Merge branch 'feat_basic_connectivity_check'

Author: Byron

Cargo.lock

@@ -281,6 +281,7 @@ members = [
     "gix-archive",
     "gix-worktree-stream",
     "gix-revwalk",
+    "gix-fsck",
 
     "tests/tools",
 

Cargo.toml

@@ -140,6 +140,7 @@ is usable to some extent.
   * [gix-tui](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-tui)
   * [gix-tix](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-tix)
   * [gix-bundle](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-bundle)
+  * [gix-fsck](https://github.com/Byron/gitoxide/blob/main/crate-status.md#gix-fsck)
 
 ### Stress Testing
   * [x] Verify huge packs

README.md

@@ -775,6 +775,23 @@ See its [README.md](https://github.com/Byron/gitoxide/blob/main/gix-lock/README.
 * [x] validate submodule names
 * [x] [validate][tagname-validation] tag names
 
+### gix-fsck
+* [x] validate connectivity and find missing objects starting from…
+    - [x] commits
+    - [ ] tags
+    - [ ] tree-cache in the `index` or any entry within
+* [ ] validate object hashes during connectivity traversal
+* [ ] progress reporting and interruptability
+* [ ] skipList to exclude objects which are known to be broken
+* [ ] validate blob hashes (connectivity check
+* [ ] identify objects that exist but are not reachable (i.e. what remains after a full graph traversal from all valid starting points)
+* [ ] write dangling objects to the `.git/log-found` directory structure
+* [ ] `strict` mode, to check for tree objects with `g+w` permissions
+* [ ] consider reflog entries from `ref` starting points
+* [ ] when reporting reachable objects, provide the path through which they are reachable, i.e. ref-log@{3} -> commit -> tree -> path-in-tree
+* [ ] limit search to ODB without alternates (default is equivalent to `git fsck --full` due to ODB implementation)
+* [ ] all individual [checks available in `git fsck`](https://git-scm.com/docs/git-fsck#_fsck_messages) (*too many to print here*)
+
 ### gix-ref
 * [ ] Prepare code for arrival of longer hashes like Sha256. It's part of the [V2 proposal][reftable-v2] but should work for loose refs as well.
 * **Stores**

crate-status.md

@@ -15,6 +15,7 @@ function indent () {
 }
 
 echo "in root: gitoxide CLI"
+(enter gix-fsck && indent cargo diet -n --package-size-limit 10KB)
 (enter gix-actor && indent cargo diet -n --package-size-limit 10KB)
 (enter gix-archive && indent cargo diet -n --package-size-limit 10KB)
 (enter gix-worktree-stream && indent cargo diet -n --package-size-limit 40KB)

etc/check-package-size.sh

@@ -49,6 +49,7 @@ gix-pack-for-configuration-only = { package = "gix-pack", version = "^0.44.0", p
 gix-transport-configuration-only = { package = "gix-transport", version = "^0.38.0", path = "../gix-transport", default-features = false }
 gix-archive-for-configuration-only = { package = "gix-archive", version = "^0.6.0", path = "../gix-archive", optional = true, features = ["tar", "tar_gz"] }
 gix-status = { version = "^0.2.0", path = "../gix-status" }
+gix-fsck = { version = "^0.1.0", path = "../gix-fsck" }
 serde = { version = "1.0.114", optional = true, default-features = false, features = ["derive"] }
 anyhow = "1.0.42"
 thiserror = "1.0.34"

