v0.11.2

CHANGES:

• Private_key.decode_{pem,der} now has a ~sloppy option to recover from
  bad keys (where e.g. the private exponent d is wrong).
  (#135 by @hannesm, reported by @mattjbray in mirage/mirage-crypto#62)

v0.11.1

CHANGES:

• open variant for errors to make the composable (#133 by @dinosaure,
  review by @hannesm)
• BUGFIX avoid fractional seconds in generalized_time: truncate on serialising,
  validate them to be 0 on deseariasing, as required in RFC 5280 4.1.2.5.2
  (#134 by @hannesm, reported by @ansiwen)

v0.11.0

CHANGES:

• BREAKING Validation.validate_raw_signature results in a
  (unit, signature_error) result and logs (Logs.warn) if a weak (non-SHA2)
  hash algorithm was used. This function is used for verifying signatures
  on certificates, signing requests, and certificate revocation lists.
• The `CAInvalidSelfSignature constructor (Validation.ca_error) and
  `ChainInvalidSignature constructor (Validation.chain_validation_error)
  have been removed.
• BREAKING The polymorphic variant Validation.chain_error is now flat (the
  tags `Chain and `Leaf have been removed)
• BREAKING Adapted return type of CRL.validate and CRL.verify
• The pretty-printer Public_key.pp is now provided
• All implemented by @hannesm in #132 based on private conversation with @cfcs
  (who kindly reviewed the changes)

v0.10.0

CHANGES:

• BREAKING #131 use mirage-crypto and mirage-crypto-pk instead of nocrypto
  raise lower OCaml bound to 4.07.0, test 4.10.0 in CI
• BREAKING #131 Certificate.Host_set is now Host.Set, and has pretty-printers
• BREAKING #131 Authenticator.null was removed (fixes #130, suggested by @emillon)
• BREAKING #131 Authenticator.t now requires (non-optional) ~host and ~time
  arguments which may return an option. This allows to more easily spot the
  cases where None is provided. Also, time is (unit -> Ptime.t option) instead
  of Ptime.t -- thus the timestamp at the time of certificate verification is
  used instead of the time of Authenticator construction. Similar changes in
  Validation for verify_chain, verify_chain_of_trust, trust_key_fingerprint,
  trust_cert_fingerprint.
  fixes #130, suggested by @emillon

v0.9.0

CHANGES:

• BREAKING add a whitelist of hash algorithms used for signatures. The default
  whitelist is the SHA-2 family (without SHA-224), Validation.valid_ca{,s} use
  all algorithms as default
  reported by @emillon in #123, fixed in #128
• BREAKING Certificate.hostnames and Signing_request.hostnames (new) return a
  set of [Wildcard|Strict] * [`host] Domain_name.t (Certificate.Host_set.t)
  reported by @mmaker in #88, fixed in #127
• BREAKING #127 Signing_request.sign returns a result type now, an error is
  returned if the signing request was not properly signed
• BREAKING #127 Validation.{verify_chain_of_trust, trust_key_fingerprint,
  trust_cert_fingerptint} and the type Authenticator.t changed, no longer use
  of a Certificate.host, but instead a [`host] Domain_name.t (previously, it was
  a pair)
• BUGFIX support AlgorithmIdentifier of RSA signature algorithms with parameter
  not present
  reported by @Ulrar in #108, fixed in #129
• BUGFIX #127 preserve a signed signing request (Country in a DN sometimes uses
  a non-utf8 string encoding)
• remove deprecation from Validation.trust_cert_fingerprint and
  Authenticator.server_cert_fingerprint
  requested by @mben-romdhane in #125, fixed in #126
• Certificate.signature_algorithm, CRL.signature_algorithm, and
  Signing_request.signature_algorithm are now provided, returning a
  ([RSA|ECDSA] * Nocrypto.Hash.hash) option
  requested by @psafont in #123, fixed in #128

v0.8.1

CHANGES:

• export Private_key.decode_der and encode_der

v0.8.0

CHANGES:

• export X509.Distinguished_name.common_name : t -> string option, which
  extracts the common name of a distinguished name
• Distinguished_name.t is now a Relative_distinguished_name.t list, a
  Relative_distinguished_name is a Set.S with element type attribute, a variant.
  It used to be an attribute (expressed as GADT) Gmap.t, but this representation
  did not conform to RFC 5280, reported by @paurkedal (#117, fixed by #118)
• Now using Set.find_first_opt, which bumps lower OCaml bound to 4.05.0
• Improved pretty-printing for DNs including RFC 4514 conformance (@paurkedal, #119).
• Extension.pp now outputs extension key and its value (#120)
• rename Distinguished_name.SP constructor (stateOrProvince) to ST, as widely used (#121)
• support Street and UID in Distinguished_name to satisfy RFC 4514 demands (#121)

v0.7.1

CHANGES:

• revert General_name.t (DNS and IP components) to string/Cstruct.t list
  • NameConstraints uses DNSname with strings with leading dots (.example.com),
    which are not valid RFC1034 domain names
  • NameConstraints uses IP with IP/netmask, i.e. using 8 octets for an IPv4
    address
  • X509.Certificate.hostnames still return a Domain_name.Set.t
  • reported by @reynir
• drop ipaddr dependency

0.7.0

CHANGES:

• major restructuring, it is unlikely any pre-0.7.0 users will work with 0.7.0+
• remove sexp de&encoders
• provide pretty-printers for validation errors (and types) instead of to_string functions
• use result type and Rresult instead of custom result types and control monad
• use a GADT map for certificate & csr extensions, distinguished names, general names (avoiding multiple extensions with the same OID, uses the gmap library)
• use domain-name library for hostname validation (instead of custom string matching)
• use ipaddr library for IPs in SubjectAlternativeName extension
• remove Encoding module, provide {en,de}code_{der,pem} in the respective modules (which decoders return (_, [> `Msg of string ]) result, no exceptions raised)
• fix DistributionPoint extension: the CRLissuer is a GeneralName, not a DistinguishedName
• remove Extension.reason_code (Extension.reason was there before, and is now used)
• remove bindings from toplevel, t is now Certificate.t, public_key is now Public_key.t
• use alcotest instead of oUnit

0.6.3

CHANGES:

• provide X509.Encoding.distinguished_name_of_cs -- similar to #87 which provided distinguished_name_to_cs
• provide X509.Encoding.{public_key_of_cstruct,public_key_to_cstruct}, as requested by @dinosaure
• support of cstruct 4.0.0, which split up the sexp de&encoders
• removes result dependency (now requires >= 4.04.2)
• upgrades opam file to version 2.0
• build system is now dune