[CLSA-2026:1790760686] Fix CVE(s): CVE-2026-15806, CVE-2026-17084
Type:
security
Severity:
Moderate
Release date:
2026-09-30 09:31:39 UTC
Description:
* SECURITY UPDATE: urllib.request handed out credentials registered for an https:// URI to the matching http:// URI. HTTPPasswordMgr.reduce_uri() drops the scheme, and add_password()/find_user_password() compared only the reduced (authority, path) pair, so an attacker able to force an HTTPS-to-HTTP downgrade or redirect received the Basic-auth credentials in cleartext. HTTPPasswordMgrWithPriorAuth had the same flaw in update_authenticated()/is_authenticated(). Credentials are now scoped by scheme; a URI registered without a scheme still matches any scheme, which keeps proxy authentication working. - debian/patches/CVE-2026-15806.patch: backport of cpython a2773a34183b7d94a243bb98fd658926cc5348ce (gh-155694, GH-155696), the 3.13-branch form of the fix. The code change and the three new tests in Lib/test/test_urllib2.py are upstream's verbatim; only hunk offsets were re-cut. - CVE-2026-15806 * SECURITY UPDATE: the stringprep module, used by the "idna" codec (IDNA 2003), applied post-Unicode-3.2.0 case mappings that RFC 3454 forbids. map_table_b3() falls back to str.lower(), which uses the interpreter's own Unicode 11.0.0 data, for every codepoint missing from the generated b3_exceptions table, so characters that gained a lowercase mapping after 3.2.0 (Cherokee, Cyrillic Palochka, Georgian Mtavruli, Adlam, ...) were case-folded, and different Python versions derived different A-labels for the same domain name. - debian/patches/CVE-2026-17084.patch: backport of cpython 1e54caa096678a38afcabecabb1ff72400dd6bae (gh-155292, GH-155293), the 3.14-branch form of the fix, covering Lib/stringprep.py, Tools/unicode/mkstringprep.py and the new idna test in Lib/test/test_codecs.py. - The b3_exceptions table is generated from the running interpreter's Unicode database, so upstream's Unicode 16.0.0 table does not apply. Lib/stringprep.py was regenerated with the patched mkstringprep.py under this version's own interpreter (Unicode 11.0.0), adding 634 identity entries. The new unidata_version assertion makes a future Unicode database bump fail loudly instead of leaving the table stale. - CVE-2026-17084 * The Misc/NEWS.d fragments and documentation changes of both upstream commits are omitted, as is the Tools/unicode/makeunicodedata.py hook, which only matters when the bundled Unicode database is rebuilt.
Updated packages:
  • alt-python37_3.7.17-34_amd64.deb
    sha:4595a7a55dc6d0e7c3a8c35e55df2d37c1e359bb
  • alt-python37-debug_3.7.17-34_amd64.deb
    sha:bdb08c8b47aa0c6164e1415ef266a697884998d5
  • alt-python37-devel_3.7.17-34_amd64.deb
    sha:0f5349f1b0df6afff00539556d9581914ce72566
  • alt-python37-libs_3.7.17-34_amd64.deb
    sha:cd5112b1332f36280622e95bbdfdab9dbcf56ba2
  • alt-python37-test_3.7.17-34_amd64.deb
    sha:4dd4b639edc04f24bf8b44926cd3ec9d0d986e5b
  • alt-python37-tkinter_3.7.17-34_amd64.deb
    sha:5cd6637e4bda103928132bc3335bad51a4a7b77c
  • alt-python37-tools_3.7.17-34_amd64.deb
    sha:ff102f8ba1beaa906ea184dfa051ab32010352d4
Notes:
This page is generated automatically and has not been checked for errors. For clarification or corrections please contact the CloudLinux Packaging Team.