rhg: Add RevlogEntry::data that does delta resolution

This requires keeping a `&Revlog` reference inside the `RevlogEntry` struct.
This struct already had the appropriate lifetime parameter.

Differential Revision: https://phab.mercurial-scm.org/D11960

rust/hg-core/src/operations/debugdata.rs

@@ -1,33 +1,33
 // debugdata.rs
 //
 // Copyright 2020 Antoine Cezar <[e-mail]>
 //
 // This software may be used and distributed according to the terms of the
 // GNU General Public License version 2 or any later version.
 
 use crate::repo::Repo;
 use crate::revlog::revlog::{Revlog, RevlogError};
 
 /// Kind of data to debug
 #[derive(Debug, Copy, Clone)]
 pub enum DebugDataKind {
     Changelog,
     Manifest,
 }
 
 /// Dump the contents data of a revision.
 pub fn debug_data(
     repo: &Repo,
     revset: &str,
     kind: DebugDataKind,
 ) -> Result<Vec<u8>, RevlogError> {
     let index_file = match kind {
         DebugDataKind::Changelog => "00changelog.i",
         DebugDataKind::Manifest => "00manifest.i",
     };
     let revlog = Revlog::open(repo, index_file, None)?;
     let rev =
         crate::revset::resolve_rev_number_or_hex_prefix(revset, &revlog)?;
     let data = revlog.get_rev_data(rev)?;
-    Ok(data)
+    Ok(data.into_owned())
 }

rust/hg-core/src/revlog/changelog.rs

@@ -1,67 +1,67
 use crate::errors::HgError;
 use crate::repo::Repo;
 use crate::revlog::node::NULL_NODE;
 use crate::revlog::revlog::{Revlog, RevlogError};
 use crate::revlog::Revision;
 use crate::revlog::{Node, NodePrefix};
 
 /// A specialized `Revlog` to work with `changelog` data format.
 pub struct Changelog {
     /// The generic `revlog` format.
     pub(crate) revlog: Revlog,
 }
 
 impl Changelog {
     /// Open the `changelog` of a repository given by its root.
     pub fn open(repo: &Repo) -> Result<Self, HgError> {
         let revlog = Revlog::open(repo, "00changelog.i", None)?;
         Ok(Self { revlog })
     }
 
     /// Return the `ChangelogEntry` for the given node ID.
     pub fn data_for_node(
         &self,
         node: NodePrefix,
     ) -> Result<ChangelogRevisionData, RevlogError> {
         let rev = self.revlog.rev_from_node(node)?;
         self.data_for_rev(rev)
     }
 
     /// Return the `ChangelogEntry` of the given revision number.
     pub fn data_for_rev(
         &self,
         rev: Revision,
     ) -> Result<ChangelogRevisionData, RevlogError> {
-        let bytes = self.revlog.get_rev_data(rev)?;
+        let bytes = self.revlog.get_rev_data(rev)?.into_owned();
         Ok(ChangelogRevisionData { bytes })
     }
 
     pub fn node_from_rev(&self, rev: Revision) -> Option<&Node> {
         self.revlog.node_from_rev(rev)
     }
 }
 
 /// `Changelog` entry which knows how to interpret the `changelog` data bytes.
 #[derive(Debug)]
 pub struct ChangelogRevisionData {
     /// The data bytes of the `changelog` entry.
     bytes: Vec<u8>,
 }
 
 impl ChangelogRevisionData {
     /// Return an iterator over the lines of the entry.
     pub fn lines(&self) -> impl Iterator<Item = &[u8]> {
         self.bytes
             .split(|b| b == &b'\n')
             .filter(|line| !line.is_empty())
     }
 
     /// Return the node id of the `manifest` referenced by this `changelog`
     /// entry.
     pub fn manifest_node(&self) -> Result<Node, HgError> {
         match self.lines().next() {
             None => Ok(NULL_NODE),
             Some(x) => Node::from_hex_for_repo(x),
         }
     }
 }

rust/hg-core/src/revlog/filelog.rs

@@ -1,89 +1,89
 use crate::errors::HgError;
 use crate::repo::Repo;
 use crate::revlog::path_encode::path_encode;
 use crate::revlog::revlog::{Revlog, RevlogError};
 use crate::revlog::NodePrefix;
 use crate::revlog::Revision;
 use crate::utils::files::get_path_from_bytes;
 use crate::utils::hg_path::HgPath;
 use crate::utils::SliceExt;
 use std::path::PathBuf;
 
