What is `corpus_dir` in README. Why are`data` and the `crash-*` and `slow-unit-*` always empty?

[khteh]

I run the fuzzer with -atheris_runs=100 and I always get empty data and not sure what I miss. Here is my code snippet:

import atheris, asyncio, sys
from datetime import datetime
from uuid_extensions import uuid7, uuid7str
from langchain_core.runnables import RunnableConfig, ensure_config
from src.rag_agent.RAGAgent import RAGAgent

@atheris.instrument_func
async def main(input_message):
    rag = RAGAgent()
    await rag.CreateGraph()
    config = RunnableConfig(run_name="RAG Deep Agent", configurable={"thread_id": uuid7str(), "user_id": uuid7str()})
    timestamp = datetime.now().strftime("%d-%m-%Y_%H-%M-%S")
    message = f"[Timestamp: {timestamp}]\n{input_message}\n"
    await rag.ChatAgent(config, message, ["values"], False)

@atheris.instrument_func
def FuzzEntryPoint(data):
    # Initialize the provider with raw bytes from the fuzzer
    fdp = atheris.FuzzedDataProvider(data)    
    # Consume structured data
    number = fdp.ConsumeIntInRange(100, 100000)
    input_message = fdp.ConsumeUnicodeNoSurrogates(128)
    print(f"input_messsage: {repr(input_message)}", flush=True)
    asyncio.run(main(input_message))

if __name__ == "__main__":
    atheris.Setup(sys.argv, FuzzEntryPoint)
    atheris.instrument_all()    
    atheris.Fuzz()    

This is what I see in the console:

INFO: Instrumenting functions: [58178/58178] 100%
INFO: Using built-in libfuzzer
WARNING: Failed to find function "__sanitizer_acquire_crash_state".
WARNING: Failed to find function "__sanitizer_print_stack_trace".
WARNING: Failed to find function "__sanitizer_set_death_callback".
INFO: Running with entropic power schedule (0xFF, 100).
INFO: Seed: 2817686543
INFO: -max_len is not provided; libFuzzer will not generate inputs larger than 4096 bytes
input_messsage: ''
================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-11]


================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-11]


================================== Ai Message ==================================
Tool Calls:
  write_file (311c9924-9f89-4722-8321-bfc9262c78f2)
 Call ID: 311c9924-9f89-4722-8321-bfc9262c78f2
  Args:
    file_path: /user_request_22-04-2026_19-37-11.md
    content:
================================= Tool Message =================================
Name: write_file

Updated file /user_request_22-04-2026_19-37-11.md
================================== Ai Message ==================================

It looks like the request body is empty. Could you please provide the question or topic you’d like me to address?
INFO: A corpus is not provided, starting from an empty corpus
input_messsage: ''
================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-26]


================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-26]


================================== Ai Message ==================================

Could you please provide the question or request you’d like me to address?
#2	pulse  ft: 11881 exec/s: 0 rss: 404Mb
#2	INITED cov: 11885 ft: 11881 corp: 1/1b exec/s: 0 rss: 404Mb
input_messsage: ''
================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-29]


================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-29]


================================== Ai Message ==================================

Could you let me know what you’d like assistance with?
#3	NEW    cov: 11895 ft: 12080 corp: 2/3b lim: 4 exec/s: 0 rss: 413Mb L: 2/2 MS: 1 InsertByte-
input_messsage: ''
================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-33]


================================ Human Message =================================

[Timestamp: 22-04-2026_19-37-33]


================================== Ai Message ==================================

I see the timestamp `22-04-2026_19-37-33`, but there’s no request or question following it. Could you let me know what you’d like assistance with?
#4	pulse  cov: 11895 ft: 12494 corp: 2/3b lim: 4 exec/s: 0 rss: 421Mb
#4	NEW    cov: 11911 ft: 12494 corp: 3/7b lim: 4 exec/s: 0 rss: 421Mb L: 4/4 MS: 1 CopyPart-
input_messsage: ''

As seen the Human Message is always empty except for the timestamp which is provided by the code. Question on the Exception ignored in: <function _removeHandlerRef at 0x79ced5d9a700>. Is that something caught and reported by artheris? It is not clear to me as it looks like a usual exception / python error that shows normally on a console when running app sometimes.