gitoxide-core/Cargo.toml

@@ -0,0 +1,36 @@
+use anyhow::Context;
+use gix::{objs::Kind, ObjectId};
+
+pub fn function(mut repo: gix::Repository, spec: Option<String>, mut out: impl std::io::Write) -> anyhow::Result<()> {
+    let spec = spec.unwrap_or("HEAD".into());
+
+    repo.object_cache_size_if_unset(4 * 1024 * 1024);
+    // We expect to be finding a bunch of non-existent objects here - never refresh the ODB
+    repo.objects.refresh_never();
+
+    let id = repo
+        .rev_parse_single(spec.as_str())
+        .context("Only single revisions are supported")?;
+    let commits: gix::revision::Walk<'_> = id
+        .object()?
+        .peel_to_kind(gix::object::Kind::Commit)
+        .context("Need commitish as starting point")?
+        .id()
+        .ancestors()
+        .all()?;
+
+    let on_missing = |oid: &ObjectId, kind: Kind| {
+        writeln!(out, "{oid}: {kind}").expect("failed to write output");
+    };
+
+    let mut check = gix_fsck::Connectivity::new(&repo.objects, on_missing);
+    // Walk all commits, checking each one for connectivity
+    for commit in commits {
+        let commit = commit?;
+        check.check_commit(&commit.id)?;
+        // Note that we leave parent-iteration to the commits iterator, as it will
+        // correctly handle shallow repositories which are expected to have the commits
+        // along the shallow boundary missing.
+    }
+    Ok(())
+}

gitoxide-core/src/repository/fsck.rs

@@ -35,6 +35,8 @@ pub use clone::function::clone;
 pub use fetch::function::fetch;
 
 pub mod commitgraph;
+mod fsck;
+pub use fsck::function as fsck;
 pub mod index;
 pub mod mailmap;
 pub mod odb;

gitoxide-core/src/repository/mod.rs

@@ -0,0 +1,6 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

gix-fsck/CHANGELOG.md

@@ -0,0 +1,22 @@
+[package]
+name = "gix-fsck"
+version = "0.1.0"
+repository = "https://github.com/Byron/gitoxide"
+authors = ["Cameron Esfahani <[email protected]>", "Sebastian Thiel <[email protected]>"]
+license = "MIT OR Apache-2.0"
+description = "Verifies the connectivity and validity of objects in the database"
+edition = "2021"
+include = ["src/**/*", "LICENSE-*"]
+rust-version = "1.65"
+
+[lib]
+doctest = false
+
+[dependencies]
+gix-hash = { version = "^0.13.1", path = "../gix-hash" }
+gix-hashtable = { version = "^0.4.0", path = "../gix-hashtable" }
+gix-object = { version = "^0.38.0", path = "../gix-object" }
+
+[dev-dependencies]
+gix-odb = { path = "../gix-odb" }
+gix-testtools = { path = "../tests/tools"}