 /// A specialized `Revlog` to work with file data logs.
 pub struct Filelog {
     /// The generic `revlog` format.
     revlog: Revlog,
 }
 
 impl Filelog {
     pub fn open(repo: &Repo, file_path: &HgPath) -> Result<Self, HgError> {
         let index_path = store_path(file_path, b".i");
         let data_path = store_path(file_path, b".d");
         let revlog = Revlog::open(repo, index_path, Some(&data_path))?;
         Ok(Self { revlog })
     }
 
     /// The given node ID is that of the file as found in a manifest, not of a
     /// changeset.
     pub fn data_for_node(
         &self,
         file_node: impl Into<NodePrefix>,
     ) -> Result<FilelogRevisionData, RevlogError> {
         let file_rev = self.revlog.rev_from_node(file_node.into())?;
         self.data_for_rev(file_rev)
     }
 
     /// The given revision is that of the file as found in a manifest, not of a
     /// changeset.
     pub fn data_for_rev(
         &self,
         file_rev: Revision,
     ) -> Result<FilelogRevisionData, RevlogError> {
-        let data: Vec<u8> = self.revlog.get_rev_data(file_rev)?;
+        let data: Vec<u8> = self.revlog.get_rev_data(file_rev)?.into_owned();
         Ok(FilelogRevisionData(data.into()))
     }
 }
 
 fn store_path(hg_path: &HgPath, suffix: &[u8]) -> PathBuf {
     let encoded_bytes =
         path_encode(&[b"data/", hg_path.as_bytes(), suffix].concat());
     get_path_from_bytes(&encoded_bytes).into()
 }
 
 /// The data for one revision in a filelog, uncompressed and delta-resolved.
 pub struct FilelogRevisionData(Vec<u8>);
 
 impl FilelogRevisionData {
     /// Split into metadata and data
     pub fn split(&self) -> Result<(Option<&[u8]>, &[u8]), HgError> {
         const DELIMITER: &[u8; 2] = &[b'\x01', b'\n'];
 
         if let Some(rest) = self.0.drop_prefix(DELIMITER) {
             if let Some((metadata, data)) = rest.split_2_by_slice(DELIMITER) {
                 Ok((Some(metadata), data))
             } else {
                 Err(HgError::corrupted(
                     "Missing metadata end delimiter in filelog entry",
                 ))
             }
         } else {
             Ok((None, &self.0))
         }
     }
 
     /// Returns the file contents at this revision, stripped of any metadata
     pub fn file_data(&self) -> Result<&[u8], HgError> {
         let (_metadata, data) = self.split()?;
         Ok(data)
     }
 
     /// Consume the entry, and convert it into data, discarding any metadata,
     /// if present.
     pub fn into_file_data(self) -> Result<Vec<u8>, HgError> {
         if let (Some(_metadata), data) = self.split()? {
             Ok(data.to_owned())
         } else {
             Ok(self.0)
         }
     }
 }

rust/hg-core/src/revlog/manifest.rs

@@ -1,193 +1,193
 use crate::errors::HgError;
 use crate::repo::Repo;
 use crate::revlog::revlog::{Revlog, RevlogError};
 use crate::revlog::Revision;
 use crate::revlog::{Node, NodePrefix};
 use crate::utils::hg_path::HgPath;
 use crate::utils::SliceExt;
 
 /// A specialized `Revlog` to work with `manifest` data format.
 pub struct Manifestlog {
     /// The generic `revlog` format.
     revlog: Revlog,
 }
 
 impl Manifestlog {
     /// Open the `manifest` of a repository given by its root.
     pub fn open(repo: &Repo) -> Result<Self, HgError> {
         let revlog = Revlog::open(repo, "00manifest.i", None)?;
         Ok(Self { revlog })
     }
 
     /// Return the `Manifest` for the given node ID.
     ///
     /// Note: this is a node ID in the manifestlog, typically found through
     /// `ChangelogEntry::manifest_node`. It is *not* the node ID of any
     /// changeset.
     ///
     /// See also `Repo::manifest_for_node`
     pub fn data_for_node(
         &self,
         node: NodePrefix,
     ) -> Result<Manifest, RevlogError> {
         let rev = self.revlog.rev_from_node(node)?;
         self.data_for_rev(rev)
     }
 
