fix(windows): size all D3D11 render pools to five slots — restores the #267 pool fix lost in merge

[petercr]

Summary

The #267 merge landed the ScreenCamera smokes, diagnostics, and drain pacing, but kept the pre-#262 pool sizing (windows_d3d11_render_pool_slots returns 3 for screen-only, 5 for camera sessions). The reviewed #267 stack sized all pools to five slots unconditionally — post-#262 stop/preview-restore accounting raised steady-state churn for every session shape, and a 3-slot screen-only pool starves CFR ticks into silent single-frame skips under fence lag.

Evidence (win11-x64-i5-8400-gtx1650-super, GTX 1650 SUPER)

With main's sizing, a screen-only direct-D3D11 session shows texturePoolPressureEvents > 0 and cadence loss; with all pools at 5 slots, pressure events are 0 across runs and observed fps recovers above the issue floor. This sizing is also the leading suspect for the open P0 second-session decay (#271 release record), since decay reproduces only with real camera sessions where lease residency stretches furthest.

Changes

• windows_d3d11_render_pool_slots returns 5 unconditionally (still inside the per-format pool ceiling of 8)
• Test renamed to windows_d3d11_sessions_size_deep_render_pools asserting both shapes = 5

Testing

• cargo test -p videorc-backend: 1519 passed, 0 failed
• cargo fmt --check --all: clean
• Real-app screen-only smoke (--d3d11 --require-d3d11): PASS, pressure 0, 29.5fps observed on 1080p30

Related: open P0 "second-session recording lag" in docs/releases/0.9.71.md

Summary by CodeRabbit

• Bug Fixes
  • Improved rendering consistency for screen-only sessions by ensuring they use the same pool capacity as sessions with camera input.
  • Updated internal validation to reflect the standardized rendering capacity.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 6ef276f8-22e0-4b05-8481-942bcba38f42

📥 Commits

Reviewing files that changed from the base of the PR and between 92f9149 and 30b1dd8.

📒 Files selected for processing (1)

• crates/videorc-backend/src/windows_d3d11_session.rs

Included review availability: Your plan provides up to 2 included reviews per hour; 1 remains after this review.

---

📝 Walkthrough

Walkthrough

The D3D11 render pool now always allocates five slots. The existing test was renamed and updated to verify five slots for non-camera sessions.

Changes

D3D11 render pool sizing

Layer / File(s)	Summary
Set and validate fixed pool size crates/videorc-backend/src/windows_d3d11_session.rs	windows_d3d11_render_pool_slots returns five slots for all sessions. The test name and non-camera assertion now reflect the fixed size.

Estimated code review effort: 1 (Trivial) | ~5 minutes

Merge Risk: 🔵 Low · up to 30b1d

The pool-sizing change has no identified product-impacting defect, but the Rust lint check did not complete. The PR is mergeable with explicit owner follow-up to rerun and complete that check; no actionable runtime merge blocker is otherwise supported.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly and concisely describes the Windows D3D11 render pool sizing change to five slots.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 4 functions across 1 files.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[petercr]

Rebased onto current main — still needed

Refreshed the branch onto fd49a359 (0.9.81-alpha.1 line, post-#305/#307/#308): the diff is now just the 17-line sizing change. Confirmed main still sizes screen-only pools to 3 slots (windows_d3d11_render_pool_slots → if camera_required { 5 } else { 3 }), and nothing in #278/#279/#303/#305/#307/#312 changed pool depth or the per-format ceiling of 8.

Still applies and still matters because:

• The evidence stands: on win11-x64-i5-8400-gtx1650-super, screen-only direct-D3D11 sessions at 60fps show texturePoolPressureEvents > 0 and CFR cadence loss with 3-slot pools; 5 slots → 0 pressure events.
• The fix(windows): reset D3D11 drain state, always restore preview after stop, release direct-D3D11 consumer, frame accounting at stop #262 stop/preview-restore accounting (still in main) keeps steady-state churn high for every session shape, which is exactly what a 3-slot screen-only pool cannot absorb.
• Orthogonal to fix(windows): Intel Iris Xe probe fallback ladder for E_UNEXPECTED at 6000 CBR #307/fix(windows): finalize fallback recordings as MP4 #312 (those fix encoder selection/finalization; this fixes render-pool depth under fence lag).

Gates: cargo test -p videorc-backend windows_d3d11_sessions_size pass (renamed test asserts 5 for both shapes), cargo fmt --check --all clean. CI should now run against the clean one-file diff.