diff options
| author | Bryan Newbold <bnewbold@archive.org> | 2021-10-04 12:49:45 -0700 |
|---|---|---|
| committer | Bryan Newbold <bnewbold@archive.org> | 2021-10-04 12:49:45 -0700 |
| commit | 28557119a7caa0699d5b44b3aa4bddf6d5cf7d3c (patch) | |
| tree | ed3c8c1a5e235a87a3b233ca63c8951f48f30004 | |
| parent | ba3ee68b7789243921f0063461df1e2f7da65256 (diff) | |
| download | sandcrawler-28557119a7caa0699d5b44b3aa4bddf6d5cf7d3c.tar.gz sandcrawler-28557119a7caa0699d5b44b3aa4bddf6d5cf7d3c.zip | |
html ingest: report dt with broken CDX records
| -rw-r--r-- | python/sandcrawler/html_ingest.py | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/python/sandcrawler/html_ingest.py b/python/sandcrawler/html_ingest.py index e81a16d..f11cac4 100644 --- a/python/sandcrawler/html_ingest.py +++ b/python/sandcrawler/html_ingest.py @@ -168,7 +168,7 @@ def fetch_html_resources(resources: List[dict], wayback_client: WaybackClient, w raise NoCaptureError(f"HTML sub-resource not found: {resource['url']}") file_meta = gen_file_metadata(wayback_resp.body, allow_empty=True) if file_meta['sha1hex'] != wayback_resp.cdx.sha1hex: - raise WaybackContentError(f"wayback payload sha1hex mismatch: {wayback_resp.cdx.url}") + raise WaybackContentError(f"wayback payload sha1hex mismatch: {wayback_resp.cdx.datetime} {wayback_resp.cdx.url}") full.append(WebResource( surt=wayback_resp.cdx.surt, timestamp=parse_cdx_datetime(wayback_resp.cdx.datetime), |