| Commit message (Collapse) | Author | Age | Files | Lines |
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The job of an SNI callback is typically to swap out the context
associated with an SSL depending on the domain the client is trying to
talk to. Typically, only the callbacks associated with the current
context are used, but this is not the case for the SNI callback.
If SNI is run for a second time on a connection (i.e. in a
renegotiation) and the context was replaced with one that didn't itself
register an SNI callback, the old callback would run but wouldn't be
able to find its state in the context's ex data. To work around this, we
pass the pointer to the callback data directly to the callback to make
sure it's always available. It still lives in ex data to handle the
lifetime management.
Closes #979
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
This is a revival of #684 to see if I can help push it across the finish line!
Closes #580
|
| |\
| |
| | |
Add wrapper for SSL_CTX_set_psk_server_callback
|
| | | |
|
| | | |
|
| | |
| |
| |
| | |
Hopefully it works on CI servers now
|
| | | |
|
| |/
|
|
|
|
|
|
|
| |
The initial session ticket is now sent as part of SSL_accept, so some
tests need to write a single byte through the stream to make sure that
both ends have fully completed to avoid test flakes.
TLSv1.3 cipher suite control has been extracted from the normal cipher
list into a separate method: SslContextBuilder::set_ciphersuites.
|
| | |
|
| |
|
|
| |
Closes #690
|
| |
|
|
| |
Also expose hostname verification on libressl
|
| |
|
|
|
| |
Their implementations of the accessors don't behave expected with no
bounds, so we ignore those bits of the tests.
|
| |
|
|
|
| |
They're flickering, and I'm assuming it's just because that version is
so old.
|
| |
|
|
| |
Closes #911
|
| | |
|
| | |
|
| |
|
|
| |
Closes #852
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
We should keep the version features totally separate for now.
|
| |
|
|
|
| |
Google yells at you when using TLS 1.3 without SNI by sending a bogus
self-signed cert!
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
```
% cargo +stable test --lib ssl::test::test_connect_with_alpn_successful_single_match --features=v102
Finished dev [unoptimized + debuginfo] target(s) in 0.0 secs
Running /Users/nga/devel/left/rust-openssl/target/debug/deps/openssl-a38e12a3527f6932
running 1 test
test ssl::test::test_connect_with_alpn_successful_single_match ... FAILED
failures:
---- ssl::test::test_connect_with_alpn_successful_single_match stdout ----
thread 'ssl::test::test_connect_with_alpn_successful_single_match' panicked at 'server exited: exit code: 1', src/ssl/test.rs:91:24
note: Run with `RUST_BACKTRACE=1` for a backtrace.
failures:
ssl::test::test_connect_with_alpn_successful_single_match
test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 159 filtered out
```
|
| |
|
|
|
|
| |
Also document their specific formats.
Closes #502
|
| |
|
|
|
|
| |
These constants have the same values, but X509_FILETYPE_DEFAULT doesn't
work in the Ssl methods and using the SSL_* names is a bit less
confusing.
|
| | |
|
| |
|
|
|
|
|
|
| |
There was previously a lot of behind the scenes magic. We now bind much
more directly to the relevant functions.
Also remove APN support. That protocol is supersceded by ALPN - let's
see if anyone actually needs to use it.
|
| |
|