chore: update iroh from 0.33.0 to 0.34.0

[link2xt]

No description provided.

[link2xt]

@dignifiedquire @flub What is the transition plan for raw keys? I enable tls_x509 for now, does it mean that it is not compatible with iroh clients that is not using tls_x509 and we have to keep this indefinitely, or can we drop it after a few releases and have a smooth transition?

EDIT: the answer is that tls_x509 is not compatible to tls_raw_public_keys, we just keep using tls_x509 and at some point (e.g. iroh 1.0 breaking compatibility anyway) we can switch to tls_raw_public_keys. DCBACKUP3 scheme will use raw keys.

[Hocuri]

When I try to build this for Android, I get the following error:

$ ./ndk-make.sh
warning: Git tree '/home/user/Development/deltachat-android' is dirty
starting time: Tue Mar 25 12:24:31 CET 2025
Setting CARGO_TARGET environment variables.
Quick debug build that will produce a slower app. DO NOT UPLOAD THE APK ANYWHERE.
/home/user/Development/deltachat-android/jni
/home/user/Development/deltachat-android/jni/deltachat-core-rust
-- cross compiling to aarch64-linux-android (arm64) --
    Finished `dev` profile [optimized] target(s) in 0.32s
-- ndk-build --
[arm64-v8a] SharedLibrary  : libnative-utils.so
ld.lld: error: undefined symbol: preadv
>>> referenced by weak.rs:168 (std/src/sys/pal/unix/weak.rs:168)
>>>               std-993242cf06a7b510.std.5a80935b0be77383-cgu.0.rcgu.o:(_$LT$std..fs..File$u20$as$u20$std..os..unix..fs..FileExt$GT$::read_vectored_at::hab075818bf071fad) in archive jni/arm64-v8a/libdeltachat.a
>>> referenced by weak.rs:168 (std/src/sys/pal/unix/weak.rs:168)
>>>               std-993242cf06a7b510.std.5a80935b0be77383-cgu.0.rcgu.o:(_$LT$std..fs..File$u20$as$u20$std..os..unix..fs..FileExt$GT$::read_vectored_at::hab075818bf071fad) in archive jni/arm64-v8a/libdeltachat.a
>>> referenced by weak.rs:169 (std/src/sys/pal/unix/weak.rs:169)
>>>               std-993242cf06a7b510.std.5a80935b0be77383-cgu.0.rcgu.o:(_$LT$std..fs..File$u20$as$u20$std..os..unix..fs..FileExt$GT$::read_vectored_at::hab075818bf071fad) in archive jni/arm64-v8a/libdeltachat.a
>>> referenced 4 more times
>>> did you mean: readv
>>> defined in: /nix/store/r7v7as0l7a2asrh6m15avrmjsnf72spa-ndk-27-2-12479018-27.2.12479018/toolchains/llvm/prebuilt/linux-x86_64/bin/../sysroot/usr/lib/aarch64-linux-android/21/libc.so

ld.lld: error: undefined symbol: pwritev
>>> referenced by weak.rs:168 (std/src/sys/pal/unix/weak.rs:168)
>>>               std-993242cf06a7b510.std.5a80935b0be77383-cgu.0.rcgu.o:(_$LT$std..fs..File$u20$as$u20$std..os..unix..fs..FileExt$GT$::write_vectored_at::h8f497cf1b2446893) in archive jni/arm64-v8a/libdeltachat.a
>>> referenced by weak.rs:168 (std/src/sys/pal/unix/weak.rs:168)
>>>               std-993242cf06a7b510.std.5a80935b0be77383-cgu.0.rcgu.o:(_$LT$std..fs..File$u20$as$u20$std..os..unix..fs..FileExt$GT$::write_vectored_at::h8f497cf1b2446893) in archive jni/arm64-v8a/libdeltachat.a
>>> referenced by weak.rs:169 (std/src/sys/pal/unix/weak.rs:169)
>>>               std-993242cf06a7b510.std.5a80935b0be77383-cgu.0.rcgu.o:(_$LT$std..fs..File$u20$as$u20$std..os..unix..fs..FileExt$GT$::write_vectored_at::h8f497cf1b2446893) in archive jni/arm64-v8a/libdeltachat.a
>>> referenced 4 more times

