Incremental Update: Ashutosh Bapat responds to Peter Eisentraut's concerns
Ashutosh's reply addresses the two specific blockers Peter raised in his previous message:
1. Patch Application Issue — Clarified but Not Fully Resolved
Ashutosh explains that he maintains all SQL/PGQ fixes in a single local branch, which is why the patch was numbered 5/5. He reports that both git cherry-pick and git am succeed on a fresh branch from his end, and asks Peter to confirm whether the issue persists. This doesn't fully resolve the concern — Peter may have been applying against a different base (plain master without patches 1–4), which would naturally fail. The underlying hygiene issue of submitting a dependent patch without its prerequisites remains, though Ashutosh seems unaware that this is the root cause.
2. Test Reproducibility — Key Admission and Fix
This is the substantive progress: Ashutosh acknowledges that most of the test queries from Satya's original patch do not actually reproduce the bug on current master. He states that only two of the original test queries trigger the failure. He has now:
- Included a test query that demonstrably fails without the code fix, with the expected error message included — directly addressing Peter's request for a reproducing test case on master.
- Removed the separate test section, folding the query into one of the earliest existing sections in the test file, which is a minor organizational cleanup.
This is the critical fix to the submission: the prior version's test passed even without the code change (as Peter discovered), making the patch appear unnecessary. The new version should have a test that actually turns red without the fix, proving the bug still exists on master.
What's Still Open
- No new patch file appears to be attached in this message (or at least none is quoted), so Peter will need to evaluate the updated submission when it arrives.
- The
p_resolve_unknownscorrectness question (raised by Ashutosh himself in earlier rounds) is not revisited here — presumably the code still respects it. - Peter has not yet responded to this clarification.