From 18d83b4da9564aed796ab2b734f5cd1c3ded760b Mon Sep 17 00:00:00 2001
From: Teknium <127238744+teknium1@users.noreply.github.com>
Date: Wed, 22 Jul 2026 21:07:23 -0700
Subject: [PATCH] test(agent): pin the #61932 all-oversized-tail dead-end shape
 as compressible

Regression test for the exact issue #61932 report: head + an 8-message
protected tail made exclusively of oversized tool pairs.  Pre-fix,
compress_start >= compress_end made compress() a pure no-op and the
retry loop ended in 'Cannot compress further'; post-fix the Phase-1
pressure demotion reclaims the tail in one pass while preserving
tool_call/tool_result pairing.
---
 .../test_protected_tail_pressure_61932.py     | 58 +++++++++++++++++++
 1 file changed, 58 insertions(+)

diff --git a/tests/agent/test_protected_tail_pressure_61932.py b/tests/agent/test_protected_tail_pressure_61932.py
index 59e5bf4c2762..ce7222e2326c 100644
--- a/tests/agent/test_protected_tail_pressure_61932.py
+++ b/tests/agent/test_protected_tail_pressure_61932.py
@@ -210,6 +210,64 @@ class TestProtectedTailPressure61932:
         # progress (never a pure no-op dead-end above the window).
         assert tok < c.threshold_tokens or last_progress
 
+    def test_all_oversized_tail_dead_end_shape_now_compresses(
+        self, compressor_128k
+    ):
+        """Exact #61932 dead-end: the protected tail ALONE holds everything.
+
+        Head (3 messages) + an 8-message tail of exclusively oversized tool
+        pairs.  The tail token budget + the ``_MAX_TAIL_MESSAGE_FLOOR`` (8)
+        floor protect every non-head message, so ``compress_start >=
+        compress_end`` — pre-fix ``compress()`` returned the transcript
+        UNCHANGED, incremented ``_ineffective_compression_count``, and the
+        retry loop died with "Cannot compress further".  Post-fix the Phase-1
+        pressure pass demotes the oversized tool bodies even though the
+        summary window is empty, so the same call materially shrinks the
+        transcript below the context window.
+        """
+        c = compressor_128k
+        msgs: list[dict] = [
+            {"role": "system", "content": "You are Hermes."},
+            {"role": "user", "content": "Investigate thoroughly"},
+            {"role": "assistant", "content": "OK"},
+        ]
+        for i in range(4):
+            msgs.extend(_unique_tool_pair(i, 200_000))
+        assert len(msgs) == 11  # 3 head + 8-message all-oversized tail
+
+        before = estimate_messages_tokens_rough(msgs)
+        assert before > c.context_length, "fixture must start over-context"
+
+        out = c.compress(list(msgs), current_tokens=before)
+        after = estimate_messages_tokens_rough(out)
+
+        # The dead-end is broken: one pass reclaims the bulk of the tail.
+        assert after < c.context_length, (
+            f"still over context: {after:,} >= {c.context_length:,}"
+        )
+        assert after < before * 0.25, (
+            f"expected the oversized tail to demote: {before:,} → {after:,}"
+        )
+
+        # tool_call/tool_result pairing must survive demotion — never orphan
+        # a tool result or a tool call (provider 400s otherwise).  Whole
+        # pairs may legitimately be summarized away together.
+        call_ids = {
+            tc["id"]
+            for m in out
+            if m.get("role") == "assistant"
+            for tc in (m.get("tool_calls") or [])
+            if isinstance(tc, dict)
+        }
+        tool_result_ids = [
+            m.get("tool_call_id") for m in out if m.get("role") == "tool"
+        ]
+        assert tool_result_ids, "expected surviving tool pairs in the tail"
+        for rid in tool_result_ids:
+            assert rid in call_ids, f"orphaned tool result {rid!r}"
+        for cid in call_ids:
+            assert cid in tool_result_ids, f"orphaned tool call {cid!r}"
+
     def test_light_tail_still_keeps_recent_tool_bodies(self, compressor_128k):
         """Pressure demotion must not fire on a normal-sized protected tail."""
         c = compressor_128k