     /// Return the `Manifest` of a given revision number.
     ///
     /// Note: this is a revision number in the manifestlog, *not* of any
     /// changeset.
     ///
     /// See also `Repo::manifest_for_rev`
     pub fn data_for_rev(
         &self,
         rev: Revision,
     ) -> Result<Manifest, RevlogError> {
-        let bytes = self.revlog.get_rev_data(rev)?;
+        let bytes = self.revlog.get_rev_data(rev)?.into_owned();
         Ok(Manifest { bytes })
     }
 }
 
 /// `Manifestlog` entry which knows how to interpret the `manifest` data bytes.
 #[derive(Debug)]
 pub struct Manifest {
     /// Format for a manifest: flat sequence of variable-size entries,
     /// sorted by path, each as:
     ///
     /// ```text
     /// <path> \0 <hex_node_id> <flags> \n
     /// ```
     ///
     /// The last entry is also terminated by a newline character.
     /// Flags is one of `b""` (the empty string), `b"x"`, `b"l"`, or `b"t"`.
     bytes: Vec<u8>,
 }
 
 impl Manifest {
     pub fn iter(
         &self,
     ) -> impl Iterator<Item = Result<ManifestEntry, HgError>> {
         self.bytes
             .split(|b| b == &b'\n')
             .filter(|line| !line.is_empty())
             .map(ManifestEntry::from_raw)
     }
 
     /// If the given path is in this manifest, return its filelog node ID
     pub fn find_by_path(
         &self,
         path: &HgPath,
     ) -> Result<Option<ManifestEntry>, HgError> {
         use std::cmp::Ordering::*;
         let path = path.as_bytes();
         // Both boundaries of this `&[u8]` slice are always at the boundary of
         // an entry
         let mut bytes = &*self.bytes;
 
         // Binary search algorithm derived from `[T]::binary_search_by`
         // <https://github.com/rust-lang/rust/blob/1.57.0/library/core/src/slice/mod.rs#L2221>
         // except we don’t have a slice of entries. Instead we jump to the
         // middle of the byte slice and look around for entry delimiters
         // (newlines).
         while let Some(entry_range) = Self::find_entry_near_middle_of(bytes)? {
             let (entry_path, rest) =
                 ManifestEntry::split_path(&bytes[entry_range.clone()])?;
             let cmp = entry_path.cmp(path);
             if cmp == Less {
                 let after_newline = entry_range.end + 1;
                 bytes = &bytes[after_newline..];
             } else if cmp == Greater {
                 bytes = &bytes[..entry_range.start];
             } else {
                 return Ok(Some(ManifestEntry::from_path_and_rest(
                     entry_path, rest,
                 )));
             }
         }
         Ok(None)
     }
 
     /// If there is at least one, return the byte range of an entry *excluding*
     /// the final newline.
     fn find_entry_near_middle_of(
         bytes: &[u8],
     ) -> Result<Option<std::ops::Range<usize>>, HgError> {
         let len = bytes.len();
         if len > 0 {
             let middle = bytes.len() / 2;
             // Integer division rounds down, so `middle < len`.
             let (before, after) = bytes.split_at(middle);
             let is_newline = |&byte: &u8| byte == b'\n';
             let entry_start = match before.iter().rposition(is_newline) {
                 Some(i) => i + 1,
                 None => 0, // We choose the first entry in `bytes`
             };
             let entry_end = match after.iter().position(is_newline) {
                 Some(i) => {
                     // No `+ 1` here to exclude this newline from the range
                     middle + i
                 }
                 None => {
                     // In a well-formed manifest:
                     //
                     // * Since `len > 0`, `bytes` contains at least one entry
                     // * Every entry ends with a newline
                     // * Since `middle < len`, `after` contains at least the
                     //   newline at the end of the last entry of `bytes`.
                     //
                     // We didn’t find a newline, so this manifest is not
                     // well-formed.
                     return Err(HgError::corrupted(
                         "manifest entry without \\n delimiter",
                     ));
                 }
             };
             Ok(Some(entry_start..entry_end))
         } else {
             // len == 0
             Ok(None)
         }
     }
 }
 
