The GitHub Actions job "Required Checks" on texera.git/loop-feb has failed.
Run started by GitHub user aglinxinyuan (triggered by aglinxinyuan).

Head commit for run:
ce05a66c4a8bb859128c7c0111797a19b2a9b97a / Xinyuan Lin <[email protected]>
refactor(amber): simplify loop runtime and trim duplicated tests/narrative

Review-driven cleanup of the loop feature (net -235 lines), no mechanism
changes:

- _eval_loop_expr evaluates the user expression with eval() instead of
  exec("output = " + expr), so "output" stops being a reserved name and
  _RESERVED_STATE_KEYS shrinks to {"table"}. Two deliberate user-visible
  changes: a loop variable named `output` now persists like any other,
  and multi-statement expressions now raise SyntaxError instead of
  silently executing.
- Drop the dead scratch-key strip in _jump_to_loop_start (run_update is
  @overrides.final and always persists _strip_reserved(...) before any
  jump can fire) and inline the one-line _compute_loop_start_id wrapper.
- Drop the dead .withSuggestedWorkerNum(1) (WorkerConfig forces
  workerCount=1 for non-parallelizable ops) and its mixin pin.
- Deduplicate tests: merge the twin loop_counter>0 pass-through tests;
  merge the two happy-path jump tests into one that also pins
  RPC-before-write ordering via a shared event log; fold the
  envelope-exclusion assertions into the consume test; drop the 8 pyb
  tricky-input descriptor tests (PythonTemplateBuilderSpec on main pins
  the base64 guarantee) and the engine-simulation single-loop test
  (LoopIntegrationSpec runs it for real with identical expressions).
- Keep one canonical home per design narrative (proto field doc for the
  setup-config story, table_to_ipc_bytes for Arrow-IPC-not-pickle, the
  reset_output_storage call site for the inner-reset story); other
  copies reduced to pointers. Fix three stale TestLoopCounterRuntime
  cross-references.

Report URL: https://github.com/apache/texera/actions/runs/28586671928

With regards,
GitHub Actions via GitBox

Reply via email to