.dockerignore reinclusion retains unrelated files in a glob-excluded directory
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
- docker, python
- 领域
- build-system
调研方向
Start from docker.utils.tar and the exclude/reinclusion pattern matching behind it (docker/utils/build.py) to see how **/cache combined with !cache/keep.txt is evaluated. Run the reproducer script from the issue (reproduce-ancestor.py) to confirm the current output keeps cache/drop.txt, then compare with the stated BuildKit expectation: only cache/keep.txt survives while src/cache/drop.txt is dropped. Done means the reproducer prints ['cache/keep.txt'] for the fixture files and the existing unit tests for build-context tar creation still pass.
由索引模型根据 Issue 内容生成。
描述
🤖 this issue description was generated by an LLM. i reviewed it before submitting. 🤖
I used Codex for the investigation, reproduction scripts, and this draft.
When I generate a build context with docker.utils.tar, the rules **/cache and !cache/keep.txt retain both cache/keep.txt and cache/drop.txt. Docker's BuildKit keeps only cache/keep.txt with the same input. The SDK therefore includes an unrelated file that the ignore rule excludes.
Reproduction
Save this script as reproduce-ancestor.py:
import tarfile
from pathlib import Path
from tempfile import TemporaryDirectory
import docker
from docker.utils import tar
patterns = ['**/cache', '!cache/keep.txt']
files = ['cache/drop.txt', 'cache/keep.txt', 'src/cache/drop.txt']
with TemporaryDirectory() as directory:
root = Path(directory)
(root / 'Dockerfile').write_text('FROM scratch\nCOPY . /\n')
(root / '.dockerignore').write_text('\n'.join(patterns) + '\n')
for relative in files:
path = root / relative
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(relative)
with tar(str(root), exclude=patterns.copy()) as context:
with tarfile.open(fileobj=context) as archive:
actual = sorted(name for name in archive.getnames() if name in files)
print('docker:', docker.__version__)
print('patterns:', patterns)
print('SDK context files:', actual)
I ran it in an isolated uv environment, using commit 56343ddf8f0c44281e151c2dad016c16cdb8393d:
uv run --isolated --no-project --no-config --no-cache \
--python /home/jyn/.local/share/mise/installs/python/3.14.7/bin/python3 \
--with 'docker @ git+https://github.com/docker/docker-py@56343ddf8f0c44281e151c2dad016c16cdb8393d' \
python reproduce-ancestor.py
The --python path selects my tested Python installation; use the path to your Python executable on another machine. The pinned requirement selects the upstream SDK rather than an installed copy. I did not modify the SDK or replace its modules with mocks. This reproducer creates disposable files and examines the SDK's context tar without contacting a Docker daemon.
Expected behavior
The SDK should retain only cache/keep.txt from the fixture payload files. It should exclude cache/drop.txt and src/cache/drop.txt. I built the same files with FROM scratch and COPY . / using BuildKit, and its local output retained only cache/keep.txt.
Actual output
docker: 7.2.1.dev26+g56343ddf8
patterns: ['**/cache', '!cache/keep.txt']
SDK context files: ['cache/drop.txt', 'cache/keep.txt']
BuildKit control
I ran the following standalone control with Docker's active lima context and lima builder. It creates the same input files, runs docker buildx build with a local filesystem export, and enumerates every exported regular file before removing the temporary directories. The command uses the active Docker context and builder; another machine can use its own working configuration.
Buildx command and exported-file inspection
Save this script as buildkit-control-ancestor.py:
import subprocess
from pathlib import Path
from tempfile import TemporaryDirectory
patterns = ['**/cache', '!cache/keep.txt']
files = ['cache/drop.txt', 'cache/keep.txt', 'src/cache/drop.txt']
with TemporaryDirectory() as directory:
context = Path(directory) / 'input'
context.mkdir()
output = Path(directory) / 'output'
(context / 'Dockerfile').write_text('FROM scratch\nCOPY . /\n')
(context / '.dockerignore').write_text('\n'.join(patterns) + '\n')
for relative in files:
path = context / relative
path.parent.mkdir(parents=True, exist_ok=True)
path.write_text(relative)
subprocess.run([
'docker', 'buildx', 'build',
'--progress=plain', '--provenance=false',
'--output', f'type=local,dest={output}', str(context),
], check=True)
print('BuildKit output files:', sorted(
path.relative_to(output).as_posix()
for path in output.rglob('*') if path.is_file()
))
Run it with:
python3 buildkit-control-ancestor.py
I got this stdout; Buildx wrote its progress log to stderr and exited successfully:
BuildKit output files: ['.dockerignore', 'Dockerfile', 'cache/keep.txt']
Environment
- Docker SDK for Python:
7.2.1.dev26+g56343ddf8, using upstream commit56343ddf8f0c44281e151c2dad016c16cdb8393d. - Python:
3.14.7 (main, Sep 29 2026, 15:01:40) [Clang 22.1.3 ]. - OS/distribution: CachyOS Linux, rolling release (
BUILD_ID=rolling;/etc/os-releasedoes not defineVERSION_ID). - Kernel/platform:
Linux-7.2.9-1-cachyos-x86_64-with-glibc2.44. - Docker CLI:
29.8.2; Docker Engine:29.8.0(linux/amd64). - Buildx:
github.com/docker/buildx 0.37.2 2d379c0c3f22da0d2759d132a0ec81ca949098f0. - The SDK tar reproducer requires no engine connection; the separate BuildKit comparison used the versions above.
- 主要语言
- Python
- 星标
- 7.2k
- 派生
- 1.7k
- PR 合并指标
- 30 天内没有已合并 PR
环境准备
- 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
- 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。
docker/docker-py 的其他 Issue
-
.dockerignore matching ignores filename case on Linux可能已有人在做 @locker95 于 13 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 82/100
-
Add `pull` to the list of kwargs to pass to create and run可能已有人在做 @zainnadeem786 于 143 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 75/100
-
难度 1/5 1 小时以内 新手友好度 68/100
-
难度 2/5 1-3 小时 新手友好度 65/100
-
containers.run() returns None for the local logging driver, now the Docker Desktop 4.94 default可能已有人在做 @locker95 于 7 天前认领。 未关闭
难度 3/5 1-2 天 新手友好度 76/100
相似的 Issue
-
documentation good first issue
难度 2/5 1-3 小时 新手友好度 82/100
selimfirat/pysad#379 ·
维护者通常 1 天内回复
-
bug
难度 2/5 1-3 小时 新手友好度 75/100
open-webui/open-webui#32805 ·
维护者通常 1 天内回复
-
fix: [bug description] HTML CVE metric table shows only the last CVE for each product可能已有人在做 @Chris-wh233 今天认领。 未关闭bug
难度 2/5 1-3 小时 新手友好度 66/100
ossf/cve-bin-tool#5976 ·
维护者通常 1 天内回复
-
needs-triage
难度 2/5 1-3 小时 新手友好度 72/100
gke-labs/kube-agents#3615 ·
维护者通常 1 天内回复
-
deferred
难度 2/5 1-3 小时 新手友好度 70/100
维护者通常 1 天内回复