Composer vendored repair (and re-vendor) rebuilds a deleted uuid dir without its socket-patch.vendor.json marker, leaving git showing the marker as deleted
Les mainteneurs répondent en général sous 1 jour
Personne n'a encore pris cette issue.
Évaluation
- Difficulté
- 2/5
- Temps estimé
- 1-3 heures
- Accessibilité débutants
- 90/100
Piste de recherche
Commencez dans crates/socket-patch-core/src/vendor/composer_lock.rs:311-345 et comparez le comportement de reconstruction avec crates/socket-patch-core/src/vendor/cargo.rs:928-940. Utilisez test_wired_missing_copy_rebuilds_artifact_only dans composer_lock.rs comme test de départ ; le travail est terminé lorsqu’un répertoire uuid supprimé reconstruit à la fois l’artefact et son marqueur socket-patch.vendor.json.
Rédigé par le modèle d'indexation à partir du texte de l'issue.
Description
[agent] Found by the scheduled Composer bug-hunt routine (ledger #321).
Summary
When a vendored Composer unit's uuid dir (.socket/vendor/composer/<uuid>/) is missing, for example because a checkout lost it or someone deleted it, socket-patch repair and a re-run of vendor download the patched copy again. Neither writes back the unit's socket-patch.vendor.json marker. The rebuild then reports success, but git status shows the marker as deleted. Committing that "repair" leaves the vendored unit without the marker that CLI_CONTRACT.md says every unit carries. Later vendor and repair runs (already_vendored, success []) never put it back.
The Cargo backend handles the same rebuild path correctly: it restores the marker when it's missing (crates/socket-patch-core/src/vendor/cargo.rs:928-940, pinned by test_wired_deleted_uuid_dir_rebuild_restores_marker). The Composer artifact-only rebuild branch doesn't.
Impact
Low. The install itself is fine: composer install from the repaired tree installs the patched bytes, and vex never reads the marker. But the repair leaves the committed vendored unit incomplete. External tools that rely on the marker ("belt-and-braces for tools that have the tree but not the lockfile") lose it, and the repair diff a user is asked to commit contains a surprise deletion.
Repro (Linux, Composer 1.10.28 and 2.10.3, PHP 8.3, main 61cfb9b)
The patch service is a local mock: POST /v0/orgs/acme/patches/package returns a granted result with an sha512 SRI, and GET serves a single-top-dir dist zip of the patched package. A manifest plus blob is staged under .socket/ for pkg:composer/acme/[email protected], from an inline package repository with a file:// zip dist.
composer update -n # installs acme/tool 1.0.0
printf '/vendor/\n' > .gitignore
git init -q && git add -A && git commit -qm init
socket-patch vendor --json --cwd . --api-url http://127.0.0.1:8765 --api-token fake --org acme
git add -A && git commit -qm vendored
git ls-files .socket/vendor | grep vendor.json
# .socket/vendor/composer/4d5e6f7a-…/socket-patch.vendor.json <- committed
rm -rf .socket/vendor/composer/4d5e6f7a-8b9c-4a1b-8c2d-0123456789ab
socket-patch repair --cwd . --api-url http://127.0.0.1:8765 --api-token fake --org acme
# Redownloaded pkg:composer/acme/[email protected] (.socket/vendor/composer/4d5e6f7a-…/acme/[email protected])
# Repair complete.
git status --short
# D .socket/vendor/composer/4d5e6f7a-8b9c-4a1b-8c2d-0123456789ab/socket-patch.vendor.json
socket-patch vendor --json --cwd . … # -> already_vendored
socket-patch repair --json --cwd . … # -> success, events: []
git status --short # marker still deleted
A vendor re-run instead of repair takes the same path. The JSON shows rebuilt with reason: vendor_artifact_missing, redownloaded: true, and the marker is still missing afterwards.
Expected vs actual
- Expected: CLI_CONTRACT.md ("Each vendored unit also carries an informational
socket-patch.vendor.jsonmarker … next to the artifact").repair"redownload[s] missing/corrupt vendored artifacts". The doc comment onwrite_marker_or_warn(crates/socket-patch-core/src/vendor/state.rs:808-812) says "Every backend's fresh and rebuild paths report it through here". After the rebuild the unit should be complete again (copy + marker), as it is for Cargo. - Actual: only the copy under
<uuid>/acme/[email protected]/is recreated. The marker stays deleted for good.