What's the corpus_dir in README? Is it data or the modules to be executed by the fuzzer?

Why are these files always seem to be empty?

$ ll crash-* slow-unit-*
-rw-rw-r-- 1 khteh khteh 1 Apr 20 13:35 crash-adc83b19e793491b1c6ea0fd8b46cd9f32e592fc
-rw-rw-r-- 1 khteh khteh 0 Apr 21 10:24 crash-da39a3ee5e6b4b0d3255bfef95601890afd80709
-rw-rw-r-- 1 khteh khteh 2 Apr 21 12:00 slow-unit-0a493543ed2c9330548e26865cbd5029a954b928
-rw-rw-r-- 1 khteh khteh 2 Apr 21 11:58 slow-unit-454ca76b8adc45fa3cb7347b6fe5f12339f93b7c

But I am able to generate html report with coverage data in it:

$ ll htmlcov/
total 924
drwxrwxr-x  2 khteh khteh   4096 Apr 20 15:36 ./
drwxrwxr-x 16 khteh khteh   4096 Apr 21 12:00 ../
-rw-rw-r--  1 khteh khteh     27 Apr 20 15:36 .gitignore
-rw-rw-r--  1 khteh khteh  29803 Apr 20 15:36 class_index.html
-rw-rw-r--  1 khteh khteh  25450 Apr 20 15:36 coverage_html_cb_dd2e7eb5.js
-rw-rw-r--  1 khteh khteh   1732 Apr 20 15:36 favicon_32_cb_c827f16f.png
-rw-rw-r--  1 khteh khteh  45023 Apr 20 15:36 function_index.html
-rw-rw-r--  1 khteh khteh  15143 Apr 20 15:36 index.html
-rw-rw-r--  1 khteh khteh   9004 Apr 20 15:36 keybd_closed_cb_900cfef5.png
-rw-rw-r--  1 khteh khteh   8817 Apr 20 15:36 status.json
-rw-rw-r--  1 khteh khteh  16104 Apr 20 15:36 style_cb_9ff733b0.css
-rw-rw-r--  1 khteh khteh   4570 Apr 20 15:36 z_145eef247bfb46b6___init___py.html
-rw-rw-r--  1 khteh khteh  39574 Apr 20 15:36 z_145eef247bfb46b6_config_py.html
-rw-rw-r--  1 khteh khteh  21584 Apr 20 15:36 z_af79f3804eddddbf_Configuration_py.html
-rw-rw-r--  1 khteh khteh   6721 Apr 20 15:36 z_af79f3804eddddbf_ContextSchema_py.html
-rw-rw-r--  1 khteh khteh   5477 Apr 20 15:36 z_af79f3804eddddbf_Prompts_py.html
-rw-rw-r--  1 khteh khteh  18078 Apr 20 15:36 z_af79f3804eddddbf_State_py.html
-rw-rw-r--  1 khteh khteh   4572 Apr 20 15:36 z_af79f3804eddddbf___init___py.html
-rw-rw-r--  1 khteh khteh  12239 Apr 20 15:36 z_b491f7ea070f9a13_Context_py.html
-rw-rw-r--  1 khteh khteh  67088 Apr 20 15:36 z_b491f7ea070f9a13_Tools_py.html
-rw-rw-r--  1 khteh khteh   5110 Apr 20 15:36 z_b491f7ea070f9a13___init___py.html
-rw-rw-r--  1 khteh khteh  35465 Apr 20 15:36 z_c969fe9432d0cef8_EmailPrompts_py.html
-rw-rw-r--  1 khteh khteh 140145 Apr 20 15:36 z_c969fe9432d0cef8_EmailRAG_py.html
-rw-rw-r--  1 khteh khteh  30704 Apr 20 15:36 z_c969fe9432d0cef8_RobustEmailModelParser_py.html
-rw-rw-r--  1 khteh khteh   5144 Apr 20 15:36 z_cb3a919737131835___init___py.html
-rw-rw-r--  1 khteh khteh  51564 Apr 20 15:36 z_cb3a919737131835_prompts_py.html
-rw-rw-r--  1 khteh khteh  51140 Apr 20 15:36 z_defad1a87623d8bf_EmailModel_py.html
-rw-rw-r--  1 khteh khteh  30917 Apr 20 15:36 z_defad1a87623d8bf_EscalationModel_py.html
-rw-rw-r--  1 khteh khteh   7691 Apr 20 15:36 z_defad1a87623d8bf___init___py.html
-rw-rw-r--  1 khteh khteh  37329 Apr 20 15:36 z_defad1a87623d8bf_schema_py.html
-rw-rw-r--  1 khteh khteh   4516 Apr 20 15:36 z_efda893aa850b0c0___init___py.html
-rw-rw-r--  1 khteh khteh  21808 Apr 20 15:36 z_efda893aa850b0c0_sample_emails_py.html
-rw-rw-r--  1 khteh khteh   4574 Apr 20 15:36 z_f1b38b22aeb65474___init___py.html
-rw-rw-r--  1 khteh khteh  12120 Apr 20 15:36 z_f1b38b22aeb65474_image_py.html
-rw-rw-r--  1 khteh khteh   7247 Apr 20 15:36 z_ff2b9fe67d372814_Backend_py.html
-rw-rw-r--  1 khteh khteh  19141 Apr 20 15:36 z_ff2b9fe67d372814_PostgreSQLSetup_py.html
-rw-rw-r--  1 khteh khteh  64674 Apr 20 15:36 z_ff2b9fe67d372814_VectorStore_py.html