gix-fsck/Cargo.toml

@@ -0,0 +1 @@
+../LICENSE-APACHE

gix-fsck/LICENSE-APACHE

@@ -0,0 +1 @@
+../LICENSE-MIT

gix-fsck/LICENSE-MIT

@@ -0,0 +1,107 @@
+//! A library for performing object database integrity and connectivity checks
+#![deny(rust_2018_idioms, unsafe_code, missing_docs)]
+
+use gix_hash::ObjectId;
+use gix_hashtable::HashSet;
+use gix_object::{tree::EntryMode, Exists, FindExt, Kind};
+use std::collections::VecDeque;
+
+/// Perform a connectivity check.
+pub struct Connectivity<T, F>
+where
+    T: FindExt + Exists,
+    F: FnMut(&ObjectId, Kind),
+{
+    /// ODB handle to use for the check
+    db: T,
+    /// Closure to invoke when a missing object is encountered
+    missing_cb: F,
+    /// Set of Object IDs already (or about to be) scanned during the check
+    seen: HashSet,
+    /// A buffer to keep a single object at a time.
+    buf: Vec<u8>,
+}
+
+impl<T, F> Connectivity<T, F>
+where
+    T: FindExt + Exists,
+    F: FnMut(&ObjectId, Kind),
+{
+    /// Instantiate a connectivity check.
+    pub fn new(db: T, missing_cb: F) -> Connectivity<T, F> {
+        Connectivity {
+            db,
+            missing_cb,
+            seen: HashSet::default(),
+            buf: Default::default(),
+        }
+    }
+
+    /// Run the connectivity check on the provided commit `oid`.
+    ///
+    /// ### Algorithm
+    ///
+    /// Walk the trees and blobs referenced by the commit and verify they exist in the ODB.
+    /// Any objects previously encountered by this instance will be skipped silently.
+    /// Any referenced blobs that are not present in the ODB will result in a call to the  `missing_cb`.
+    /// Missing commits or trees will cause an error to be returned.
+    ///     - TODO: consider how to handle a missing commit (invoke `missing_cb`, or possibly return a Result?)
+    pub fn check_commit(&mut self, oid: &ObjectId) -> Result<(), gix_object::find::existing_object::Error> {
+        // Attempt to insert the commit ID in the set, and if already present, return immediately
+        if !self.seen.insert(*oid) {
+            return Ok(());
+        }
+        // Obtain the commit's tree ID
+        let tree_id = {
+            let commit = self.db.find_commit(oid, &mut self.buf)?;
+            commit.tree()
+        };
+
+        let mut tree_ids = VecDeque::from_iter(Some(tree_id));
+        while let Some(tree_id) = tree_ids.pop_front() {
+            if self.seen.insert(tree_id) {
+                self.check_tree(&tree_id, &mut tree_ids);
+            }
+        }
+
+        Ok(())
+    }
+
+    /// Blobs are checked right away, trees are stored in `tree_ids` for the parent to iterate them, and only
+    /// if they have not been `seen` yet.
+    fn check_tree(&mut self, oid: &ObjectId, tree_ids: &mut VecDeque<ObjectId>) {
+        let Ok(tree) = self.db.find_tree(oid, &mut self.buf) else {
+            (self.missing_cb)(oid, Kind::Tree);
+            return;
+        };
+
+        for entry_ref in tree.entries.iter() {
+            match entry_ref.mode {
+                EntryMode::Tree => {
+                    let tree_id = entry_ref.oid.to_owned();
+                    if self.seen.insert(tree_id) {
+                        tree_ids.push_back(tree_id);
+                    }
+                }
+                EntryMode::Blob | EntryMode::BlobExecutable | EntryMode::Link => {
+                    let blob_id = entry_ref.oid.to_owned();
+                    if self.seen.insert(blob_id) {
+                        check_blob(&self.db, &blob_id, &mut self.missing_cb);
+                    }
+                }
+                EntryMode::Commit => {
+                    // Skip submodules as it's not in this repository!
+                }
+            }
+        }
+    }
+}
+
+fn check_blob<F>(db: impl Exists, oid: &ObjectId, mut missing_cb: F)
+where
+    F: FnMut(&ObjectId, Kind),
+{
+    if !db.exists(oid) {
+        missing_cb(oid, Kind::Blob);
+    }
+}

