Merge branch 'fuzz-gix-commitgraph'

gix-commitgraph/fuzz/.gitignore

@@ -0,0 +1,4 @@
+target
+corpus
+artifacts
+coverage

gix-commitgraph/fuzz/Cargo.toml

@@ -0,0 +1,30 @@
+[package]
+name = "gix-commitgraph-fuzz"
+version = "0.0.0"
+publish = false
+edition = "2021"
+
+[package.metadata]
+cargo-fuzz = true
+
+[dependencies]
+anyhow = "1.0.76"
+arbitrary = { version = "1.3.2", features = ["derive"] }
+libfuzzer-sys = "0.4"
+memmap2 = "0.9.0"
+
+[dependencies.gix-commitgraph]
+path = ".."
+
+# Prevent this from interfering with workspaces
+[workspace]
+members = ["."]
+
+[profile.release]
+debug = 1
+
+[[bin]]
+name = "fuzz_file"
+path = "fuzz_targets/fuzz_file.rs"
+test = false
+doc = false

gix-commitgraph/fuzz/fuzz_targets/fuzz_file.rs

@@ -0,0 +1,29 @@
+#![no_main]
+
+use anyhow::Result;
+use gix_commitgraph::File;
+use libfuzzer_sys::fuzz_target;
+use std::hint::black_box;
+
+fn fuzz(data: &[u8]) -> Result<()> {
+    let data = {
+        let mut d = memmap2::MmapMut::map_anon(data.len())?;
+        d.copy_from_slice(data);
+        d.make_read_only()?
+    };
+    let file = File::new(data, "does not matter".into())?;
+
+    _ = black_box(file.iter_base_graph_ids().count());
+    _ = black_box(file.iter_commits().count());
+    _ = black_box(file.iter_ids().count());
+
+    let _ = black_box(file.checksum());
+    let _ = black_box(file.verify_checksum());
+    let _ = black_box(file.object_hash());
+
+    Ok(())
+}
+
+fuzz_target!(|data: &[u8]| {
+    _ = black_box(fuzz(data));
+});

gix-commitgraph/src/file/init.rs

@@ -1,3 +1,4 @@
+use std::path::PathBuf;
 use std::{
     convert::{TryFrom, TryInto},
     path::Path,
@@ -62,24 +63,13 @@ impl File {
     pub fn at(path: impl AsRef<Path>) -> Result<File, Error> {
         Self::try_from(path.as_ref())
     }
-}
 
-impl TryFrom<&Path> for File {
-    type Error = Error;
-
-    fn try_from(path: &Path) -> Result<Self, Self::Error> {
-        let data = std::fs::File::open(path)
-            .and_then(|file| {
-                // SAFETY: we have to take the risk of somebody changing the file underneath. Git never writes into the same file.
-                #[allow(unsafe_code)]
-                unsafe {
-                    Mmap::map(&file)
-                }
-            })
-            .map_err(|e| Error::Io {
-                err: e,
-                path: path.to_owned(),
-            })?;
+    /// A lower-level constructor which constructs a new instance directly from the mapping in `data`,
+    /// assuming that it originated from `path`.
+    ///
+    /// Note that `path` is only used for verification of the hash its basename contains, but otherwise
+    /// is not of importance.
+    pub fn new(data: memmap2::Mmap, path: PathBuf) -> Result<File, Error> {
         let data_size = data.len();
         if data_size < MIN_FILE_SIZE {
             return Err(Error::Corrupt(
@@ -240,13 +230,33 @@ impl TryFrom<&Path> for File {
             extra_edges_list_range,
             fan,
             oid_lookup_offset,
-            path: path.to_owned(),
+            path,
             hash_len: object_hash.len_in_bytes(),
             object_hash,
         })
     }
 }
 
+impl TryFrom<&Path> for File {
+    type Error = Error;
+
+    fn try_from(path: &Path) -> Result<Self, Self::Error> {
+        let data = std::fs::File::open(path)
+            .and_then(|file| {
+                // SAFETY: we have to take the risk of somebody changing the file underneath. Git never writes into the same file.
+                #[allow(unsafe_code)]
+                unsafe {
+                    Mmap::map(&file)
+                }
+            })
+            .map_err(|e| Error::Io {
+                err: e,
+                path: path.to_owned(),
+            })?;
+        Self::new(data, path.to_owned())
+    }
+}
+
 // Copied from gix-odb/pack/index/init.rs
 fn read_fan(d: &[u8]) -> ([u32; FAN_LEN], usize) {
     let mut fan = [0; FAN_LEN];