microlinkhq / browserless / 28332040892 / 2

85%

master: 85%

push

github

perf(screencast): frame backpressure (#812)

* perf(screencast): add frame backpressure

* fix(screencast): harden frame backpressure (review follow-up)

Addresses review findings on the frame-backpressure change:

- recorder: write a frame's cluster header and payload back-to-back, then
  await drain once, instead of deferring the payload across an await. This
  stops a concurrent flush() emit from splicing between a frame's header and
  payload and corrupting the Matroska stream; removes the now-dead writeChunks.
- recorder: stop writing once ffmpeg stdin errors/closes, so a backpressured
  frame is not written to a dead stream.
- screencast: swallow a synchronous onFrame throw (symmetric with the async
  rejection path) rather than rethrowing into puppeteer's CDP dispatch loop.
- screencast: guard the deferred ack so a frame that settles after stop() does
  not ack a torn-down session.
- tests for each.

Also carries pre-existing working-tree changes: screencast jpeg/quality option
defaults (DEFAULT_OPTS) with their matching README and test.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(screencast): clear stopped flag on start() so restarts ack again

stop() sets `stopped` to suppress a late ack from a frame that settles after
teardown, but start() never cleared it — so a stop()/start() on the same
screencast instance left promise-returning onFrame handlers un-acked, hanging
the restarted capture. Reset it in start(). (Cursor Bugbot, medium.)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

37 of 38 branches covered (97.37%)

Branch coverage included in aggregate %.

110 of 110 relevant lines covered (100.0%)

6.52 hits per line