOS 26 on attempt 1, both iOS 27 lanes, Android, all consumers/minimum-runtime checks, 16 shared outcomes across four configurations, and the exact-candidate gate.a6951c298f016ebc440f715a6c8647ace66d0b8a: two reset regressions fail before the fix; all 17 example tests, lint and typecheck pass afterward. The rebuilt Release app passes live repeated links, pristine Reset, a real drag and post-drag Reset on iOS 26.5. Earlier 16 contract/preflight tests passed.The original cold-touch PR2362 remains draft: empty-event preparation has not demonstrated an improvement. The successful matrix did not use that proposal. This PR remains draft while cold-start reliability and upstream review are followed up; earlier evidence does not establish an OS-only cause.