 /// `Manifestlog` entry which knows how to interpret the `manifest` data bytes.
 #[derive(Debug)]
 pub struct ManifestEntry<'manifest> {
     pub path: &'manifest HgPath,
     pub hex_node_id: &'manifest [u8],
 
     /// `Some` values are b'x', b'l', or 't'
     pub flags: Option<u8>,
 }
 
 impl<'a> ManifestEntry<'a> {
     fn split_path(bytes: &[u8]) -> Result<(&[u8], &[u8]), HgError> {
         bytes.split_2(b'\0').ok_or_else(|| {
             HgError::corrupted("manifest entry without \\0 delimiter")
         })
     }
 
     fn from_path_and_rest(path: &'a [u8], rest: &'a [u8]) -> Self {
         let (hex_node_id, flags) = match rest.split_last() {
             Some((&b'x', rest)) => (rest, Some(b'x')),
             Some((&b'l', rest)) => (rest, Some(b'l')),
             Some((&b't', rest)) => (rest, Some(b't')),
             _ => (rest, None),
         };
         Self {
             path: HgPath::new(path),
             hex_node_id,
             flags,
         }
     }
 
     fn from_raw(bytes: &'a [u8]) -> Result<Self, HgError> {
         let (path, rest) = Self::split_path(bytes)?;
         Ok(Self::from_path_and_rest(path, rest))
     }
 
     pub fn node_id(&self) -> Result<Node, HgError> {
         Node::from_hex_for_repo(self.hex_node_id)
     }
 }

rust/hg-core/src/revlog/revlog.rs

@@ -1,421 +1,431
 use std::borrow::Cow;
 use std::io::Read;
 use std::ops::Deref;
 use std::path::Path;
 
 use flate2::read::ZlibDecoder;
 use micro_timer::timed;
 use sha1::{Digest, Sha1};
 use zstd;
 
 use super::index::Index;
 use super::node::{NodePrefix, NODE_BYTES_LENGTH, NULL_NODE};
 use super::nodemap;
 use super::nodemap::{NodeMap, NodeMapError};
 use super::nodemap_docket::NodeMapDocket;
 use super::patch;
 use crate::errors::HgError;
 use crate::repo::Repo;
 use crate::revlog::Revision;
 use crate::{Node, NULL_REVISION};
 
 #[derive(derive_more::From)]
 pub enum RevlogError {
     InvalidRevision,
     /// Working directory is not supported
     WDirUnsupported,
     /// Found more than one entry whose ID match the requested prefix
     AmbiguousPrefix,
     #[from]
     Other(HgError),
 }
 
 impl From<NodeMapError> for RevlogError {
     fn from(error: NodeMapError) -> Self {
         match error {
             NodeMapError::MultipleResults => RevlogError::AmbiguousPrefix,
             NodeMapError::RevisionNotInIndex(_) => RevlogError::corrupted(),
         }
     }
 }
 
 impl RevlogError {
     fn corrupted() -> Self {
         RevlogError::Other(HgError::corrupted("corrupted revlog"))
     }
 }
 
 /// Read only implementation of revlog.
 pub struct Revlog {
     /// When index and data are not interleaved: bytes of the revlog index.
     /// When index and data are interleaved: bytes of the revlog index and
     /// data.
     index: Index,
     /// When index and data are not interleaved: bytes of the revlog data
     data_bytes: Option<Box<dyn Deref<Target = [u8]> + Send>>,
     /// When present on disk: the persistent nodemap for this revlog
     nodemap: Option<nodemap::NodeTree>,
 }
 
 impl Revlog {
     /// Open a revlog index file.
     ///
     /// It will also open the associated data file if index and data are not
     /// interleaved.
     #[timed]
     pub fn open(
         repo: &Repo,
         index_path: impl AsRef<Path>,
         data_path: Option<&Path>,
     ) -> Result<Self, HgError> {
         let index_path = index_path.as_ref();
         let index = {
             match repo.store_vfs().mmap_open_opt(&index_path)? {
                 None => Index::new(Box::new(vec![])),
                 Some(index_mmap) => {
                     let index = Index::new(Box::new(index_mmap))?;
                     Ok(index)
                 }
             }
         }?;
 
         let default_data_path = index_path.with_extension("d");
 
