Fix TTL extraction [RHELDST-20510]

pubtools/_pulp/cdn.py

@@ -26,7 +26,7 @@ class CdnClient(object):
     _EXPONENT = float(os.environ.get("CDN_RETRY_EXPONENT", "3.0"))
     _MAX_SLEEP = float(os.environ.get("CDN_RETRY_MAX_SLEEP", "120.0"))
 
-    TTL_REGEX = re.compile(r".*/(\d+[smhd])/.*")
+    TTL_REGEX = re.compile(r"/(\d+[smhd])/")
     CACHE_KEY_HEADER = HeaderPair("akamai-x-get-cache-key", "X-Cache-Key")
 
     def __init__(self, url, max_retry_sleep=_MAX_SLEEP, **kwargs):
@@ -104,7 +104,7 @@ def _get_ttl(self, path):
         out = self._get_headers_for_path(path, headers)
 
         def _parse_ttl(value):
-            parsed = re.match(
+            parsed = re.search(
                 self.TTL_REGEX, value.get(self.CACHE_KEY_HEADER.response) or ""
             )
             return parsed.group(1) if parsed else None

tests/cdn/test_cdn_client.py

@@ -24,8 +24,14 @@ def test_format_arl_template(requests_mock, caplog):
             ("https://cdn.example.com/content/foo/test-path-2/other-file.xml", "30m"),
         ]
 
+        # ARLs are generated from the template using the {ttl} placeholder, which is replaced with
+        # the real TTL value. The real TTL value is extracted from the cache key header of the real
+        # request for the given path using '/(\d+[smhd])/' regex.
+        # The /1h/foo in the mocked header here is to test that if the path contains a component
+        # that also matches the TTL regex ('/1h/'), it will still find the correct value
+        # ('/10h/' or '/30m/').
         for url, ttl in url_ttl:
-            headers = {"X-Cache-Key": f"/fake/cache-key/{ttl}/something"}
+            headers = {"X-Cache-Key": f"/fake/cache-key/{ttl}/something/1h/foo"}
             requests_mock.register_uri("HEAD", url, headers=headers)
 
         fts = []