python / python/cpython

Exceptions raised while tracing certain statements cannot be caught

未關閉
#148,278 4 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視

還沒有人認領這個 Issue。

3.13 3.14 3.15 interpreter-core type-bug
主要語言
Python
星號
77.2k
分支
36k
PR 合併指標
PR 指標待擷取

描述

Bug report

So I found a really odd bug today. except blocks don't catch exceptions raised when tracing certain statements that themselves "cannot raise". For example:

def f():
    breakpoint()
    try:
        pass  # If you raise while tracing this line, the exception isn't caught.
    except:
        ...
    try:
        42  # Ditto.
    except:
        ...
    try:
        return  # Ditto.
    except:
        ...

A couple of observations:

  • Exceptions raised while tracing other "normal" statements are handled the normal way.
  • The code for the exception handling is still present in these examples, even though it's statically unreachable.
  • If two or more of these "non-raising" statements occur together in the same block (like several pass statements in a row), only the last one will have the bug:
def f():
    breakpoint()
    try:
        pass  # No bug
        pass  # No bug
        pass  # Bug
    except:
        ...

I haven't invested any time into figuring out what the root cause is, I've just been poking at it with code examples. It's definitely a bug in the bytecode compiler though, not the interpreter (dis shows missing exception table entries). Probably something to do with how we emit exception tables for NOP or anything that unwinds the block stack, if I had to guess?

(This also affects 3.12, and maybe even earlier versions, but only tagging for 3.13 and newer since I don't think this is a "security" issue.)

Linked PRs
  • gh-148333

貢獻指南

開啟貢獻指南

從這裡開始

  1. 先讀完整個 Issue,再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫,在一個分支上完成修改。
  4. 送出 Pull Request,並在描述裡引用這個 Issue 編號。

研究方向

重現 issue 中的範例並檢查其 dis 輸出,重點關注 pass、42 和 return 缺少的例外表項目。將這些情況與一般陳述式的追蹤進行比較,並檢查編譯器對 NOP 的處理以及區塊堆疊展開。完成的標準是,except 區塊能夠捕捉追蹤這些陳述式時引發的例外,並且為所述案例提供回歸涵蓋率。

由索引模型根據 Issue 內容生成。

評估

技術堆疊
python
領域
compilers
Issue 類型
缺陷
難度
4/5
預估耗時
3-5 天
活躍度
停滯
描述清晰度
需要釐清
新手友好度
25/100

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。