From db8bc40731d2dde3eab96292b463a7008355c1f5 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Sat, 4 Apr 2026 23:16:26 +0800
Subject: [PATCH] add some warnings if shifting fails

---
 gpttype_adapter.cpp | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
index 35620c3af..558dc005f 100644
--- a/gpttype_adapter.cpp
+++ b/gpttype_adapter.cpp
@@ -2030,11 +2030,12 @@ bool DoContextShifting(llama_context * ctx, llama_context * draft_ctx, std::vect
         int found = ArrFindIndexOf(current_context_tokens,shared);
         if(found>=0 && found > trimstart)
         {
+            bool ok = true;
             if(!dryrun)
             {
                 //extract the unwanted tokens out from context and KV
                 int diff = found - trimstart;
-                llama_memory_seq_rm(llama_get_memory(ctx), 0, trimstart, trimstart + diff);
+                ok = llama_memory_seq_rm(llama_get_memory(ctx), 0, trimstart, trimstart + diff);
                 llama_memory_seq_add(llama_get_memory(ctx), 0, trimstart + diff, -1, -diff);
                 if(draft_ctx)
                 {
@@ -2045,7 +2046,14 @@ bool DoContextShifting(llama_context * ctx, llama_context * draft_ctx, std::vect
                 {
                     current_context_tokens[i - diff] = current_context_tokens[i];
                 }
-                printf("\n[Context Shifting: Erased %d tokens at position %d]", diff, trimstart + 1);
+                if(ok)
+                {
+                    printf("\n[Context Shifting: Erased %d tokens at position %d]", diff, trimstart + 1);
+                }
+                else
+                {
+                    printf("\n[Warning: Context Shifting FAILED to erase %d tokens at position %d]", diff, trimstart + 1);
+                }
                 current_context_tokens.resize(current_context_tokens.size() - diff);
             }
             return true;