From 05a40db4e3bf4f043e5666277828f92b48067439 Mon Sep 17 00:00:00 2001
From: Stephen Toub <stoub@microsoft.com>
Date: Fri, 26 Jun 2026 12:16:01 -0400
Subject: [PATCH] Make abort E2E snapshots tolerate timing variants

Add cassette alternatives for valid abort histories where an in-flight tool result is interrupted and where streaming abort retains only the original user prompt before recovery.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---
 .../should_abort_during_active_streaming.yaml | 10 +++++++
 .../session/should_abort_a_session.yaml       | 28 +++++++++++++++++++
 2 files changed, 38 insertions(+)

diff --git a/test/snapshots/abort/should_abort_during_active_streaming.yaml b/test/snapshots/abort/should_abort_during_active_streaming.yaml
index ea70c0d536..70981ee597 100644
--- a/test/snapshots/abort/should_abort_during_active_streaming.yaml
+++ b/test/snapshots/abort/should_abort_during_active_streaming.yaml
@@ -35,3 +35,13 @@ conversations:
         content: Say 'abort_recovery_ok'.
       - role: assistant
         content: abort_recovery_ok
+  - messages:
+      - role: system
+        content: ${system}
+      - role: user
+        content: Write a very long essay about the history of computing, covering every decade from the 1940s to the 2020s in
+          great detail.
+      - role: user
+        content: Say 'abort_recovery_ok'.
+      - role: assistant
+        content: abort_recovery_ok
diff --git a/test/snapshots/session/should_abort_a_session.yaml b/test/snapshots/session/should_abort_a_session.yaml
index dbbbd32aa7..f1217f7f62 100644
--- a/test/snapshots/session/should_abort_a_session.yaml
+++ b/test/snapshots/session/should_abort_a_session.yaml
@@ -50,3 +50,31 @@ conversations:
         content: What is 2+2?
       - role: assistant
         content: 2 + 2 = 4
+  - messages:
+      - role: system
+        content: ${system}
+      - role: user
+        content: run the shell command 'sleep 100' (note this works on both bash and PowerShell)
+      - role: assistant
+        content: I'll run the sleep command for 100 seconds.
+        tool_calls:
+          - id: toolcall_0
+            type: function
+            function:
+              name: report_intent
+              arguments: '{"intent":"Running sleep command"}'
+          - id: toolcall_1
+            type: function
+            function:
+              name: ${shell}
+              arguments: '{"command":"sleep 100","description":"Run sleep 100 command","mode":"sync","initial_wait":105}'
+      - role: tool
+        tool_call_id: toolcall_0
+        content: The execution of this tool, or a previous tool was interrupted.
+      - role: tool
+        tool_call_id: toolcall_1
+        content: The execution of this tool, or a previous tool was interrupted.
+      - role: user
+        content: What is 2+2?
+      - role: assistant
+        content: 2 + 2 = 4