From e39eba26f3921402aca0d2600cf65abdd1fe9e55 Mon Sep 17 00:00:00 2001
From: Matt <matt@wayouthere.co.uk>
Date: Wed, 15 Apr 2026 00:24:57 -0700
Subject: [PATCH] read n_ctx back after making llama_context (#21939)

---
 examples/diffusion/diffusion-cli.cpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/examples/diffusion/diffusion-cli.cpp b/examples/diffusion/diffusion-cli.cpp
index e9780407da..403b9b4744 100644
--- a/examples/diffusion/diffusion-cli.cpp
+++ b/examples/diffusion/diffusion-cli.cpp
@@ -602,8 +602,8 @@ int main(int argc, char ** argv) {
 
     int n_input = input_tokens.size();
 
-    if (n_input >= params.n_ctx) {
-        LOG_ERR("error: input too long (%d tokens), max context is %d\n", n_input, params.n_ctx);
+    if (static_cast<uint32_t>(n_input) >= llama_n_ctx(ctx)) {
+        LOG_ERR("error: input too long (%d tokens), max context is %d\n", n_input, llama_n_ctx(ctx));
         llama_free(ctx);
         llama_model_free(model);
         return 1;