Bump version to 0.104.0-alpha.6

Correct selection of candidate CRLs by Distribution Point and Issuing Distribution Point. If a certificate had more than one distributionPoint, then only the first distributionPoint would be considered against each CRL's IssuingDistributionPoint distributionPoint, and then the certificate's subsequent distributionPoints would be ignored.

The impact was that correct provided CRLs would not be consulted to check revocation. With UnknownStatusPolicy::Deny (the default) this would lead to incorrect but safe Error::UnknownRevocationStatus. With UnknownStatusPolicy::Allow this would lead to inappropriate acceptance of revoked certificates.

This vulnerability is thought to be of limited impact. This is because both the certificate and CRL are signed -- an attacker would need to compromise a trusted issuing authority to trigger this bug. An attacker with such capabilities could likely bypass revocation checking through other more impactful means (such as publishing a valid, empty CRL.)

More likely, this bug would be latent in normal use, and an attacker could leverage faulty revocation checking to continue using a revoked credential.

This vulnerability is identified by GHSA-pwjx-qhcg-rvj4. Thank you to @1seal for the report.

What's Changed

• Take semver-compatible dependency updates by @djc in #448
• fix rust 1.94 ambiguous panic macro warnings by @cpu in #449
• avoid std::prelude imports by @cpu in #450
• Take semver-compatible dependency updates by @djc in #451
• Tweak SECURITY.md to provide guidance to sloperators by @djc in #454
• Fix formatting by @ctz in #456
• Take semver-compatible updates by @ctz in #459
• Prepare 0.104.0-alpha.5 by @ctz in #457

Full Changelog: v/0.104.0-alpha.4...v/0.104.0-alpha.5

What's Changed

• tests: port client auth revocation tests to Rust by @djc in #442
• Drop dependency on crypto crates by @djc in #444
• Implement StdError for sct::Error by @djc in #446

Full Changelog: v/0.104.0-alpha.3...v/0.104.0-alpha.4

0.104.0-alpha.3 Pre-release

Pre-release

Compare

Choose a tag to compare

Fixes docs.rs build errors.

What's Changed

• Cargo: avoid all-features in docs.rs by @cpu in #530

Full Changelog: https://github.com/rustls/webpki/compare/v/0.103.14..v/0.103.15

What's Changed

• [0.103] Support stable ML-DSA by @djc in #523
• [0.103] Mark ML-DSA algorithms as FIPS submitted by @djc in #525

Full Changelog: v/0.103.13...v/0.103.14

• Fix reachable panic in parsing a CRL. This was reported to us as GHSA-82j2-j2ch-gfr8. Users who don't use CRLs are not affected.
• For name constraints on URI names, we incorrectly processed excluded subtrees in a way which inverted the desired meaning. See #471. This was a case missing in the fix for GHSA-965h-392x-2mh5.

What's Changed

• Actually fail closed for URI matching against excluded subtrees by @djc in #473
• Prepare 0.103.13 by @ctz in #474

Full Changelog: v/0.103.12...v/0.103.13

This release fixes two bugs in name constraint enforcement:

• GHSA-965h-392x-2mh5: name constraints for URI names were ignored and therefore accepted. URI name constraints are now rejected unconditionally. Note this library does not provide an API for asserting URI names, and URI name constraints are otherwise not implemented.
• GHSA-xgp8-3hg3-c2mh: permitted subtree name constraints for DNS names were accepted for certificates asserting a wildcard name. This was incorrect because, given a name constraint of accept.example.com, *.example.com could feasibly allow a name of reject.example.com which is outside the constraint. This is very similar to CVE-2025-61727.

Since name constraints are restrictions on otherwise properly-issued certificates, these bugs are reachable only after signature verification and require misissuance to exploit.

What's Changed

• Prepare 0.103.12 by @djc in #470

Full Changelog: v/0.103.11...v/0.103.12

In response to #464, we've slightly relaxed requirements for anchor_from_trust_cert() to ignore unknown extensions even if they're marked as critical. This only affects parsing a TrustAnchor from DER, for which most extensions are ignored anyway.

What's Changed

• Backport parsing trust anchors with unknown critical extensions to 0.103 by @djc in #466

Correct selection of candidate CRLs by Distribution Point and Issuing Distribution Point. If a certificate had more than one distributionPoint, then only the first distributionPoint would be considered against each CRL's IssuingDistributionPoint distributionPoint, and then the certificate's subsequent distributionPoints would be ignored.

The impact was that correctly provided CRLs would not be consulted to check revocation. With UnknownStatusPolicy::Deny (the default) this would lead to incorrect but safe Error::UnknownRevocationStatus. With UnknownStatusPolicy::Allow this would lead to inappropriate acceptance of revoked certificates.

This vulnerability is thought to be of limited impact. This is because both the certificate and CRL are signed -- an attacker would need to compromise a trusted issuing authority to trigger this bug. An attacker with such capabilities could likely bypass revocation checking through other more impactful means (such as publishing a valid, empty CRL.)

More likely, this bug would be latent in normal use, and an attacker could leverage faulty revocation checking to continue using a revoked credential.

This vulnerability is identified by GHSA-pwjx-qhcg-rvj4. Thank you to @1seal for the report.

What's Changed

• Freshen up rel-0.103 by @ctz in #455
• Prepare 0.103.10 by @ctz in #458

Full Changelog: v/0.103.9...v/0.103.10