Matrix
| OS | Composer | Reproduces |
|---|---|---|
| Linux | 1.10.28 (PHP 8.3) | yes |
| Linux | 2.10.3 (PHP 8.3) | yes (×2) |
| macOS / Windows | n/a | untested. The code path is pure Rust filesystem logic, independent of OS and Composer version. |
First bad version: not a regression. v4.0.0 behaves the same way (rebuilt, marker left deleted).
Suspect code
crates/socket-patch-core/src/vendor/composer_lock.rs:311-345: the "Wired but the committed copy is missing/stale: rebuild the ARTIFACT only" branch calls composer_service_copy and mirror_filters::heal_or_warn, then returns. It never calls write_marker_or_warn, which only runs on the fresh path (composer_lock.rs:437-439). Porting Cargo's "restore the marker only when it is missing" block (cargo.rs:928-940) into this branch should fix it. test_wired_missing_copy_rebuilds_artifact_only (composer_lock.rs:1643) could assert the marker as well.
No probe runs: this is OS-independent.
- Langage dominant
- Rust
- Étoiles
- 8
- Forks
- 0
- Merge moyen
- 18 h 4 min
- PR mergées (30 j)
- 70
Préparer son environnement
- Aucun Dockerfile ni fichier Docker Compose
- Aucun modèle de pull request
- Lire le guide de contribution
Par où commencer
- Lisez l'issue en entier, puis le guide de contribution du projet.
- Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
- Forkez le dépôt et travaillez sur une branche.
- Ouvrez une pull request qui référence le numéro de l'issue.
Autres issues de SocketDev/socket-patch
-
agent:triaged bug bughunt pm:npm priority:p1
Difficulté 2/5 1-3 heures Accessibilité débutants 82/100
SocketDev/socket-patch#464 · 1 commentaire ·
Les mainteneurs répondent en général sous 1 jour
-
agent:triaged bug bughunt pm:npm priority:p1
Difficulté 2/5 1-3 heures Accessibilité débutants 82/100
SocketDev/socket-patch#433 · 1 commentaire ·
Les mainteneurs répondent en général sous 1 jour
-
agent:triaged bug bughunt pm:uv priority:p1
Difficulté 2/5 1-3 heures Accessibilité débutants 78/100
SocketDev/socket-patch#408 · 1 commentaire ·
Les mainteneurs répondent en général sous 1 jour
-
agent:triaged bug bughunt pm:yarn-berry priority:p1
Difficulté 2/5 1-3 heures Accessibilité débutants 82/100
SocketDev/socket-patch#370 · 2 commentaires ·
Les mainteneurs répondent en général sous 1 jour
-
Hosted Gradle snippet is always Groovy DSL, so pasting it into a build.gradle.kts fails to compileOuverteagent:triaged bug bughunt pm:gradle priority:p3
Difficulté 2/5 1-3 heures Accessibilité débutants 78/100
SocketDev/socket-patch#348 · 1 commentaire ·
Les mainteneurs répondent en général sous 1 jour
Toutes les issues de SocketDev/socket-patch
Issues similaires
-
Difficulté 2/5 1-3 heures Accessibilité débutants 78/100
Les mainteneurs répondent en général sous 1 jour
-
Difficulté 1/5 Moins d'une heure Accessibilité débutants 78/100
Les mainteneurs répondent en général sous 1 jour
-
Difficulté 2/5 1-3 heures Accessibilité débutants 78/100
Les mainteneurs répondent en général sous 1 jour
-
Difficulté 2/5 1-3 heures Accessibilité débutants 86/100
dani-garcia/vaultwarden#7801 ·
Les mainteneurs répondent en général sous 1 jour
-
Difficulté 2/5 1-3 heures Accessibilité débutants 90/100
boxlite-ai/boxlite#1814 ·
Les mainteneurs répondent en général sous 1 jour