From 28557119a7caa0699d5b44b3aa4bddf6d5cf7d3c Mon Sep 17 00:00:00 2001 From: Bryan Newbold Date: Mon, 4 Oct 2021 12:49:45 -0700 Subject: html ingest: report dt with broken CDX records --- python/sandcrawler/html_ingest.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'python/sandcrawler') 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), -- cgit v1.2.3