Skip to content

Fix measurement lifecycle reviews and preserve sampler batching - #2

Merged
prematurely merged 3 commits into
mainfrom
codex/review-fixes-20261110
Nov 11, 2026
Merged

prematurely merged 3 commits into
mainfrom
codex/review-fixes-20261110

Conversation

@prematurely

@prematurely prematurely commented Nov 9, 2026 •

Copy link
Copy Markdown
Owner

Fix three post-merge measurement reviews from #1 while preserving the existing generic sampler optimization. A multithreaded Reset could deadlock with a Present callback reading diagnostics; a secondary device could claim the selected capture's thread; ANSI module enumeration could break symbol lookup for Unicode installation paths.

Changes:

  • Use device → observer lock order for lifecycle registration and notification, retaining synchronous unregister/final-release guarantees and same-device API reentry.
  • Reject non-selected devices before capture ownership, key handling or state progression.
  • Enumerate modules through Win32 W APIs and convert paths explicitly to UTF-8.
  • Retain sampler recording, first-write Journal capture, ordered batch mutation and existing pixel behavior.
  • Keep the measurement HUD clear of the frame-time graph in the tested full-HUD layouts. Move it to a right column when needed and cache ellipsized long lines.

Validation:

  • Each original review has an observed native RED → GREEN regression: old Reset/Present deadlock (watchdog exit 86) → 33 checks passed; secondary-device capture → 8 checks passed; actual Unicode DLL + matching RSYM + executable address under ACP 936 → 9 checks passed.
  • Full native measurement runtime: 82 checks passed.
  • Complete local pipeline on code commit f6876b2f: 85 passed / 0 failed / 2 skipped. Both skips require external source/config inputs; GPU and environment-sensitive gates ran. All 87 per-gate logs were integrity checked and hashed. The final documentation commit changes no product or test code.
  • Full HUD pixel checks passed at 2560×1600 and 1280×720 with eight real-format diagnostic lines, including a long function name. Arbitrary HUD scale, very narrow windows and unusually tall memory-details lists remain outside that matrix.
  • Final sampler A/B: 12 independent processes × 4 rounds; A/A: 6 × 4. The prior sampler behavior and cost benefits are retained. Present controls are reported with their noise: final no-observer cycles +8.61%, observer cycles −0.98%; these are not game FPS claims.
  • A fresh whole-branch review found no Critical, Important or Minor defects.

The earlier full run's two setup failures are retained: the historical ABI gate requires the original archived reference, while its default selected the currently installed game DLL. The complete suite passed with the explicit hash-verified reference, without weakening assertions. Local Clang 22 instability was handled by pinning compiler descendants only; runtime/benchmark affinity stayed normal.

Validation report · Raw paired evidence and gate hashes

Public interfaces, rendering quality and Vulkan staging/copy paths are retained. No game DLL, configuration or symbol file was replaced; all eight installed-file hashes match the pre-task snapshot. The game installation remains unchanged, and this PR will remain open for review.

Remote validation completed:

  • Windows CI run 42667461457, attempt 1, passed the clean Release x86 build, tests, packaging, package-layout verification and all uploads. The required Build, test, and package check is SUCCESS for head aadc8e0c48474f71c7137725b6f15fab44c6df2d.
  • Downloaded and hash-verified the exact results artifact: 39 passed / 0 failed / 48 configured skips on the hosted runner. These skips cover local GPU/environment/reference requirements and the already-current build; local native/GPU validation is recorded separately above.
  • CI tested temporary merge commit 33bd259532b4879cca8a8eb6af29d1150a67d2ea. Its tree 2bd1b3f5dc9f521aad2c509e4f69416bfd51a243 exactly matches the PR head tree.
  • Original review replies: lifecycle lock order, selected-device ownership, Unicode module paths. They link this follow-up PR; the original threads remain open pending integration.
  • The follow-up remains unmerged, with no game deployment.

Copilot AI balanced review requested due to automatic review settings November 9, 2026 17:43

Copilot AI 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.

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.

@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 9, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-09T17:46:07.377213Z aadc8e0 PR opened
🔒 Security Review ✅ Completed 2026-11-09T17:47:38.572418Z aadc8e0 PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@prematurely prematurely assigned Copilot and unassigned Copilot Nov 9, 2026
@prematurely
prematurely merged commit 0d5680c into main Nov 11, 2026
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants