From 7fafaaa2d1c1556d0d769d574a86758aad9052ef Mon Sep 17 00:00:00 2001
From: Beforerr <zzj956959688@gmail.com>
Date: Thu, 30 Apr 2026 13:28:17 -0700
Subject: [PATCH] feat: make Test.jl exception handling preserves native output
 format

---
 go/client_test.go            | 38 ++++++++++++++++++++++++++++++++++++
 go/daemon.go                 |  7 +++++++
 go/julia_client_runtime.jl   | 24 +++++++++++++++++++++--
 skills/julia-client/SKILL.md |  2 +-
 4 files changed, 68 insertions(+), 3 deletions(-)

diff --git a/go/client_test.go b/go/client_test.go
index d6d18e1..005c2fa 100644
--- a/go/client_test.go
+++ b/go/client_test.go
@@ -273,6 +273,44 @@ func TestEvalErrorTraceSaved(t *testing.T) {
 	require.Contains(t, resp.Error, "Stacktrace:")
 }
 
+func TestEvalTestFailureKeepsNativeNoiseLevel(t *testing.T) {
+	socketPath, stop, _ := startTestDaemon(t)
+	defer stop()
+
+	cwd, err := os.Getwd()
+	require.NoError(t, err)
+
+	resp := sendRequest(t, socketPath, protocolRequest{
+		Action: "eval",
+		Code:   `using Test; @testset "x" begin @test false end`,
+		Cwd:    cwd,
+	})
+	require.Contains(t, resp.Output, "x: Test Failed")
+	require.Contains(t, resp.Output, "Test Summary:")
+	require.Equal(t, "ERROR: Some tests did not pass: 0 passed, 1 failed, 0 errored, 0 broken.", resp.Error)
+	require.NotContains(t, resp.Error, "Stacktrace:")
+	require.NotContains(t, resp.Error, "Trace saved")
+	require.NotContains(t, resp.Error, "TestSetException")
+
+	trace := sendRequest(t, socketPath, protocolRequest{
+		Action:     "trace",
+		Cwd:        cwd,
+		TraceLevel: "full",
+	})
+	require.Empty(t, trace.Error)
+	require.Contains(t, trace.Output, "TestSetException")
+
+	resp = sendRequest(t, socketPath, protocolRequest{
+		Action: "eval",
+		Code:   `using Test; @test false`,
+		Cwd:    cwd,
+	})
+	require.Contains(t, resp.Output, "Test Failed")
+	require.Equal(t, "ERROR: There was an error during testing", resp.Error)
+	require.NotContains(t, resp.Error, "Stacktrace:")
+	require.NotContains(t, resp.Error, "FallbackTestSetException")
+}
+
 func TestRevisePicksUpPackageChanges(t *testing.T) {
 	socketPath, stop, _ := startTestDaemon(t)
 	defer stop()
diff --git a/go/daemon.go b/go/daemon.go
index 728bf51..a8e02b6 100644
--- a/go/daemon.go
+++ b/go/daemon.go
@@ -123,12 +123,19 @@ func normalizedTraceLevel(level string) string {
 }
 
 func formatJuliaError(err *juliaEvalError, level string) string {
+	traceHint := strings.TrimSpace(err.smart) != strings.TrimSpace(err.short)
 	switch normalizedTraceLevel(level) {
 	case "short":
+		if !traceHint {
+			return err.short
+		}
 		return err.short + "\n\nTrace saved: run `julia-client trace --trace [smart|full]` to inspect"
 	case "full":
 		return err.full
 	default:
+		if !traceHint {
+			return err.short
+		}
 		return err.smart + "Trace saved: run `julia-client trace` to inspect"
 	}
 }
diff --git a/go/julia_client_runtime.jl b/go/julia_client_runtime.jl
index 3fbced9..0f1b41c 100644
--- a/go/julia_client_runtime.jl
+++ b/go/julia_client_runtime.jl
@@ -138,18 +138,38 @@ function _display_error(err)
     return err
 end
 
+function _is_test_summary_exception(err)
+    err = _display_error(err)
+    name = nameof(typeof(err))
+    mod = parentmodule(typeof(err))
+    return string(nameof(mod)) == "Test" && string(name) in ("TestSetException", "FallbackTestSetException")
+end
+
 function _render_error(err, bt)
     frames = stacktrace(bt)
     cut = findfirst(frame -> frame.func === :include_string || String(frame.file) == "none", frames)
     cut === nothing || (frames = frames[1:max(cut - 2, 1)])
     display_err = _display_error(err)
-    short = "ERROR: " * sprint(showerror, display_err)
-    smart = short * "\n" * _render_selected(frames; include_omitted=true)
+    test_exception = _is_test_summary_exception(display_err)
+    short_body = if test_exception
+        sprint() do io
+            Base.invokelatest(showerror, io, display_err, bt)
+        end
+    else
+        sprint(showerror, display_err)
+    end
+    short = "ERROR: " * short_body
+    if test_exception
+        smart = short
+    else
+        smart =  short * "\n" * _render_selected(frames; include_omitted=true)
+    end
     full = sprint(showerror, err, bt)
     return short, smart, full
 end
 
 function _write_error(start_marker, end_marker, short, smart, full)
+    flush(stderr)
     write(stdout, "\n")
     println(stdout, start_marker)
     println(stdout, bytes2hex(Vector{UInt8}(codeunits(short))))
diff --git a/skills/julia-client/SKILL.md b/skills/julia-client/SKILL.md
index 3d4d94b..10cb331 100644
--- a/skills/julia-client/SKILL.md
+++ b/skills/julia-client/SKILL.md
@@ -6,7 +6,7 @@ description: "Run Julia code with session state persistence, project env auto-de
 ## Running code
 
 ```bash
-julia-client -e 'x=1' # Evaluate
+julia-client -e 'x=setupOnce()' # Evaluate
 julia-client -E 'x' # Evaluate and display
 
 # Long-running tasks (pkg install, compile, plot, heavy compute): set longer timeout or disable timeout (0)