Arch Linux User Repository

A manual makepkg -si installed fine just now but still fails with yay --devel. I have no idea what changed. @keilmillerjr Can you also try again?

This ran fine on a clean working tree latest pull of the master branch of helix as of right now:

cargo build --locked --profile opt && cargo test --workspace --locked && RUSTFLAGS="--cfg tokio_unstable" cargo integration-test --locked

One notable thing is this line in the output:

test helix-lsp-types/src/lib.rs - MarkupContent (line 2716) ... ignored

Are there any troubleshooting hints anyone can give? Thank you.

I share the same issue as @SalmanFarooq.

@SalmanFarooq, it builds fine for me, both before and after that change. The weird thing is that it seems to be treating something as a doctest that shouldn't be at all, because it's marked ${language}. Doesn't fundamentally seem like a stable vs nightly issue to me, but falling back on the upstream rust-toolchain.toml seemed like a good idea regardless.

It is still failing with the same error even after the update you pushed just now.

The build has been failing for me since some time. I have been able to get it to work with:

diff --git a/PKGBUILD b/PKGBUILD
index f47d06c..bd46893 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -49,7 +49,7 @@ build() {

 check() {
     cd "${_pkgname}"
-    export RUSTUP_TOOLCHAIN=stable
+    export RUSTUP_TOOLCHAIN=nightly
     export CARGO_TARGET_DIR=target
     cargo test --workspace --locked
     export RUSTFLAGS="${RUSTFLAGS} --cfg tokio_unstable"

More info:

> RUSTUP_TOOLCHAIN=stable rustc --version
rustc 1.80.1 (3f5fd8dd4 2024-08-06)

> RUSTUP_TOOLCHAIN=stable cargo --version
cargo 1.80.1 (376290515 2024-07-16)

Error is a test compilation fails:

---- helix-lsp-types/src/lib.rs - MarkedString (line 2571) stdout ----
error: expected expression, found `$`
 --> helix-lsp-types/src/lib.rs:2572:1
  |
3 | ${value}
  | ^ expected expression

error: aborting due to 1 previous error

Took me a while to confirm, as the pkgbuild always just worked for me, but

• Enabling lto (which I believe is only fat lto, no way to choose thin lto) in /etc/makepkg.conf (which is disabled by default) does seem to trigger linker failures
• for tokio_unstable, the same observation, that the tests simply compiled and passed for me. In fact, they still do, even after enabling (and consequently disabling on the PKGBUILD level) lto. I'm not sure where the problem stems from, maybe some difference in compiler version, somehow? Either way, I'll add the flag before the integration test in the hopes it alleviates the issues people are having.
• I can't confirm rustc 1.70 being used, both checking embedded strings in the final binary and the content of ~/.rustup/toolchains gives me the expected current stable
• Adding the license key. That was mostly just an oversight/not being present when I took over the maintenance