gix-fsck/src/lib.rs

@@ -0,0 +1,78 @@
+use gix_fsck::Connectivity;
+use gix_hash::ObjectId;
+use gix_hashtable::HashMap;
+use gix_object::Kind;
+use gix_testtools::once_cell::sync::Lazy;
+
+use crate::hex_to_id;
+
+fn check_missing<'a>(repo_name: &str, commits: impl IntoIterator<Item = &'a ObjectId>) -> HashMap<ObjectId, Kind> {
+    let db = {
+        let fixture_path = gix_testtools::scripted_fixture_read_only("make_test_repos.sh")
+            .expect("fixture path")
+            .join(repo_name)
+            .join(".git")
+            .join("objects");
+        let mut db = gix_odb::at(fixture_path).expect("valid odb");
+        db.refresh_never();
+        db
+    };
+
+    let mut missing: HashMap<ObjectId, Kind> = HashMap::default();
+    let record_missing_and_assert_no_duplicate = |oid: &ObjectId, kind: Kind| {
+        missing.try_insert(*oid, kind).expect("no duplicate oid");
+    };
+
+    let mut check = Connectivity::new(db, record_missing_and_assert_no_duplicate);
+    for commit in commits.into_iter() {
+        check.check_commit(commit).expect("commit is present")
+    }
+    missing
+}
+
+fn hex_to_ids<'a>(hex_ids: impl IntoIterator<Item = &'a str>) -> Vec<ObjectId> {
+    hex_ids.into_iter().map(hex_to_id).collect()
+}
+
+fn hex_to_objects<'a>(hex_ids: impl IntoIterator<Item = &'a str>, kind: Kind) -> HashMap<ObjectId, Kind> {
+    hex_to_ids(hex_ids).into_iter().map(|id| (id, kind)).collect()
+}
+
+// Get a `&Vec<ObjectID` for each commit in the test fixture repository
+fn all_commits() -> &'static [ObjectId] {
+    static ALL_COMMITS: Lazy<Vec<ObjectId>> = Lazy::new(|| {
+        hex_to_ids([
+            "5d18db2e2aabadf7b914435ef34f2faf8b4546dd",
+            "3a3dfaa55a515f3fb3a25751107bbb523af6a1b0",
+            "734c926856a328d1168ffd7088532e0d1ad19bbe",
+        ])
+    });
+    &ALL_COMMITS
+}
+
+#[test]
+fn no_missing() {
+    // The "base" repo is the original, and has every object present
+    assert_eq!(check_missing("base", all_commits()), HashMap::default());
+}
+
+#[test]
+fn missing_blobs() {
+    // The "blobless" repo is cloned with `--filter=blob:none`, and is missing one blob
+    let expected = hex_to_objects(["c18147dc648481eeb65dc5e66628429a64843327"], Kind::Blob);
+    assert_eq!(check_missing("blobless", all_commits()), expected);
+}
+
+#[test]
+fn missing_trees() {
+    // The "treeless" repo is cloned with `--filter=tree:0`, and is missing two trees
+    // NOTE: This repo is also missing a blob, but we have no way of knowing that, as the tree referencing it is missing
+    let expected = hex_to_objects(
+        [
+            "9561cfbae43c5e2accdfcd423378588dd10d827f",
+            "fc264b3b6875a46e9031483aeb9994a1b897ffd3",
+        ],
+        Kind::Tree,
+    );
+    assert_eq!(check_missing("treeless", all_commits()), expected);
+}

gix-fsck/tests/connectivity/mod.rs

@@ -0,0 +1 @@
+make_test_repos.tar.xz

gix-fsck/tests/fixtures/generated-archives/.gitignore

@@ -0,0 +1,32 @@
+#!/bin/bash
+set -x
+set -euo pipefail
+
+# We override the global config with our own local one (see below)
+export GIT_CONFIG_GLOBAL="$PWD/.gitconfig"
+
+# We need to be able to do partial clones, so enable it
+# - needs to be present in the persistent gitconfig, as a clone with `--no-local`
+git config --global uploadpack.allowFilter true
+
+# First build out a base repository
+git init base
+(
+    cd base
+
+    echo "blob 1" > blob-1
+    git add -A
+    git commit -m "commit 1"
+    echo "blob-2" > blob-2
+    git add -A
+    git commit -m "commit 2"
+    git rm blob-1
+    git add -A
+    git commit -m "commit 3"
+)
+
+# Blobless clone
+git clone --no-local --no-hardlinks --filter=blob:none ./base blobless
+
+# Treeless (and blobless) clone
+git clone --no-local --no-hardlinks --filter=tree:0 ./base treeless

gix-fsck/tests/fixtures/make_test_repos.sh

@@ -0,0 +1,7 @@
+use gix_hash::ObjectId;
+
+pub fn hex_to_id(hex: &str) -> ObjectId {
+    ObjectId::from_hex(hex.as_bytes()).expect("40 bytes hex")
+}
+
+mod connectivity;