         // type annotation required
         // won't recognize Mmap as Deref<Target = [u8]>
         let data_bytes: Option<Box<dyn Deref<Target = [u8]> + Send>> =
             if index.is_inline() {
                 None
             } else {
                 let data_path = data_path.unwrap_or(&default_data_path);
                 let data_mmap = repo.store_vfs().mmap_open(data_path)?;
                 Some(Box::new(data_mmap))
             };
 
         let nodemap = if index.is_inline() {
             None
         } else {
             NodeMapDocket::read_from_file(repo, index_path)?.map(
                 |(docket, data)| {
                     nodemap::NodeTree::load_bytes(
                         Box::new(data),
                         docket.data_length,
                     )
                 },
             )
         };
 
         Ok(Revlog {
             index,
             data_bytes,
             nodemap,
         })
     }
 
     /// Return number of entries of the `Revlog`.
     pub fn len(&self) -> usize {
         self.index.len()
     }
 
     /// Returns `true` if the `Revlog` has zero `entries`.
     pub fn is_empty(&self) -> bool {
         self.index.is_empty()
     }
 
     /// Returns the node ID for the given revision number, if it exists in this
     /// revlog
     pub fn node_from_rev(&self, rev: Revision) -> Option<&Node> {
         if rev == NULL_REVISION {
             return Some(&NULL_NODE);
         }
         Some(self.index.get_entry(rev)?.hash())
     }
 
     /// Return the revision number for the given node ID, if it exists in this
     /// revlog
     #[timed]
     pub fn rev_from_node(
         &self,
         node: NodePrefix,
     ) -> Result<Revision, RevlogError> {
         if node.is_prefix_of(&NULL_NODE) {
             return Ok(NULL_REVISION);
         }
 
         if let Some(nodemap) = &self.nodemap {
             return nodemap
                 .find_bin(&self.index, node)?
                 .ok_or(RevlogError::InvalidRevision);
         }
 
         // Fallback to linear scan when a persistent nodemap is not present.
         // This happens when the persistent-nodemap experimental feature is not
         // enabled, or for small revlogs.
         //
         // TODO: consider building a non-persistent nodemap in memory to
         // optimize these cases.
         let mut found_by_prefix = None;
         for rev in (0..self.len() as Revision).rev() {
             let index_entry =
                 self.index.get_entry(rev).ok_or(HgError::corrupted(
                     "revlog references a revision not in the index",
                 ))?;
             if node == *index_entry.hash() {
                 return Ok(rev);
             }
             if node.is_prefix_of(index_entry.hash()) {
                 if found_by_prefix.is_some() {
                     return Err(RevlogError::AmbiguousPrefix);
                 }
                 found_by_prefix = Some(rev)
             }
         }
         found_by_prefix.ok_or(RevlogError::InvalidRevision)
     }
 
     /// Returns whether the given revision exists in this revlog.
     pub fn has_rev(&self, rev: Revision) -> bool {
         self.index.get_entry(rev).is_some()
     }
 
     /// Return the full data associated to a revision.
     ///
     /// All entries required to build the final data out of deltas will be
     /// retrieved as needed, and the deltas will be applied to the inital
     /// snapshot to rebuild the final data.
     #[timed]
-    pub fn get_rev_data(&self, rev: Revision) -> Result<Vec<u8>, RevlogError> {
+    pub fn get_rev_data(
+        &self,
+        rev: Revision,
+    ) -> Result<Cow<[u8]>, RevlogError> {
         if rev == NULL_REVISION {
-            return Ok(vec![]);
+            return Ok(Cow::Borrowed(&[]));
         };
-        // Todo return -> Cow
-        let mut entry = self.get_entry(rev)?;
-        let mut delta_chain = vec![];
-
-        // The meaning of `base_rev_or_base_of_delta_chain` depends on
-        // generaldelta. See the doc on `ENTRY_DELTA_BASE` in
-        // `mercurial/revlogutils/constants.py` and the code in
-        // [_chaininfo] and in [index_deltachain].
-        let uses_generaldelta = self.index.uses_generaldelta();
-        while let Some(base_rev) = entry.base_rev_or_base_of_delta_chain {
-            let base_rev = if uses_generaldelta {
-                base_rev
-            } else {
-                entry.rev - 1
-            };
-            delta_chain.push(entry);
-            entry = self.get_entry_internal(base_rev)?;
-        }
-
-        // TODO do not look twice in the index
-        let index_entry = self
-            .index
-            .get_entry(rev)
-            .ok_or(RevlogError::InvalidRevision)?;
-
-        let data: Vec<u8> = if delta_chain.is_empty() {
-            entry.data_chunk()?.into()
-        } else {
-            Revlog::build_data_from_deltas(entry, &delta_chain)?
-        };
-
-        if self.check_hash(
-            index_entry.p1(),
-            index_entry.p2(),
-            index_entry.hash().as_bytes(),
-            &data,
-        ) {
-            Ok(data)
-        } else {
-            Err(RevlogError::corrupted())
-        }
+        self.get_entry(rev)?.data()
     }
 
