Bug: llama-cli does not show the results of the performance test when SIGINT

[ownia]

What happened?

When running llama-cli in conversation mode, press Ctrl+C to interject it did not print the results of the performance info.

git bisect show it's 6262d13.

e6deac3

>
llama_perf_sampler_print:    sampling time =      17.31 ms /   124 runs   (    0.14 ms per token,  7164.32 tokens per second)
llama_perf_context_print:        load time =    2548.25 ms
llama_perf_context_print: prompt eval time =    4104.68 ms /    25 tokens (  164.19 ms per token,     6.09 tokens per second)
llama_perf_context_print:        eval time =    6035.09 ms /   109 runs   (   55.37 ms per token,    18.06 tokens per second)
llama_perf_context_print:       total time =   36065.20 ms /   134 tokens
localhost:~/code/kleidiai/llama.cpp #

6262d13: (empty output after Ctrl+C)

> localhost:~/code/kleidiai/llama.cpp #

Name and Version

localhost:~/code/kleidiai/llama.cpp # ./llama-cli --version
version: 3787 (6026da5)
built with cc (SUSE Linux) 14.2.0 for aarch64-suse-linux

What operating system are you seeing the problem on?

Linux

Relevant log output

No response

[ggerganov]

Works on my end:

make -j && ./llama-cli -m models/gemma-2-27b-it/ggml-model-f16.gguf -cnv -s 1 -p "You are helpful AI assistant."

0.02.626.446 I sampler chain: logits -> logit-bias -> penalties -> top-k -> tail-free -> typical -> top-p -> min-p -> temp-ext -> softmax -> dist 
0.02.626.447 I generate: n_ctx = 8192, n_batch = 2048, n_predict = -1, n_keep = 1

== Running in interactive mode. ==
 - Press Ctrl+C to interject at any time.
 - Press Return to return control to the AI.
 - To return control without starting a new line, end your input with '/'.
 - If you want to submit another line, end your input with '\'.


> Hi
Hello! 👋 How can I help you today? 😊

>        
0.15.759.453 I llama_perf_sampler_print:    sampling time =       1.06 ms /    28 runs   (    0.04 ms per token, 26490.07 tokens per second)
0.15.759.468 I llama_perf_context_print:        load time =    2625.69 ms
0.15.759.472 I llama_perf_context_print: prompt eval time =    4617.03 ms /    17 tokens (  271.59 ms per token,     3.68 tokens per second)

[ggerganov]

Ignore my previous comment - I've pushed a fix in #9559. Let me know if this resolves the problem.

[ownia]

LGTM!