Exit cleanly after test failures

.github/pull_request_template.md

@@ -0,0 +1,12 @@
+## Summary
+
+<!-- Describe the problem and the resulting behavior. Link related issues. -->
+
+## Check list
+
+- [ ] Add tests
+- [ ] Update `CHANGELOG.md`
+
+## Compatibility
+
+<!-- Describe changes to the public API, behavior, package dependencies, or required Jo/Python/Ruby versions. Note breaking changes and migration steps, or state that there are none. -->

.github/workflows/ci.yml

@@ -31,19 +31,6 @@ jobs:
       - run: jo run test-python
       - run: jo run test-ruby
       - run: python3 scripts/check-output.py
-      - name: Check runner failure exit status
-        run: |
-          for module in test-python-failure test-ruby-failure; do
-            output=$(mktemp)
-            if jo run "$module" >"$output" 2>&1; then
-              cat "$output"
-              echo "$module unexpectedly succeeded"
-              exit 1
-            fi
-            grep -q 'tests failed' "$output"
-            cat "$output"
-            rm "$output"
-          done
       - name: Build release artifacts
         run: |
           jo package core

CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## [Unreleased]
+
+### Fixed
+
+- `Runner.runWithArgs` uses normal process exit with status 1 after reporting
+  failed tests on Python and Ruby, removing the extra runner exception
+  traceback. ([#2](https://github.com/typescope/jtest/issues/2))
+
 ## [0.2.0] - 2026-10-13
 
 ### Breaking changes

README.md

@@ -59,7 +59,7 @@ jo run tests -- "math/adds numbers"
 
 Assertions accumulate failures and retain their source locations. Unexpected
 exceptions fail the current test, and execution continues. `Runner.runWithArgs`
-exits with a non-zero status if any selected test fails. Use `Runner.run` when
+exits with status 1 if any selected test fails. Use `Runner.run` when
 you want to handle its Boolean result yourself.
 
 Filters are test-path prefixes. The root's name is not part of a path. A filter

scripts/check-output.py

@@ -1,4 +1,4 @@
-"""Check observable reporting parity after building the conformance apps."""
+"""Check reporting parity and clean failure exits on both backends."""
 
 import re
 import subprocess
@@ -20,7 +20,28 @@ def report(platform):
     return re.sub(r" in [0-9.]+s", " in <elapsed>s", output)
 
 
+def failure_report(platform):
+    result = subprocess.run(
+        ["jo", "run", f"test-{platform}-failure"], capture_output=True, text=True
+    )
+    output = result.stdout + result.stderr
+    assert result.returncode == 1, (result.returncode, output)
+    assert (
+        "intentional failure: intentional (test-fixtures/failure/Run.jo:"
+        in result.stdout
+    ), output
+    assert "0 passed, 1 failed" in result.stdout, output
+    assert "Traceback" not in output, output
+    assert "SystemExit" not in output, output
+    assert "tests failed" not in output, output
+    assert "src/runner/Runner.jo" not in output, output
+    return re.sub(r" in [0-9.]+s", " in <elapsed>s", result.stdout)
+
+
 if __name__ == "__main__":
     assert Path("jo.toml").exists(), "run from the repository root"
     assert report("python") == report("ruby"), "backend reports differ"
-    print("Python and Ruby reports match")
+    assert failure_report("python") == failure_report("ruby"), (
+        "backend failure reports differ"
+    )
+    print("Python and Ruby reports match; failure exits are clean")

src/python/Host.jo

@@ -45,6 +45,10 @@ private def hostFlush(): Unit =
   py.module("sys").stdout.flush()
   pass
 
+private def hostExit(status: Int): Unit =
+  py.module("sys").exit(status)
+  pass
+
 private def hostTry(body: () => Unit): Result[Bool, String] =
   match py.try(body())
   case Ok(_) => Ok(true)

src/ruby/Host.jo

@@ -42,6 +42,10 @@ private def hostFlush(): Unit =
   rb.const("STDOUT").flush()
   pass
 
+private def hostExit(status: Int): Unit =
+  rb.const("Kernel").exit(status)
+  pass
+
 private def hostTry(body: () => Unit): Result[Bool, String] =
   match rb.try(body())
   case Ok(_) => Ok(true)

src/runner/Runner.jo

@@ -160,12 +160,12 @@ section Runner
 
   //[ Run tests using the conventional first-argument filter.
     !
-    ! Aborts with a non-zero process status when any selected test fails. Use
+    ! Exits with process status 1 when any selected test fails. Use
     ! `run` when the caller needs to handle the Boolean result itself.
   //]
   def runWithArgs(root: Testing.Suite): Unit receives stdout, args =
     if !run(root, filterFromArgs(args)) then
-      abort("tests failed")
+      hostExit(1)
 
   // The root contributes no path segment, so a filter reads as `providers/openai`.
   private def walk(