`test_workspace_config` test fails

[Kladki]

Steps to reproduce the bug with the above code

Version being used: 0.45.0
The test setup can be seen here (Just running ./scripts/regenerate_test_rustdocs.sh, followed by cargo test --frozen)

Actual Behaviour

When trying to package cargo-semver-checks for Alpine Linux, the aforementioned test failed on all architectures, with the following stdout:

running 7 tests
test test_workspace_config ... FAILED
test test_workspace_key_both_missing ... ok
test test_only_read_config_from_new_manifest ... ok
test test_workspace_key_metadata_true ... ok
test test_workspace_key_cargo_true ... ok
test test_only_warnings ... ok
test test_allow_skipped ... ok
failures:
---- test_workspace_config stdout ----
thread 'test_workspace_config' (28677) panicked at /rustc/ed61e7d7e242494fb7057f2657300d9e77bb4fcb/library/core/src/ops/function.rs:250:5:
Unexpected stderr, failed var.is_match(FAIL(.*)minor(.*)function_missing)
├── var: error: failed to retrieve local crate data
│   
│   Caused by:
│       0: possibly due to errors: [
│            failed to parse old/Cargo.toml: no `package` table,
│          ]
│       1: package `pkg` not found in old
└── var as str: error: failed to retrieve local crate data
    
    Caused by:
        0: possibly due to errors: [
             failed to parse old/Cargo.toml: no `package` table,
           ]
        1: package `pkg` not found in old
command=`cd "test_crates/manifest_tests/workspace_overrides" && "/builds/Matthias/aports/testing/cargo-semver-checks/src/cargo-semver-checks-0.45.0/target/debug/cargo-semver-checks" "semver-checks" "--baseline-root" "old" "--manifest-path" "new/pkg" "-v"`
code=1
stdout=""
stderr=```
error: failed to retrieve local crate data
Caused by:
    0: possibly due to errors: [
         failed to parse old/Cargo.toml: no `package` table,
       ]
    1: package `pkg` not found in old

note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
failures:
test_workspace_config
test result: FAILED. 6 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.04s
error: test failed, to rerun pass --test lint_config

### Expected Behaviour

For the test to pass.

### Generated System Information

#### Software version

cargo-semver-checks 0.45.0

#### Operating system

- OS: Linux (Alpine Linux 3.21.0_alpha20240923)
- Kernel: 6.6.63-0-lts

#### Command-line

```bash
/builds/Matthias/aports/testing/cargo-semver-checks/src/cargo-semver-checks-0.45.0/target/release/cargo-semver-checks semver-checks --bugreport 

cargo version

> cargo -V 
cargo 1.91.1 (ea2d97820 2025-10-10) (Alpine Linux Rust 1.91.1-r0)

Compile time information

• Profile: release
• Target triple: x86_64-alpine-linux-musl
• Family: unix
• OS: linux
• Architecture: x86_64
• Pointer width: 64
• Endian: little
• CPU features: fxsr,sse,sse2
• Host: x86_64-alpine-linux-musl

Build Configuration

[build]
jobs = 48
rustflags = ["-Clink-arg=-Wl,-z,pack-relative-relocs"]

Additional Context

No response

[obi1kenobi]

Thanks for reporting this! It's quite a strange issue: I'm not able to reproduce it using Rust 1.91.1 on my Linux machine.

Let me try to suggest a few things you could try on your end to help us narrow down on what's causing this.

Try removing your build configuration settings

I doubt this is the cause, but it'll be very good to rule it out quickly. Build configuration has caused issues in the past, which is why it's in the issue reporting form now :)

Try deleting the test_crates directory, then checking it out from git again

Sometimes stray files in the test crates directory can change test outcomes. For example, a .cargo/config.toml file somewhere in that directory (or above it) could definitely cause some havoc.

Try applying this patch and see if the test outcome changes

diff --git a/tests/lint_config.rs b/tests/lint_config.rs
index d2684a7..96ee3e3 100644
--- a/tests/lint_config.rs
+++ b/tests/lint_config.rs
@@ -23,7 +23,7 @@ fn command_for_workspace(workspace: &str, package: &str) -> Command {
         .args([
             "semver-checks",
             "--baseline-root",
-            "old",
+            &format!("old/{package}"),
             "--manifest-path",
             &format!("new/{package}"),
             "-v", // needed to show pass/fail of individual lints

In principle cargo should have been able to locate the package and workspace in either case, so this should do nothing. If it changes the outcome, we'll know where to look more closely.

I'm happy to help as best I can, but I can mostly just give suggestions and advice until I can reproduce the issue on my end. I'd appreciate your help in getting this to reproduce reliably!

[Ashutosh0x]

Hi @obi1kenobi, I tried reproducing this on Windows with v0.45.0. I'll apply the patch from the comment above and report back with results. If I can get a reliable reproduction, I'd be happy to put together a fix.

[Ashutosh0x]

Hey! I just opened PR #1611 adding a union_field_marked_deprecated lint — it mirrors the existing struct_field_marked_deprecated pattern for union fields. I initially tried union_marked_deprecated (PR #1610) but realized it duplicated ype_marked_deprecated which already covers unions via ImplOwner. Happy to take on more lints if this one looks good!