Repository navigation
fix(metal-replay): presentation replay failed on VideoOut label store (#494) - #500
Merged
Merged
Conversation
…he replay executables The replay build's guest memory check only accepts borrowed guest ranges, but VideoOut's buffer labels live in its host-side config. The flip completion store added in #494 therefore threw 'not borrowed' and the presentation replay failed. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DgW4n3qYgi5jAAohEmXRQj
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configuration
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
RINNECODER
added a commit
that referenced
this pull request
Oct 14, 2026
anyps5_metal_presentation_replay exits 77 (ctest Skipped) when the runner has no console session or the screen is locked, which hid a real regression (#500). CI now writes JUnit and fails the job on any skip. CMake skip semantics are unchanged for local runs. Claude-Session: https://claude.ai/code/session_01DgW4n3qYgi5jAAohEmXRQj Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
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.
Cause
anyps5_metal_presentation_replayfailed withNative guest memory address 0x... is not borrowed(plus a 10 s queue stall message as a side effect of the aborted flip).#494 made VideoOut store its buffer labels on flip completion (
VideoOutState.cpp:AgcDriverStoreGuestLabel_nid_postfix(&cfg->bufferLabels[previous], 0)) and addedMetalReplay/GuestLabelStore.cppfor the replay executables. That replay store calledGuestMemory::CheckRange, which in the replay build is the borrowed-range check. The label array is VideoOut's own host memory, never part of the ranges a replay borrows, so the check threw on the first flip that replaced a frame.First bad commit
6ef7160ef(#494, "sync: small HLE libraries from upstream ...").git bisectbetween361d4263d(good, test passes) andorigin/main(bad); the parentf5f20ad1c(#495) is good. It is not caused by batching or copy-back deferral, which is why those switches did not change it, nor by #495/#496/#497/#498.Fix
The replay label store is a plain release store, with a comment why it is not range-checked. The production Metal
DirectDriverstore is unchanged (its check is host-accessibility, which holds for the label array).Evidence
ctest -R anyps5_metal_presentation_replayfails on main (arm64 Release, Ninja, guest-replay flags from macos-metal.yml).Why CI was not red
The guest-replay job did run it on main (
1eee7d2ad), but the test returned exit 77 (no console session / screen locked on the self-hosted runner), which the CMakeSKIP_RETURN_CODE 77reports asSkipped:44/60 anyps5_metal_presentation_replay ... ***Skipped 0.15 sec. A skip does not fail the job, so the regression was invisible whenever the runner screen was locked.Codex review findings
🤖 Generated with Claude Code