next.js

7e01f56f - Skip reader task locking for immutable Turbo Tasks reads (#96179)

Skip reader task locking for immutable Turbo Tasks reads (#96179) ## What changed Tracked output and cell reads now acquire the target task first. If the target is already immutable, the read skips locking the reader task because no dependency edge can be added and the target cannot be invalidated. Mutable targets drop the speculative target lock and reacquire the target/reader pair together, preserving the existing invalidation-race guarantee. `verify_immutable` deliberately retains the pair-lock path. Listener diagnostics also preserve the tracked reader ID when the fast path skips its guard. The separate current-task ID cache is in #96180. This PR is now backend-only so the lock-ordering change can receive focused backend review. ## Why The previous path locked both the target and reader task for every tracked read. That synchronization is necessary for mutable targets but redundant once the target is immutable. Instrumentation found that 1,541,880 of 4,501,364 sampled tracked reads (34.254%) could use the bypass. ## Performance evidence Measured at `0950e8ae05db20c33302f02257114f5b857edabc` on a dedicated 8-vCPU/16-GiB Vercel Sandbox with frozen side-by-side binaries. These measurements are directional evidence, not precise estimates. Small percentage changes are difficult to resolve reliably. This campaign and #96180's incremental campaign used separate run sets: the absolute endpoint here does not match #96180's starting point, which demonstrates cross-campaign noise. The values must not be compared across PRs, and their percentages must not be added. ### Browser-driven HMR, 30-second B/C/C/B - Evaluation: 20.771360 → 19.176645 ms (**-7.677%**) - Commit: 21.714832 → 20.371425 ms (**-6.187%**) ### Production builds - First 12-campaign matrix: **-1.931%** - Complementary 12-campaign matrix: **-1.990%** - Combined 24-campaign aggregate: **-1.959%** Five of six balanced blocks favored the candidate. The conservative six-block 95% interval was -4.042% to +0.248%, so the production estimate is supporting evidence rather than a precise claim. ### Causal profiles - HMR `Storage::access_pair_mut` self time fell from 3.19%/3.19% to 1.21%/1.18%. - HMR voluntary context switches fell 32.898%. - Production pair-access self time and voluntary context switches also fell; production voluntary switches changed -14.552%. - All accepted perf captures reported zero lost samples. ## Validation On the exact backend-only head (`9ccfd6753526f716466a9584fba8c8753b7d83bf`): - Complete `turbo-tasks-backend` package tests, integrations, and doctests - Complete backend package matrix with `--features verify_immutable` - `cargo fmt --check -p turbo-tasks-backend` - `cargo clippy -p turbo-tasks-backend --all-targets -- -D warnings` - `git diff --check` - Conflict-free merge-tree check against current `canary` (`5cdcbbb05b`)