Repository navigation
v1.5.0 — Fix Cash Converters scraping + diagnostic logging
Bug fix
🔒 Cash Converters scraping works again
Cash Converters recently rotated their TLS certificate and their server now sends only the site certificate, omitting the intermediate that signs it. Browsers paper over this automatically; Node cannot, so every request failed and each scrape finished instantly with 0 results.
The missing intermediate is now bundled with the add-on, so the certificate chain is complete again. Full certificate verification stays enabled — this supplies the missing link rather than turning off security checks.
Diagnostics
🔍 Logs that actually say what went wrong
Previously a failure only ever logged fetch failed. The add-on log now shows the real cause and a plain-English hint, for example:
ERROR [cc] GET https://shop.cashconverters.co.nz/Browse?... failed on attempt 3/3 after 166ms:
fetch failed | cause: UNABLE_TO_VERIFY_LEAF_SIGNATURE: unable to verify the first certificate
| hint: the site did not send its full TLS certificate chain
WARN [cc] 1/1 search term(s) failed: eufy
ERROR [cc] Every search term failed — the site was unreachable or rejected every request
Also added: startup diagnostics (Node version, database path, certificate bundle status), per-run summaries, and retry attempts with timings. Set LOG_LEVEL=debug for a line per HTTP request.
Other fixes found while investigating
- Failed scrapes no longer delete your listings. The "remove listings that disappeared" step ran even when searches failed, so a network outage could wrongly mark healthy listings as ended. It's now skipped unless the search phase fully succeeded.
- Failed scrapes are reported as failures. A run where every search failed was recorded as "completed" with 0 results. It now reports as failed, and the UI says the site couldn't be reached instead of showing a misleading "Found 0 listings".
- The add-on can no longer crash on a scraper error, which previously could restart it mid-scrape.
- Requests now time out after 30 seconds instead of potentially hanging forever (configurable via
REQUEST_TIMEOUT_MS).