This adds a unit test to compare a run against google with the supplied openssl
version vs /usr/bin/openssl .
This would fix#2626.
It looks like there are still points to clarify
* NPN output is different (bug)
* Newer openssl version claims it's ECDH 253 instead of ECDH 256.
* Newer openssl version claims for 130x cipher it's ECDH 253, via sockets it´s ECDH/MLKEM. This seems a bug (@dcooper)
A todo is also restricing the unit test to the one where openssl is being used. E.g. the ROBOT check and more aren't done with openssl. So there's no value checking this here.
In some cases OpenSSL returns an "unsupported" message rather than a "not found" message if X25519 and X448 are not supported. This commit changes the check for whether X5519 and X448 are supported for checking for either response.
This commit adds support for the three code points in draft-kwiatkowski-tls-ecdhe-mlkem and the code point 0x6399 from draft-tls-westerbaan-xyber768d00. The group 0x6399 uses a pre-standard version of Kyber and is considered obsolete.
Currently `compare_server_name_to_cert()` only indicates whether the server's host name matches a wildcard name in the certificate. So, it does not indicate if the certificate includes a wildcard name that does not match the server's host name. As a result, if a certificate includes the names "api.sub.example.tld" and "*.api.sub.example.tld," then a wildcard certificate warning will be issued for host names such as www.api.sub.example.tld, but not for api.sub.example.tld.
This commit changes `compare_server_name_to_cert()` to indicate whether the certificate is a wildcard certificate in addition to providing information about how the certificate matches the server's host name. Functions that use this function's response are then changed to extract the information they need (matching or wildcard) from the return value.
The server mentioned in #2615 has a bug, which results in it sending a handshake_failure alert rather than a successful connection if the signature_algorithms extension lists RSA+MD5 before one of the signature algorithms that it supports.
This commit works around this issue by reversing the order in which it lists the signature algorithms in the signature_algorithms extension, thus (generally) listing stronger options first.
This change should not affect the testing, except that it will result in the order of the supported signature algorithms being reversed in the output, if the server respects the client's preferences.
... to avoid redirects on the client side and to make repo migration better visible.
Also amend 'Status' and 'Contributing' in Readme.md. bluesky added, albeit mastodon
seems more interactive. Clarify twitter account is not in use anymore.
testssl.sh worked as expected.
Under the hood, broken pipes are expected as part of the fast loop exit
strategy that relies as little as possible on timeout detection.
But under the CI, testssl.sh output is garbled by the subshells stderr
outputs, catched for some reason by 'prove -v'.
Simply redirecting the stderr output of the offending command to
/dev/null fixes the problem.
This commit fixes#2599 by not wrapping fileout() messages in a "clientProblem" wrapper if TLS13_ONLY is set. The TLS13_ONLY flag being set is an indicator that fileout_banner() has already been called.
- Recover the "not vulnerable" case (no mitigation) printing, cosmetic
fix.
- With the removing of all s_client invocation other than the main loop
one, fix the init of the ERRFILE and TMPFILE: no need to append, no
need to remove, inconditionally zap the content before the loop.
In the wait loop, I was relying on a 1s sleep to eliminate a possible
late zero return value server close on the last attempt.
- do globaly one more harmless "for" iteration
and remove the sleep 1 for faster and more robust result
- correct the non HTTP case iteration value
- adjust the timeout to the conservative 6s in the non HTTP case,
for HTTP case it become 33s
- improve comments
All cases could be handled by the single openssl s_client invocation
loop:
- dispatch and adjust comments to not loose them
- remove the first s_client invocation: stuck connections are allready
handled by the main loop
- remove the second s_client invocation: normal case and server closed
connections are allready handled by the main loop. The loop take care
of the race between server connection close and s_client terminating
too by doing another loop run, not closing STDIN.
- special non HTTP case equivalent to ssl_reneg_attempts=2
- specialcase only the HTTP result printing to not change the output
- openssl-timeout option clashe badly with the main loop logic:
Introduce $OPENSSL_NOTIMEOUT
This commit adds support for the two cipher suites in RFC 9150, TLS_SHA256_SHA256 and TLS_SHA384_SHA384. These are authentication and integrity-only cipher suites.
LibreSSL 4.0.0 was recently released. This commit modified the version check in determine_trust() so that there isn't an incorrect warning suggesting that LibreSSL 4.0.0 "<= 1.0.2 might be too unreliable to determine trust."
This commit modifies check_revocation_ocsp() to check the revocation status of a certificate in a stapled OCSP response whether the response uses SHA-1 or SHA-256 in CertID.