ld.lld: error: undefined symbol: process_vm_writev
>>> referenced by nix.2ab134c738ef5a9-cgu.3
>>>               nix-ff443a8fa36385d6.nix.2ab134c738ef5a9-cgu.3.rcgu.o:(nix::sys::uio::process_vm_writev::h212ec9fb4e353a69) in archive jni/arm64-v8a/libdeltachat.a

ld.lld: error: undefined symbol: process_vm_readv
>>> referenced by nix.2ab134c738ef5a9-cgu.3
>>>               nix-ff443a8fa36385d6.nix.2ab134c738ef5a9-cgu.3.rcgu.o:(nix::sys::uio::process_vm_readv::h2396be9a216dccc4) in archive jni/arm64-v8a/libdeltachat.a
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make: *** [/nix/store/r7v7as0l7a2asrh6m15avrmjsnf72spa-ndk-27-2-12479018-27.2.12479018/build/core/build-binary.mk:661: obj/local/arm64-v8a/libnative-utils.so] Error 1

[link2xt]

For me it compiles successfully. I am using NDK that is pinned in the nix flake.

EDIT: scripts/ndk-make.sh --debug arm64-v8a (with debug) fails after this update and succeeds before.

process_vm_readv and process_vm_writev were added in rust-lang/libc#1878 and nix-rust/nix#1557, quite some time ago. They are not available on Android 5, but this has not been causing us problems so far.

The changes in dependencies are replacing backoff with backon and addition of gloo-timers.

nix-ff443a8fa36385d6 is the version nix 0.26.4 according to ./aarch64-linux-android/debug/deps/nix-ff443a8fa36385d6.d
It is used by rtnetlink 0.13.1, all other dependencies use different versions of nix in our Cargo.lock.

So iroh probably started using something from rtnetlink 0.13.1 that was not used before.

netwatch 0.4.0 depends on both rtnetlink 0.13.1 and rtnetlink 0.14.1. It was already the case with previous netwatch and it did not compile, so it is some user of netwatch that results in addition of more code to the binary.

netwatch depends on rtnetlink 0.13.1 on Android because of rust-netlink/rtnetlink#83:
https://github.com/n0-computer/net-tools/blob/5bc90c98121e6462d8b0f98cd65a2140d1fcf332/netwatch/Cargo.toml#L58-L62

Inside of nix 0.26.4 process_vm_readv is not called at all, it is a public API that is not used internally: https://docs.rs/nix/0.26.4/nix/sys/uio/fn.process_vm_readv.html

rtnetlink 0.13.1 also does not call process_vm_readv, so it got compiled in for no good reason.

Maybe the problem is portmapper 0.4.0 that pulled in code from rtnetlink with this change that happened between 0.3.1 and 0.4.0:
n0-computer/net-tools@c82d4e4

---

I bisected iroh to the commit that broke the build: n0-computer/iroh@7acfe39

This commit updated netwatch from 0.3 to 0.4 and portmapper from 0.3 to 0.4.

[link2xt]

In the end I did this.

Cloned iroh-gossip, checked out tag v0.34.0 (96173fd30934a49021a015a97b90fe43d5c88969), patched:

diff --git a/Cargo.toml b/Cargo.toml
index d75a009..10975bf 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -52,7 +52,7 @@ anyhow = { version = "1", optional = true }
 futures-lite = { version = "2.3", optional = true }
 futures-concurrency = { version = "7.6.1", optional = true }
 futures-util = { version = "0.3.30", optional = true }
-iroh = { version = "0.34", default-features = false, optional = true }
+iroh = { path = "../iroh/iroh", default-features = false, optional = true }
 tokio = { version = "1", optional = true, features = ["io-util", "sync"] }
 tokio-util = { version = "0.7.12", optional = true, features = ["codec"] }
 tracing = "0.1"
@@ -69,7 +69,7 @@ clap = { version = "4", features = ["derive"], optional = true }
 [dev-dependencies]
 tokio = { version = "1", features = ["io-util", "sync", "rt", "macros", "net", "fs"] }
 clap = { version = "4", features = ["derive"] }
-iroh = { version = "0.34", default-features = false, features = ["metrics", "test-utils"] }
+iroh = { path = "../iroh/iroh", default-features = false, features = ["metrics", "test-utils"] }
 rand_chacha = "0.3.1"
 testresult = "0.4.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }

