Bug 2572288

Summary: libHX: source signature does not verify with openpgpverify (SHA-1 data signature and key self-signatures)
Product: [Fedora] Fedora Reporter: Jakub Jelen <jjelen>
Component: libHXAssignee: Till Maas <opensource>
Status: NEW --- QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: rawhideCC: aflyhorse, opensource, steve
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: ---
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On:    
Bug Blocks: 2569052    

Description Jakub Jelen 2026-11-12 08:48:36 UTC
As part of the Fedora 45 Change Sequoia openpgpverify (https://fedoraproject.org/wiki/Changes/Sequoia_openpgpverify), %gpgverify (GnuPG) is being replaced by %openpgpverify (Sequoia sqv). For libHX, the switch can't be done in a simple PR: the current check with gpgv accepts the current upstream signature, but %openpgpverify rejects it.

Failing check (dist-git commit c99e81fdfdb3):

    %openpgpverify --keyring=gpgkey-B56B8B9D9915AA8796EDC013DFFF2CDB19FC338D.gpg --signature=libHX-3.22.tar.asc --data=-

    + xzcat /builddir/new/libHX-3.22.tar.xz
    Signing key on B56B8B9D9915AA8796EDC013DFFF2CDB19FC338D is not bound:
               No binding signature at time 2014-08-25T15:19:26Z
      because: Policy rejected non-revocation signature (PositiveCertification) requiring second pre-image resistance
      because: SHA1 is not considered secure
    0 of 1 signatures are valid (threshold is: 1).
    openpgpverify: Signature verification failed.

Why:

The release signature (SHA-1, made 2014-08-25) is by the key B56B8B9D9915AA8796EDC013DFFF2CDB19FC338D (RSA 4096 bits). sqv reports only the first problem it hits (above). All the problems found:

- The self-signatures that bind the key's user IDs to it use only SHA-1. SHA-1 is no longer considered secure: collisions have been practical since 2017 (SHAttered) and chosen-prefix collisions since 2020 (SHA-1 is a Shambles), and the crypto policy rejects it. GnuPG doesn't check self-signatures; Sequoia does, and without a valid one it doesn't use the key at all.
- The release signature itself uses SHA-1 too; for signatures over data, this won't be relaxed.

The copy of the key on keys.openpgp.org, keyserver.ubuntu.com and WKD has the same problems.

What can be done:

- Ask upstream to refresh the key's self-signatures with SHA-256 and publish the updated key (keys.openpgp.org, WKD or the project website):

      gpg --cert-digest-algo SHA256 --quick-set-expire FPR 2030-12-31
      gpg --cert-digest-algo SHA256 --quick-set-expire FPR 2030-12-31 '*'

  Use an absolute expiry date (or 0). A new SHA-256 release signature alone still fails while the key is only bound with SHA-1.
- Then ask upstream to sign releases with SHA-256 (gpg --digest-algo SHA256, or personal-digest-preferences SHA256 in gpg.conf): the current release again, or the next one. Then update the keyring and the signature in dist-git.
- If upstream already signs with a newer key, ask them to use SHA-256 with it and switch the keyring to that key.
- Until then the package can keep the current GnuPG check, or verify the tarball differently (for example against a checksum published over HTTPS).

The rest of the conversion is mechanical. Once the signature or key is sorted out, the spec change replaces the direct GnuPG call with one %{openpgpverify} call (same keyring, signature and data) and adds BuildRequires: openpgpverify; it is ready and tested up to the signature check. I'm happy to send the PR.

(The check was run with scripts from https://github.com/Jakuje/openpgpverify-migration.)