Fix Hashnode syndication recovery - #5520
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a0002945bf
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| entry = post.get(platform, {}) | ||
| return bool( | ||
| isinstance(entry, dict) | ||
| and entry.get("published") is False |
There was a problem hiding this comment.
When recovering an older failed Hashnode run, this predicate still returns false for draft-only records that predate the new published: false field. For example, scripts/website/syndication-state.json already has metal-and-skins with a https://hashnode.com/draft/... URL and no published key, so --post-slug metal-and-skins --platforms hashnode will hit state.is_syndicated(...), fail this recovery check, and skip the very draft the new recovery path is meant to unblock. Treating draft URLs, or missing published on a draft URL, as retryable would cover those legacy records.
Useful? React with 👍 / 👎.
| if not retry_failed_draft: | ||
| print(f" [{adapter.name}] already syndicated; skipping.") | ||
| continue | ||
| print(f" [{adapter.name}] retrying explicitly recorded unpublished draft.") |
There was a problem hiding this comment.
This branch says it is retrying the recorded unpublished draft, but the selected draft URL is never passed to HashnodeAdapter; submit_draft() still starts from the dashboard and clicks Write, which edits whichever current draft Hashnode opens. In a real recovery after later browser runs have left a different draft in that slot, --post-slug can overwrite or publish that other draft instead of the selected post, so the recovery path should navigate to and verify the stored draft_url before editing.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
Pull request overview
This PR tightens the Hashnode browser-syndication flow so that a retained draft (when the publish dialog fails) is no longer persisted as a “successful” syndication, and adds an explicit recovery mode to reprocess a previously recorded unpublished draft without altering the normal candidate rotation.
Changes:
- Introduces
_completed_hashnode_result()to only persist Hashnode state after a public publish succeeds (otherwise fails the adapter run). - Adds
--post-slugto target a single eligible post, with special-case logic to allow retrying a previously persisted unpublished Hashnode draft. - Adds regression tests for “unpublished draft is not treated as completed” and for the explicit failed-draft retry predicate.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| scripts/website/syndicate_browser_posts.py | Stops persisting unpublished Hashnode drafts as completed, adds --post-slug selection + retry handling, and improves runner output messaging. |
| scripts/website/test_syndicate_browser_posts.py | Adds unit tests for Hashnode completion behavior and the failed-draft recovery predicate. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| post = state.raw.get("posts", {}).get(slug, {}) | ||
| if not isinstance(post, dict): | ||
| return False | ||
| entry = post.get(platform, {}) | ||
| return bool( | ||
| isinstance(entry, dict) | ||
| and entry.get("published") is False | ||
| and entry.get("draft_url") | ||
| ) |
| print( | ||
| f"Post {candidate.slug} is not eligible on {today.isoformat()} " | ||
| f"with a {args.min_age_days}-day delay.", | ||
| file=sys.stderr, | ||
| ) |
Cloudflare Preview
|
What changed
--post-slugrecovery path for a known unpublished draft without changing the normal one-post rotation.Why
The browser flow could fail while setting tags, canonical URL, or clicking Publish, then record the retained draft URL in
syndication-state.json. Candidate selection treated any URL as complete, so the post was never retried and the health checker could not see the missing public publication.Validation
python3 scripts/website/test_syndicate_browser_posts.pypython3 scripts/website/test_syndication_health.pypython3 scripts/website/test_syndicate_blog_posts.pypython3 -m py_compile scripts/website/syndicate_browser_posts.py scripts/website/check_syndication_health.py scripts/website/test_syndicate_browser_posts.pyjq empty.