Cloned iroh, checkout out v0.34.0, patched:

diff --git a/iroh-net-report/Cargo.toml b/iroh-net-report/Cargo.toml
index 3d690a866..4a310f5d4 100644
--- a/iroh-net-report/Cargo.toml
+++ b/iroh-net-report/Cargo.toml
@@ -41,8 +41,8 @@ url = { version = "2.4" }
 # non-wasm-in-browser dependencies
 [target.'cfg(not(all(target_family = "wasm", target_os = "unknown")))'.dependencies]
 hickory-resolver = "=0.25.0-alpha.5"
-netwatch = { version = "0.4" }
-portmapper = { version = "0.4", default-features = false }
+netwatch = { path = "../../net-tools/netwatch" }
+portmapper = { path = "../../net-tools/portmapper", default-features = false }
 surge-ping = "0.8.0"

 [dev-dependencies]
diff --git a/iroh/Cargo.toml b/iroh/Cargo.toml
index 2564b21fe..ba33728e3 100644
--- a/iroh/Cargo.toml
+++ b/iroh/Cargo.toml
@@ -43,7 +43,7 @@ http = "1"
 iroh-base = { version = "0.34.0", default-features = false, features = ["key", "relay"], path = "../iroh-base" }
 iroh-relay = { version = "0.34", path = "../iroh-relay", default-features = false }
 n0-future = "0.1.2"
-netwatch = { version = "0.4" }
+netwatch = { path = "../../net-tools/netwatch" }
 pin-project = "1"
 pkarr = { version = "2", default-features = false, features = [
     "async",
@@ -103,7 +103,7 @@ parse-size = { version = "=1.0.0", optional = true } # pinned version to avoid b
 hickory-resolver = { version = "=0.25.0-alpha.5" }
 igd-next = { version = "0.15.1", features = ["aio_tokio"] }
 netdev = { version = "0.31.0" }
-portmapper = { version = "0.4", default-features = false }
+portmapper = { path = "../../net-tools/portmapper", default-features = false }
 quinn = { package = "iroh-quinn", version = "0.13.0", default-features = false, features = ["platform-verifier", "runtime-tokio", "rustls-ring"] }
 tokio = { version = "1", features = [
     "io-util",

Cloned net-tools, checked out 5bc90c98121e6462d8b0f98cd65a2140d1fcf332 (tag portmapper-v0.4.1), patched:

diff --git a/netwatch/src/interfaces/linux.rs b/netwatch/src/interfaces/linux.rs
index 4d5c117..65c6d82 100644
--- a/netwatch/src/interfaces/linux.rs
+++ b/netwatch/src/interfaces/linux.rs
@@ -25,8 +25,6 @@ pub enum Error {
     MissingDestinationField,
     #[error("mask field is missing")]
     MissingMaskField,
-    #[error("netlink")]
-    Netlink(#[from] rtnetlink::Error),
 }

 pub async fn default_route() -> Option<DefaultRouteDetails> {

And now it compiles on Android with --debug.
So the problem was indeed
n0-computer/net-tools@c82d4e4

[link2xt]

Made a PR with a fix: n0-computer/net-tools#17

[Hocuri]

I can confirm now that this a) compiles and b) a backup transfer between this and DC Desktop 1.56.0 (core v1.157.3) works fine in both direction.

What does not work is transferring the backup if the laptop is in the phone's Wi-Fi hotspot. This also doesn't work with the the latest Android release (1.56.0, core v1.157.3) & the latest Desktop release, so it's not introduced by the PR here. But this used to work in the past and we recommend it in the FAQ, so we need to investigate why it doesn't work anymore.

Not sure if we should merge this already rather than wait for the fix to be released, you are free to decide.