     /// Check the hash of some given data against the recorded hash.
     pub fn check_hash(
         &self,
         p1: Revision,
         p2: Revision,
         expected: &[u8],
         data: &[u8],
     ) -> bool {
         let e1 = self.index.get_entry(p1);
         let h1 = match e1 {
             Some(ref entry) => entry.hash(),
             None => &NULL_NODE,
         };
         let e2 = self.index.get_entry(p2);
         let h2 = match e2 {
             Some(ref entry) => entry.hash(),
             None => &NULL_NODE,
         };
 
         &hash(data, h1.as_bytes(), h2.as_bytes()) == expected
     }
 
     /// Build the full data of a revision out its snapshot
     /// and its deltas.
     #[timed]
     fn build_data_from_deltas(
         snapshot: RevlogEntry,
         deltas: &[RevlogEntry],
     ) -> Result<Vec<u8>, RevlogError> {
         let snapshot = snapshot.data_chunk()?;
         let deltas = deltas
             .iter()
             .rev()
             .map(RevlogEntry::data_chunk)
             .collect::<Result<Vec<Cow<'_, [u8]>>, RevlogError>>()?;
         let patches: Vec<_> =
             deltas.iter().map(|d| patch::PatchList::new(d)).collect();
         let patch = patch::fold_patch_lists(&patches);
         Ok(patch.apply(&snapshot))
     }
 
     /// Return the revlog data.
     fn data(&self) -> &[u8] {
         match self.data_bytes {
             Some(ref data_bytes) => &data_bytes,
             None => panic!(
                 "forgot to load the data or trying to access inline data"
             ),
         }
     }
 
     /// Get an entry of the revlog.
     fn get_entry(&self, rev: Revision) -> Result<RevlogEntry, RevlogError> {
         let index_entry = self
             .index
             .get_entry(rev)
             .ok_or(RevlogError::InvalidRevision)?;
         let start = index_entry.offset();
         let end = start + index_entry.compressed_len();
         let data = if self.index.is_inline() {
             self.index.data(start, end)
         } else {
             &self.data()[start..end]
         };
         let entry = RevlogEntry {
+            revlog: self,
             rev,
             bytes: data,
             compressed_len: index_entry.compressed_len(),
             uncompressed_len: index_entry.uncompressed_len(),
             base_rev_or_base_of_delta_chain: if index_entry
                 .base_revision_or_base_of_delta_chain()
                 == rev
             {
                 None
             } else {
                 Some(index_entry.base_revision_or_base_of_delta_chain())
             },
         };
         Ok(entry)
     }
 
     /// when resolving internal references within revlog, any errors
     /// should be reported as corruption, instead of e.g. "invalid revision"
     fn get_entry_internal(
         &self,
         rev: Revision,
     ) -> Result<RevlogEntry, RevlogError> {
         return self.get_entry(rev).map_err(|_| RevlogError::corrupted());
     }
 }
 
 /// The revlog entry's bytes and the necessary informations to extract
 /// the entry's data.
-#[derive(Debug)]
+#[derive(Clone)]
 pub struct RevlogEntry<'a> {
+    revlog: &'a Revlog,
     rev: Revision,
     bytes: &'a [u8],
     compressed_len: usize,
     uncompressed_len: usize,
     base_rev_or_base_of_delta_chain: Option<Revision>,
 }
 
 impl<'a> RevlogEntry<'a> {
     pub fn revision(&self) -> Revision {
         self.rev
     }
 
