sys.executable can be a relative path contrary to documented behavior [3.11 regression]
还没有人认领这个 Issue。
- 主要语言
- Python
- 星标
- 77.2k
- 派生
- 35.9k
- PR 合并指标
- PR 指标待抓取
描述
Bug report
Bug description:
When PATH contains a relative directory, sys.executable will also be a relative directory contrary to the current sys.executable documentation:
A string giving the absolute path of the executable binary for the Python interpreter, on systems where this makes sense. If Python is unable to retrieve the real path to its executable,
sys.executablewill be an empty string orNone.
Context: I have a script that uses a custom virtualenv for dependencies that invokes itself in a different working directory. Because cwd is different, it relies on os.executable to be absolute. Minimal reproducer, set up python3 -m venv venv, then run PATH="venv/bin:$PATH" python3 script.py; using $PWD/venv/bin instead is a workaround for this issue.
import sys, subprocess; subprocess.run([sys.executable, "--version"], cwd="/tmp")
Alternative reproducer that reveals the issue:
$ python3 -m venv venv && PATH="venv/bin:$PATH" python3 -c 'import sys;print(sys.executable, sys.version.split()[0])'
<frozen site>:101: RuntimeWarning: Unexpected value in sys.prefix, expected /home/arch/venv, got venv
<frozen site>:101: RuntimeWarning: Unexpected value in sys.exec_prefix, expected /home/arch/venv, got venv
venv/bin/python3 3.14.6
This used to work in Python 3.10 and before, and is broken in Python 3.11 (2022) up to at least 3.14.6 and 3.15.0b3. I used this reproducer to confirm:
$ podman run --rm python:3.11-slim env PATH=usr/local/bin python3 -c 'import sys;print(sys.executable, sys.version.split()[0])'
usr/local/bin/python3 3.11.15
Another mild regression is the presence of an extra leading slash when using a relative path from the root directory:
$ podman run --rm python:3.15-rc-slim usr/local/bin/python3 -c 'import sys;print(sys.executable, sys.version.split()[0])'
//usr/local/bin/python3 3.15.0b3
$ podman run --rm python:3.10-slim usr/local/bin/python3 -c 'import sys;print(sys.executable, sys.version.split()[0])'
/usr/local/bin/python3 3.10.20
I suspect that both regressions have something to do with the refactoring done in https://bugs.python.org/issue45582 (#29041 by @zooba).
Somewhat related is #124241, that report is not about a relative path, but a modified argv[0]. Its current proposed fix to rely on /proc/self/exe (#145486 by @FFY00) could fix this issue, but has new regression risks.
CPython versions tested on:
3.15, 3.14, 3.13, 3.12, 3.11
Operating systems tested on:
Linux
Linked PRs
- gh-154163
贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
调研方向
先从 issue 中最小的 Python 和容器复现程序开始,然后阅读 #29041 中的重构,以及 #124241 和 PR #145486 中的相关提案。链接的 PR #154163 表明相关工作已经在进行中;完成后应为相对 PATH 条目恢复绝对的 sys.executable,并在受影响的 Python 版本中避免额外的前导斜杠。
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- python
- 领域
- operating-systems
- Issue 类型
- 缺陷
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 活跃度
- 停滞
- 描述清晰度
- 基本清楚
- 新手友好度
- 25/100