http.cookiejar.domain_match() matches an interior substring instead of a suffix

[metsw24-max]

domain_match(A, B) is documented to return True when A has the form NB, i.e. A ends with B (RFC 2965, section 1: "x.y.com domain-matches .Y.com but not Y.com"). It computes i = A.rfind(B) and only rejects i == -1 or i == 0, so a B that appears as an interior substring of A also matches:

>>> from http.cookiejar import domain_match
>>> domain_match("www.acme.com.evil.org", ".acme.com")
True

www.acme.com.evil.org does not end with .acme.com, so this should be False. The sibling helper user_domain_match() already uses A.endswith(B). This feeds the RFC 2965 (version > 0) cookie domain checks in set_ok_domain/return_ok_domain and is_third_party().

Linked PRs

• gh-150522: anchor suffix check in http.cookiejar.domain_match #150523