mirror of https://gitlab.torproject.org/tpo/core/arti http://git.dilluti0n.com/mirrors/arti.git/atom?h=arti-v1.4.3 2025-04-09T16:25:51Z 2025-04-09T16:25:51Z Nick Mathewson nickm@torproject.org 2025-04-09T16:25:51Z urn:sha1:f79bb13bbd55957a8e67725455cc8460f5064ee0 This comes up on OSX; I hadn't seen it before, so I assume it is new with Rust 1.86. 2025-03-24T13:58:54Z Ian Jackson ijackson@chiark.greenend.org.uk 2025-03-06T15:52:03Z urn:sha1:73157920ffa84d6f1580170a0e40e5fbe3106a67 We change `NoUnixAddressSupport` to `NoAfUnixSocketSupport` because it doesn't make much sense to talk about support for the addresses separately from support for the sockets. 2025-03-24T13:58:54Z Ian Jackson ijackson@chiark.greenend.org.uk 2025-03-06T15:16:03Z urn:sha1:a60165c11c66c579d628a670de56bfcff494c1cd 2025-03-05T13:50:08Z Nick Mathewson nickm@torproject.org 2025-03-05T13:50:08Z urn:sha1:f880d8c4bea2cd82aba325b82bb68878bb769e97 Switch from x509-signature to rustls-webpki when using rustls. Closes #1824 and #1854 See merge request tpo/core/arti!2816 2025-03-04T19:45:18Z Nick Mathewson nickm@torproject.org 2025-03-03T15:17:33Z urn:sha1:d206e9b70ddfce0083546d4f1836129985b9abfc We do this so that we can make sure there's a provider installed when we run the tests. 2025-03-04T19:45:18Z Nick Mathewson nickm@torproject.org 2025-02-26T22:02:48Z urn:sha1:ede6d75567c1ee81812478e6137032dd568ddb91 The x509-signature crate is archived, and won't see any more releases. Using it is tying us to ring 0.16 internally, which means we depend on two ring versions. Fortunately, rustls-webpki relaxes some of the earlier restrictions from the vanilla webpki crate, which means that its certificate parser now accepts C tor's oddball x509 certificates as valid. With this change, we can delegate to rustls's built-in signature-checking code, and we only have to override its certificate validation. (We still override it with a pile of comments about how we don't validate link certificates much.) I've had to include a few certificates: two are for tests, but one is needed as a placeholder, since we can't construct a rustls certificate validator without a root cert, even if we'll never use it. Closes #1824. Closes #1854. 2025-03-04T19:45:18Z Nick Mathewson nickm@torproject.org 2025-02-26T21:20:03Z urn:sha1:170edfe9aa912ea9f1977475d2a2a8904c62224a (We had added this rename when rustls renamed it originally.) 2025-03-04T17:33:25Z Ian Jackson ijackson@chiark.greenend.org.uk 2025-03-04T15:39:48Z urn:sha1:11049b07c419fc7544c382a98ea8366d6b468daf Let's use Tokio terminology here. 2025-03-04T17:30:43Z Ian Jackson ijackson@chiark.greenend.org.uk 2025-02-26T16:03:32Z urn:sha1:9a491ce278bb581405aecd5dbef36d0f4aef1198 This was referenced and explained from the docs, but didn't exist yet. Here it is. Everyone except the Tokio glue, and the CompoundRuntime, just use the default implementation in terms of spawn_thread. spawn_thread has a more relaxed contract, so this is correct. 2025-03-04T17:30:43Z Ian Jackson ijackson@chiark.greenend.org.uk 2025-02-26T16:03:43Z urn:sha1:878c32bc1e9027fa3cf405c87f5323b94969b0fb Forbid re-entering the executor using ToplevelBlockOn::block_on. This was always forbidden in the case of MockExecutor, but that meant that tests using MockExecutor would malfunction if the code under test needed to re-enter the executor from sync code (since the code under test would have to use block_on, which wrong). See #1835. Provide a function which *can* do this, reenter_block_on. The MockExecutor needs to know the difference, and other runtimes may too. They are conceptually quite different operations. Introduce ToplevelRuntime as a convenience alias.