Change kcl_samples tests to run sequentially to make errors easier to understand

rust/kcl-lib/src/simulation_tests/kcl_samples.rs

@@ -2,13 +2,9 @@
 //!
 //! Use the `KCL_SAMPLES_ONLY=gear` environment variable to run only a subset of
 //! the samples, in this case, all those that start with "gear".
-use std::{
+use std::path::{Path, PathBuf};
-    collections::HashMap,
-    path::{Path, PathBuf},
-};
 
 use fnv::FnvHashSet;
-use tokio::task::JoinSet;
 
 use super::Test;
 
@@ -66,32 +62,10 @@ async fn kcl_test_execute() {
 
     assert!(!tests.is_empty(), "No KCL samples found");
 
-    // Note: This is unordered.
+    for test in &tests {
-    let mut tasks = JoinSet::new();
+        super::execute_test(test, true).await;
-    // Mapping from task ID to test index.
-    let mut id_to_index = HashMap::new();
-    // Spawn a task for each test.
-    for (index, test) in tests.iter().cloned().enumerate() {
-        let handle = tasks.spawn(async move {
-            super::execute_test(&test, true).await;
-        });
-        id_to_index.insert(handle.id(), index);
     }
 
-    // Join all the tasks and collect the failures.  We cannot just join_all
-    // because insta's error messages don't clearly indicate which test failed.
-    let mut failed = vec![None; tests.len()];
-    while let Some(result) = tasks.join_next().await {
-        let Err(err) = result else {
-            continue;
-        };
-        // When there's an error, store the test name and error message.
-        let index = *id_to_index.get(&err.id()).unwrap();
-        failed[index] = Some(format!("{}: {err}", &tests[index].name));
-    }
-    let failed = failed.into_iter().flatten().collect::<Vec<_>>();
-    assert!(failed.is_empty(), "Failed tests: {}", failed.join("\n"));
-
     // Ensure that inputs aren't missing.
     let input_names = FnvHashSet::from_iter(tests.iter().map(|t| t.name.clone()));
     let missing = expected_outputs