From 038730d432f2bcb223bc4762295524a29c240189 Mon Sep 17 00:00:00 2001
From: dawn <dawn@tangled.org>
Date: Tue, 15 Sep 2026 05:02:08 +0300
Subject: [PATCH] spindle/engines/nixery: unblock tailStep on context done and
 bound wait

Signed-off-by: dawn <dawn@tangled.org>
---
 spindle/engines/nixery/engine.go | 36 +++++++++++++++++++-------------
 1 file changed, 22 insertions(+), 14 deletions(-)

diff --git a/spindle/engines/nixery/engine.go b/spindle/engines/nixery/engine.go
index 942cab2d..484ce6ba 100644
--- a/spindle/engines/nixery/engine.go
+++ b/spindle/engines/nixery/engine.go
@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"io"
 	"log/slog"
+	"net"
 	"path"
 	"runtime"
 	"sync"
@@ -477,21 +478,24 @@ func (e *Engine) RunStep(ctx context.Context, wid models.WorkflowId, w *models.W
 	case <-tailDone:
 
 	case <-ctx.Done():
-		// cleanup will be handled by DestroyWorkflow, since
-		// Docker doesn't provide an API to kill an exec run
-		// (sure, we could grab the PID and kill it ourselves,
-		// but that's wasted effort)
-		e.l.Warn("step timed out", "step", step.Name())
-
-		<-tailDone
-
-		return engine.ErrTimedOut
+		if errors.Is(ctx.Err(), context.DeadlineExceeded) {
+			e.l.Warn("step timed out", "step", step.Name())
+		}
+		select {
+		case <-tailDone:
+		case <-time.After(2 * time.Second):
+		}
+		if errors.Is(ctx.Err(), context.DeadlineExceeded) {
+			return engine.ErrTimedOut
+		}
+		return ctx.Err()
 	}
 
-	select {
-	case <-ctx.Done():
+	if ctx.Err() != nil {
+		if errors.Is(ctx.Err(), context.DeadlineExceeded) {
+			return engine.ErrTimedOut
+		}
 		return ctx.Err()
-	default:
 	}
 
 	execInspectResp, err := e.docker.ContainerExecInspect(ctx, mkExecResp.ID)
@@ -533,19 +537,23 @@ func (e *Engine) tailStep(ctx context.Context, wfLogger models.WorkflowLogger, e
 		return nil
 	}
 
-	// This actually *starts* the command. Thanks, Docker!
 	logs, err := e.docker.ContainerExecAttach(ctx, execID, container.ExecAttachOptions{})
 	if err != nil {
 		return err
 	}
 	defer logs.Close()
 
+	stop := context.AfterFunc(ctx, func() {
+		logs.Close()
+	})
+	defer stop()
+
 	_, err = stdcopy.StdCopy(
 		wfLogger.DataWriter(stepIdx, "stdout"),
 		wfLogger.DataWriter(stepIdx, "stderr"),
 		logs.Reader,
 	)
-	if err != nil && err != io.EOF && !errors.Is(err, context.DeadlineExceeded) {
+	if err != nil && err != io.EOF && !errors.Is(err, context.DeadlineExceeded) && !errors.Is(err, context.Canceled) && !errors.Is(err, net.ErrClosed) && ctx.Err() == nil {
 		return fmt.Errorf("failed to copy logs: %w", err)
 	}
 
-- 
2.51.2