Hacktoberfest 2026: những issue maintainer đã đánh dấu cho tháng Mười, đang mở và phù hợp người mới. Xem issue Hacktoberfest

Hosted/vendored NuGet mapping isn't exclusive when nuget.config already maps the exact package id to another source, so restore races the Socket feed against nuget.org (NU1403 with a lock, silently unpatched without one)

Đang mở
  • Có gì không ổn?
#462 1 bình luận 0 reaction 0 người được giao Xem trên GitHub

Maintainer thường phản hồi trong vòng 1 ngày

Chưa có ai nhận issue này.

Đánh giá

Độ khó
3/5
Thời gian dự kiến
1-2 ngày
Mức phù hợp với người mới
72/100
Loại issue
Lỗi
Độ rõ ràng
Khá rõ ràng
Mức độ hoạt động
Sôi nổi
Công nghệ
rust
Lĩnh vực
build-system, cli, devtools

Hướng nghiên cứu

Start with the hosted mapping logic in crates/socket-patch-core/src/patch/redirect/mod.rs:4132-4140 and the vendored paths in crates/socket-patch-core/src/vendor/nuget_feed.rs:896 and :915. Use crates/socket-patch-cli/tests/e2e_nuget_dotnet_build.rs and the described NuGet fixture to verify that an existing exact-id mapping no longer leaves the Socket source tied with another source; locked and lockless restores should resolve only the patched package.

Do mô hình lập chỉ mục viết ra từ nội dung của issue.

Mô tả

agent:triaged bug bughunt pm:nuget priority:p3

[agent] Found by the scheduled NuGet / dotnet bug-hunt routine (ledger #320).

Summary

When the project's nuget.config already has a <packageSourceMapping> that names the patched package's exact id under another source (Visual Studio's package-source-mapping UI writes exactly this, e.g. <package pattern="Newtonsoft.Json" /> under nuget.org), scan --mode hosted (and --mode vendored) just adds a second exact-id mapping under socket-patch-<uuid>. Two sources now have the same most-specific pattern, so NuGet treats both as eligible and takes the package from whichever answers first.

The rewriter assumes "most specific pattern wins" (crates/socket-patch-core/src/patch/redirect/mod.rs:6696). That doesn't hold for a tie, and nothing detects the tie, warns about it, or removes the user's entry.

Impact

  • With packages.lock.json: if nuget.org answers before the Socket feed, a fresh dotnet restore --locked-mode downloads the upstream nupkg and fails with NU1403: Package content hash validation failed. Whether CI is green or red then depends on feed latency.
  • Without a lock: the same restore exits 0 and installs the unpatched upstream bytes. scan reported success and printed no warning.

Repro

This uses the repo's own real-dotnet harness (crates/socket-patch-cli/tests/e2e_nuget_dotnet_build.rs: its Backend wiremock stand-in for the API and the hosted feed, a real nuget.org fixture restore, and fresh_checkout). After restore_fixture, overwrite the fixture's config with the one below and restore it once (--locked-mode, which passes):

<configuration>
  <packageSources>
    <clear />
    <add key="nuget.org" value="https://api.nuget.org/v3/index.json" />
  </packageSources>
  <packageSourceMapping>
    <packageSource key="nuget.org">
      <package pattern="*" />
      <package pattern="Newtonsoft.Json" />
    </packageSource>
  </packageSourceMapping>
</configuration>

Then run:

  1. socket-patch scan --mode hosted --json --yes --api-url <stub> --org test-org --api-token x --patch-server-url <stub> exits 0, warnings: [].
  2. Make a fresh checkout (csproj, nuget.config, lock), then run dotnet restore --locked-mode with a cold NUGET_PACKAGES and a cold http cache.
  3. Make the stand-in's flat-container index.json answer 3 s late (ResponseTemplate::set_delay), standing in for a Socket feed that is slower than nuget.org's CDN.

Config after step 1:

  <packageSourceMapping>
    <packageSource key="socket-patch-4e4e4e4e-…">
      <package pattern="Newtonsoft.Json" />
    </packageSource>
    <packageSource key="nuget.org">
      <package pattern="*" />
      <package pattern="Newtonsoft.Json" />      <!-- still there: a tie -->
    </packageSource>
  </packageSourceMapping>

Results (2/2 runs, 2 cold restores each):

shape Socket feed delay result
user exact-id mapping, locked restore 3000 ms error NU1403, 0 requests for the Socket nupkg, cache holds the pristine LICENSE.md
user exact-id mapping, lockless restore 3000 ms exit 0, pristine bytes installed
user exact-id mapping, locked restore 0 ms / 500 ms pass (the local stand-in wins the race)
control: default nuget.config (no user mapping), locked restore 3000 ms pass (the Socket source is the only eligible one)

The control shows the failure comes from the tie, not from the delay. Vendored mode writes the same tie (socket-patch-<uuid> → .socket/vendor/nuget/<uuid> beside the user's nuget.org → Newtonsoft.Json). Its local folder feed usually wins the race, so I didn't get a failing vendored restore. The config is equally non-exclusive, though.

Expected vs actual

  • Expected: CLI_CONTRACT.md (hosted/vendored table, nuget row) says the rewrite adds "source socket-patch-<uuid> + its exclusive exact-id <packageSourceMapping>". docs/ecosystems.md (NuGet row) says hosted pins the id via source mapping plus the lock contentHash. The patched id should resolve only from the Socket source. So the rewriter should drop or override the user's same-id pattern on other sources, or at least refuse or warn.
  • Actual: the user's exact-id entry stays, the mapping isn't exclusive, restore is nondeterministic, and the lockless case silently installs the unpatched package.

Matrix

OS SDK hosted vendored
Linux 8.0.131 fail (locked: NU1403; lockless: silently unpatched) config tie reproduced; restore passed (local feed wins)
macOS / Windows, SDK 6/9/10 untested (SDK downloads from dot.net are blocked in this sandbox). The race is NuGet client behaviour, so I expect the same result untested

Main 2463257 (v5 consolidation, #277).

Suspect code

  • crates/socket-patch-core/src/patch/redirect/mod.rs:4132-4140: when a mapping already exists (!creating_mapping), it appends only the Socket <packageSource> and never looks for the same id under other sources.
  • crates/socket-patch-core/src/vendor/nuget_feed.rs:896 / :915: the same append for vendored.

Related, but not the same: #354 (the * fan-out cuts off inherited sources) and #352 (a warm cache shadows the patch).

Ngôn ngữ chính
Rust
Star
8
Fork
0
Merge trung bình
1 ngày 31 phút
Pull request đã merge (30 ngày)
151

Chuẩn bị môi trường

Bắt đầu từ đâu

  1. Đọc hết issue, rồi đọc hướng dẫn đóng góp của dự án.
  2. Bình luận trên issue rằng bạn sẽ nhận — tránh hai người làm cùng một việc.
  3. Fork repository và làm thay đổi trên một nhánh.
  4. Mở pull request có tham chiếu số hiệu của issue.

Issue khác của SocketDev/socket-patch

Tất cả issue của SocketDev/socket-patch

Issue tương tự

Thêm issue về Rust

Nhận issue mới trong hộp thư của bạn

Bản tóm tắt ngắn những issue GitHub phù hợp với người mới.