ci: normalize Windows ARM64 desktop before full E2E - #13
Merged
cm2435-hcomp merged 14 commits intoJul 27, 2026
Merged
Conversation
cm2435-hcomp
marked this pull request as ready for review
July 27, 2026 15:59
cm2435-hcomp
merged commit Jul 27, 2026
8f2bdfe
into
kai/windows-arm64-runtime
11 of 12 checks passed
cm2435-hcomp
added a commit
that referenced
this pull request
Jul 28, 2026
* Add windows-arm64 to the runtime manifest with a placeholder digest and a friendly unavailable error Windows ARM64 users previously hit a bare RuntimeError; every unpublished platform (absent or placeholder) now raises RuntimeArtifactUnavailable with the PATH / download-env-override guidance. The placeholder entry makes the next runtime release require a windows-arm64 sha, so client and pipeline ship together. Co-authored-by: Cursor <cursoragent@cursor.com> * Add Windows ARM64 installer and release scaffolding * Fix Windows ARM64 dependency bootstrap * Ship Windows ARM64 dependency wheel * Build dependency wheel with native ARM Python * Run Holo evals against HAI candidates * Pin hai-agent-runtime 0.1.10 * Prepare Windows ARM64 E2E dependencies * ci: normalize Windows ARM64 desktop before full E2E (#13) * Spike Windows ARM64 desktop readiness * Route readiness spike through full E2E workflow * Close hosted Windows first-run surfaces * Require Windows OOBE broker to exit * Probe rendered Windows privacy experience * Remove Windows readiness spike routing * ci: complete Windows ARM privacy OOBE before e2e * ci: walk Windows privacy OOBE to completion * ci: click through hosted Windows privacy OOBE * ci: gate Windows privacy state on visible UI * ci: allow an idle Windows desktop without foreground * ci: prepare Windows desktop immediately before e2e * style: format Windows readiness workflow test * ci: isolate Windows ARM64 installer smoke --------- Co-authored-by: Cursor <cursoragent@cursor.com> Co-authored-by: Charlie Masters <charlie.masters@hcompany.ai>
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.
Summary
This is a stacked infrastructure PR on top of #8. It makes GitHub-hosted Windows 11 ARM64 runners reach a measured clean desktop immediately before Holo Full E2E model execution.
Root cause
The
windows-11-armimage starts in Windows privacy OOBE and can expose Edge/WSL first-run surfaces. The existing x86_64 control uses Windows Server and does not have this state, so the earlier ARM results were dominated by runner initialization rather than runtime behavior.Changes
Shell_OOBEProxyplus UI Automation semanticsNo CUA driver or HAI runtime production behavior changes.
Validation
Local:
git diff --checkpassedFull Windows ARM64 E2E: https://github.com/hcompai/holo-desktop-cli/actions/runs/30280259801
calculator_ci_smoke: reproducible evaluator failure, display remained0foreground_visible_editor_witness: agent failure after clean readiness; sentinel was not writtenThe infrastructure gate is green. This PR does not claim 8/8 model parity, and it preserves both remaining failures as real post-readiness task evidence.