Corpus directory is dumped into and re-imported from itself every iteration

未关闭
#403 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
3/5
预计耗时
1-2 天
新手友好度
74/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
rust

调研方向

Read src/fuzzer/mod.rs around lines 413–419 and 529–530, focusing on how SyncFromDiskStage, DumpToDiskStage, and the disk-backed corpora use their directories. Trace which files the sync stage loads and check the LibAFL callback options. Done when the redundant copies and metadata files are no longer re-imported; the issue notes this was confirmed by code reading but not reproduced under SIMICS.

由索引模型根据 Issue 内容生成。

描述

Where

Problem
SyncFromDiskStage::with_from_file(corpus_directory) and DumpToDiskStage(corpus_directory, solutions_directory) target the same directories that CachedOnDiskCorpus/OnDiskCorpus already persist to. Each fuzz_one, the dump stage writes an extra id_{i}_{name} copy of every corpus entry and solution, and the sync stage loads every recent file as a raw BytesInput, with no dotfile filter: including .{name}.metadata JSON, the id_* copies and *.testcase files.

Impact
Wasted executions on duplicates, and JSON metadata executed as fuzz input (which can itself enter the corpus). Likely contributes to near-duplicate testcases.

Suggested fix
Drop DumpToDiskStage (redundant), and either sync from a separate directory or use SyncFromDiskStage::new with a callback that skips dotfiles.

Related
#101, #60

Status
Confirmed by reading the code at 79972db. Found by reading the code; not reproduced under SIMICS.

主要语言
Rust
星标
335
派生
26
平均合并
1 天 50 分钟
30 天内合并 PR
12

环境准备

在 Codespaces 中打开

在浏览器里用你自己的 GitHub 账号启动这个项目的开发容器。

  • 提供 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

intel/tsffs 的其他 Issue

查看 intel/tsffs 的全部 Issue

相似的 Issue

更多 Rust Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。