rc-docs-sync: collapse multiple unprocessed RCs to the latest per run#402
Merged
Conversation
Run 26561237947 (2026-05-28 schedule) processed both 27.8-RC1 and 27.8-RC2 because the resolver fanned out one matrix entry per unprocessed RC. The agent then independently re-documented the same wpseo_custom_fields_pre_query filter, producing PR #400 (for RC1) and PR #401 (for RC2) — identical titles, identical work, one of them obvious waste. Process only the latest unprocessed RC per product per run. The earlier unprocessed RCs are silently superseded: - The resolver collects them in a new superseded_actions list (paired with the RC that supersedes them). - A new "Mark superseded RCs" step posts a brief marker comment on the tracking issue for each, mirroring the dedup style of the safety-net step (skips if a marker already exists). - The latest RC's diff base is computed as before: if a prior RC of the same base is in processed_markers, diff against that (incremental); otherwise against the latest stable (full release cycle). When multiple never-seen RCs of the same base collapse to the latest, that's a full stable-cycle diff which captures any net public-surface changes (an RC2 that reverts an RC1 change shows neither; an RC2 that adds on top of RC1 shows both). - workflow_dispatch with an explicit input_rc_tag bypasses the collapse, preserving the manual-backfill path. Synthetic resolver tests (5 scenarios) cover: the bug case, the already-processed-prior-RC case, single-RC behavior unchanged, empty queue, and the input_rc_tag bypass — all pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Deploying yoast-developer with
|
| Latest commit: |
55bad1c
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://89f2d5fb.yoast-developer.pages.dev |
| Branch Preview URL: | https://rc-docs-sync-collapse-to-lat.yoast-developer.pages.dev |
There was a problem hiding this comment.
Pull request overview
This PR updates the RC docs sync workflow to avoid opening duplicate documentation PRs when multiple unprocessed RC tags are found in the same scheduled run.
Changes:
- Collapses multiple unprocessed RCs to only process the latest tag.
- Records skipped RCs as superseded in the tracking issue.
- Adds resolver output and workflow step support for superseded markers.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Copilot caught this on #402: if the workflow misses RCs from two different cycles (e.g. 27.8-RC1 and 27.9-RC1 both unprocessed), the previous logic collapsed both into 27.9-RC1 against the latest stable, which never sees 27.8-RC1's public surface. Group unprocessed RCs by base_version, keep only the latest per group. Each base gets its own queue entry diffed against the previous stable. Synthetic test (scenario 6) confirms the fix.
Copilot caught: posting superseded markers in the resolve job advances state past the skipped RCs before the matrix entry that's supposed to cover them has finished. If the latest RC's clone or bundle step fails (no marker for it gets posted), the next run treats the superseded RCs as `same_base_processed` and uses them as diff bases — missing changes between stable and the skipped RC. Move the marker-posting into the matrix job: each queue entry now carries its own `superseded_rcs` list, and a final per-entry step posts them only after verifying the entry's own marker is already on the tracking issue. A clone/bundle failure now leaves the superseded RCs unmarked, so the next run re-picks them up correctly.
- last_key was reading processed_markers[-1] (document order). Superseded markers are posted *after* the superseding RC's marker, so document order would identify an older (superseded) RC as the high-water mark and re-queue an already-processed RC. Use max by sort_key instead. - superseded_by_latest was initialized only inside the non-dispatch branch, so workflow_dispatch with explicit input_rc_tag would NameError when building the queue entry. Initialize at the top of each per-product iteration (also prevents accidental cross-product leakage once PRODUCTS has more than one entry). Synthetic test suite extended to 9 scenarios, including the document-order marker case and the dispatch-bypass case — both now pass.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Run 26561237947 produced duplicate PRs #400 and #401 because the resolver fanned out a matrix entry per unprocessed RC.
Fix: collapse multiple unprocessed RCs to just the latest per base version (so two cycles' RCs aren't accidentally merged). Each matrix entry carries a
superseded_rcslist; the matrix step itself posts those superseded markers, gated on the latest RC's own marker being on the tracking issue first. If clone/bundle fails, no superseded markers are posted and the next run re-picks up the skipped RCs.last_keyusesmax(processed_markers, key=sort_key)so out-of-order markers (superseded or manual backfill) can't reset state.workflow_dispatchwith explicitrc_tagbypasses collapse.9 synthetic resolver scenarios cover bug case, mid-cycle, single-RC, empty, dispatch bypass, mixed bases, document-order-after-superseded, and backfilled-older-RC.