The GitHub Actions job "Lint PR" on texera.git/state-handshake-redesign has succeeded. Run started by GitHub user aglinxinyuan (triggered by aglinxinyuan).
Head commit for run: 6f0902ecd349f16aac2167a33967a17c4e4410fa / Xinyuan Lin <[email protected]> fix(amber): keep per-task finally switches, drop run-loop end-of-body switch Earlier commits on this branch removed the per-task `finally: self._switch_context()` from process_state, process_internal_marker, and process_tuple, and relied on the run-loop's end-of-body switch (line 65) to park DataProcessor between tasks. Per review on #4560, swap to the equivalent design that keeps the per-task finally switches and drops the run-loop's end-of-body switch instead. The diff to data_processor.py is now confined to DataProcessor.run -- the three task-handler bodies are unchanged from origin/main: - process_state, process_internal_marker, process_tuple keep their original `finally: self._switch_context()`. DataProcessor parks at whichever per-task finally just ran, between tasks. - DataProcessor.run no longer does a post-init _switch_context (would burn MainLoop's first switch on a no-op handshake) and no longer does the end-of-body _switch_context after each task (the per-task finally already plays that role). The body is just dispatch + peek-then-consume. Both designs are functionally equivalent; this one minimizes the source diff and addresses the review concern about touching the tuple code path. All 8 tests in test_main_loop.py pass. Report URL: https://github.com/apache/texera/actions/runs/25141278296 With regards, GitHub Actions via GitBox