[babenek]

crash-da39a3ee5e6b4b0d3255bfef95601890afd80709 - the input is empty, so the seed file is empty
https://en.wikipedia.org/wiki/SHA-1

[khteh]

crash-da39a3ee5e6b4b0d3255bfef95601890afd80709 - the input is empty, so the seed file is empty https://en.wikipedia.org/wiki/SHA-1

?

[babenek]

atheris.FuzzedDataProvider produces empty (None) seed first. The sha1 hash of empty data is da39a3ee5e6b4b0d3255bfef95601890afd80709 - so, the file crash-da39a3ee5e6b4b0d3255bfef95601890afd80709 means something was wrong during None has passed to fuzzing.
Try to instrument all. Add printf(repr(input_message), flush=True) for debug

[khteh]

Updated my post with the new log. The input_message is always empty.

[babenek]

Perhaps, your corpus dir is not set, so the seeds files cannot stored into the dir.
My approach:
https://github.com/Samsung/CredSweeper/blob/f4568cddec5021eb747b3182b13eaf0b45bee9a4/fuzz/fuzzing.sh#L19

in your case:
@atheris.instrument_func do not neccessary

atheris.instrument_all() - put before atheris.Setup

[khteh]

To no avail!

================================== Ai Message ==================================

No request content was provided after the timestamp.
Done 100 in 968 second(s)
Exception ignored in: <function _removeHandlerRef at 0x7b8d7163a7a0>
Traceback (most recent call last):
  File "/usr/lib/python3.13/logging/__init__.py", line 891, in _removeHandlerRef
TypeError: 'NoneType' object is not callable

[khteh]

How do I interpret or use the crash-* file?

==381517== ERROR: libFuzzer: fuzz target exited
SUMMARY: libFuzzer: fuzz target exited
MS: 1 CrossOver-; base unit: adc83b19e793491b1c6ea0fd8b46cd9f32e592fc
0xa,0xa,
\012\012
artifact_prefix='./'; Test unit written to ./crash-71853c6197a6a7f222db0f1978c7cb232b87c5ee
Base64: Cgo=

[babenek]

copy crash-71853c6197a6a7f222db0f1978c7cb232b87c5ee to corpus_dir.
then you can run fuzzing without instrumentation - the data will be passed to data consumer and reproduce the crash.