René's URL Explorer Experiment

Title: stream: fix early drain after Utf8Stream reopen by mcollina · Pull Request #65633 · nodejs/node · GitHub

Open Graph Title: stream: fix early drain after Utf8Stream reopen by mcollina · Pull Request #65633 · nodejs/node

X Title: stream: fix early drain after Utf8Stream reopen by mcollina · Pull Request #65633 · nodejs/node

Description: If a 'ready' listener starts a write after reopen(), the reopen path still emits 'drain' from a nextTick before that write completes, so a listener reading the file on 'drain' can observe it empty. This is the underlying race behind the test-fastutf8stream-reopen flake that 2e8a4b1 deflaked on the test side ("The reopen path also emits a 'drain' of its own from a nextTick before the write has landed") — this fixes the library so 'drain' is never emitted while stream.writing === true. Skip the extra emit when a write is in flight: #release() emits the real 'drain' once the write completes, so no event is lost. The regression test defers the reopened file's fs.write by one setImmediate, which deterministically lands the write after the nextTick on which the premature 'drain' used to fire. It fails without the lib change (asserts writing === false on 'drain'; observed true, file empty) and passes with it. Verified with 800 runs of the test under 16× parallel CPU contention: 0 failures (28/800 before the fix, reproducing the CI signature exactly). Same fix submitted to upstream sonic-boom: pinojs/sonic-boom#255.

Open Graph Description: If a 'ready' listener starts a write after reopen(), the reopen path still emits 'drain' from a nextTick before that write completes, so a listener reading the file on 'drain...

X Description: If a 'ready' listener starts a write after reopen(), the reopen path still emits 'drain' from a nextTick before that write completes, so a listener reading the file ...

Opengraph URL: https://github.com/nodejs/node/pull/65633

X: @github

direct link

Domain: github.com

Links:

Viewport: width=device-width