next.js

faada8b4 - Turbopack: reap dead worker threads and fail their in-flight tasks

Turbopack: reap dead worker threads and fail their in-flight tasks A worker thread that died unexpectedly (boot failure, crash, OOM, a loader calling process.exit) was never reaped: the JS loaderWorkers map retained the dead Worker, the pool kept handing its id out as healthy, and each assigned task queued its payload on a channel no consumer drained and then waited forever for a response. - Report Worker error/exit to Rust (capturing threadId at creation: it is -1 by the time the exit event fires) and fail the pending creation when a worker dies before announcing itself. - Pair creations with acquirers by a nonce instead of a global FIFO that mis-paired across pools and after canceled/failed creations; terminate unclaimed workers instead of retaining them. - Wake in-flight tasks with a per-worker death signal so they fail instead of hanging, remove the dead worker from the pool and its routed channel, and never return a dead worker to the pool. The causal e2e schedules process.exit inside a loader: baseline hangs the compile forever; the fix fails it with a crash issue and recovers on a fresh worker.