+    /// The data for this entry, after resolving deltas if any.
+    pub fn data(&self) -> Result<Cow<'a, [u8]>, RevlogError> {
+        let mut entry = self.clone();
+        let mut delta_chain = vec![];
+
+        // The meaning of `base_rev_or_base_of_delta_chain` depends on
+        // generaldelta. See the doc on `ENTRY_DELTA_BASE` in
+        // `mercurial/revlogutils/constants.py` and the code in
+        // [_chaininfo] and in [index_deltachain].
+        let uses_generaldelta = self.revlog.index.uses_generaldelta();
+        while let Some(base_rev) = entry.base_rev_or_base_of_delta_chain {
+            let base_rev = if uses_generaldelta {
+                base_rev
+            } else {
+                entry.rev - 1
+            };
+            delta_chain.push(entry);
+            entry = self.revlog.get_entry_internal(base_rev)?;
+        }
+
+        // TODO do not look twice in the index
+        let index_entry = self
+            .revlog
+            .index
+            .get_entry(self.rev)
+            .ok_or(RevlogError::InvalidRevision)?;
+
+        let data = if delta_chain.is_empty() {
+            entry.data_chunk()?
+        } else {
+            Revlog::build_data_from_deltas(entry, &delta_chain)?.into()
+        };
+
+        if self.revlog.check_hash(
+            index_entry.p1(),
+            index_entry.p2(),
+            index_entry.hash().as_bytes(),
+            &data,
+        ) {
+            Ok(data)
+        } else {
+            Err(RevlogError::corrupted())
+        }
+    }
+
     /// Extract the data contained in the entry.
     /// This may be a delta. (See `is_delta`.)
-    fn data_chunk(&self) -> Result<Cow<'_, [u8]>, RevlogError> {
+    fn data_chunk(&self) -> Result<Cow<'a, [u8]>, RevlogError> {
         if self.bytes.is_empty() {
             return Ok(Cow::Borrowed(&[]));
         }
         match self.bytes[0] {
             // Revision data is the entirety of the entry, including this
             // header.
             b'\0' => Ok(Cow::Borrowed(self.bytes)),
             // Raw revision data follows.
             b'u' => Ok(Cow::Borrowed(&self.bytes[1..])),
             // zlib (RFC 1950) data.
             b'x' => Ok(Cow::Owned(self.uncompressed_zlib_data()?)),
             // zstd data.
             b'\x28' => Ok(Cow::Owned(self.uncompressed_zstd_data()?)),
             // A proper new format should have had a repo/store requirement.
             _format_type => Err(RevlogError::corrupted()),
         }
     }
 
     fn uncompressed_zlib_data(&self) -> Result<Vec<u8>, RevlogError> {
         let mut decoder = ZlibDecoder::new(self.bytes);
         if self.is_delta() {
             let mut buf = Vec::with_capacity(self.compressed_len);
             decoder
                 .read_to_end(&mut buf)
                 .map_err(|_| RevlogError::corrupted())?;
             Ok(buf)
         } else {
             let mut buf = vec![0; self.uncompressed_len];
             decoder
                 .read_exact(&mut buf)
                 .map_err(|_| RevlogError::corrupted())?;
             Ok(buf)
         }
     }
 
     fn uncompressed_zstd_data(&self) -> Result<Vec<u8>, RevlogError> {
         if self.is_delta() {
             let mut buf = Vec::with_capacity(self.compressed_len);
             zstd::stream::copy_decode(self.bytes, &mut buf)
                 .map_err(|_| RevlogError::corrupted())?;
             Ok(buf)
         } else {
             let mut buf = vec![0; self.uncompressed_len];
             let len = zstd::block::decompress_to_buffer(self.bytes, &mut buf)
                 .map_err(|_| RevlogError::corrupted())?;
             if len != self.uncompressed_len {
                 Err(RevlogError::corrupted())
             } else {
                 Ok(buf)
             }
         }
     }
 
     /// Tell if the entry is a snapshot or a delta
     /// (influences on decompression).
     fn is_delta(&self) -> bool {
         self.base_rev_or_base_of_delta_chain.is_some()
     }
 }
 
 /// Calculate the hash of a revision given its data and its parents.
 fn hash(
     data: &[u8],
     p1_hash: &[u8],
     p2_hash: &[u8],
 ) -> [u8; NODE_BYTES_LENGTH] {
     let mut hasher = Sha1::new();
     let (a, b) = (p1_hash, p2_hash);
     if a > b {
         hasher.update(b);
         hasher.update(a);
     } else {
         hasher.update(a);
         hasher.update(b);
     }
     hasher.update(data);
     *hasher.finalize().as_ref()
 }