Merge remote-tracking branch 'upstream/main' into patch-v0.64.1

.github/workflows/slow_tests.yml

@@ -5,6 +5,7 @@ env:
   ANDROID_NDK_VERSION: 27.0.12077973
   LIBSIGNAL_TESTING_ENCLAVE_SECRET: ${{secrets.SVR_ENCLAVE_SECRET}}
   LIBSIGNAL_TESTING_CDSI_ENCLAVE_SECRET: ${{secrets.CDSI_ENCLAVE_SECRET}}
+  LIBSIGNAL_TESTING_RUN_NONHERMETIC_TESTS: true
   RUST_LOG: debug
 
 on:

Cargo.toml

@@ -75,9 +75,9 @@ signal-neon-futures = { path = "rust/bridge/node/futures" }
 # Our forks of some dependencies, accessible as xxx_signal so that usages of them are obvious in source code. Crates
 # that want to use the real things can depend on those directly.
 
-boring-signal = { git = "https://github.com/signalapp/boring", tag = "signal-v4.9.0b", package = "boring", default-features = false }
+boring-signal = { git = "https://github.com/signalapp/boring", tag = "signal-v4.13.0", package = "boring", default-features = false }
 curve25519-dalek-signal = { git = 'https://github.com/signalapp/curve25519-dalek', package = "curve25519-dalek", tag = 'signal-curve25519-4.1.3' }
-tokio-boring-signal = { git = "https://github.com/signalapp/boring", package = "tokio-boring", tag = "signal-v4.9.0b" }
+tokio-boring-signal = { git = "https://github.com/signalapp/boring", package = "tokio-boring", tag = "signal-v4.13.0" }
 
 aes = "0.8.3"
 aes-gcm-siv = "0.11.1"
@@ -147,7 +147,7 @@ sha2 = "0.10"
 snow = { version = "0.9.6", default-features = false, features = ["hfs"] }
 static_assertions = "1.1"
 strum = "0.26"
-subtle = "2.5"
+subtle = "2.6"
 syn = "2.0"
 syn-mid = "0.6"
 test-case = "3.3"
@@ -156,13 +156,14 @@ thiserror = "1.0.57"
 tokio = "1"
 tokio-stream = "0.1.14"
 uuid = "1.1.2"
+visibility = "0.1.1"
 x25519-dalek = "2.0.0"
 zerocopy = "0.7.34"
 
 [patch.crates-io]
 # When building libsignal, just use our forks so we don't end up with two different versions of the libraries.
 
-boring = { git = 'https://github.com/signalapp/boring', tag = 'signal-v4.9.0b' }
+boring = { git = 'https://github.com/signalapp/boring', tag = 'signal-v4.13.0' }
 curve25519-dalek = { git = 'https://github.com/signalapp/curve25519-dalek', tag = 'signal-curve25519-4.1.3' }
 
 [profile.dev.package.argon2]

LibSignalClient.podspec

@@ -5,7 +5,7 @@
 
 Pod::Spec.new do |s|
   s.name             = 'LibSignalClient'
-  s.version          = '0.62.0'
+  s.version          = '0.64.1'
   s.summary          = 'A Swift wrapper library for communicating with the Signal messaging service.'
 
   s.homepage         = 'https://github.com/signalapp/libsignal'

TESTING.md

@@ -17,6 +17,15 @@ For the most part, libsignal is tested using each language's usual testing infra
 However, sometimes there are some more interesting test configurations; those are documented here.
 
 
+# Rust Benchmarks
+
+- If you are testing on an ARM64 device (including Desktop), you should compile with `RUSTFLAGS="--cfg aes_armv8"` to enable hardware support in the `aes` crate.
+
+- Similarly, although most tests are not very sensitive to the speed of SHA-2, you should also compile with `--features sha2/asm`. (`libsignal-message-backup` turns this on by default as a dev-dependency.) This will go away when we get to update to sha2 0.11.
+
+All of these configuration options are normally set either at the bridge crate level or in the build scripts for each bridged platform, but they may not be set when running with plain `cargo bench`.
+
+
 # Running cross-compiling Rust tests with custom runners
 
 Rust allows running tests with cross-compiled targets, but normally that only works if your system supports executing the cross-compiled binary (like Intel targets on ARM64 macOS or Windows, or 32-bit targets on 64-bit Linux or Windows). However, by overriding the "runner" setting for a particular target, we can run cross-compiled tests as well.