Git hashing for Git Fetching

Now that we have Git hashing of (the FSOs of) store objects, we should
teach the git fetcher how to use it. It can make especially good use of
it because then we can use a single hash for both identifying what to
fetch, and validating the download.

Part of RFC 133

Extracted from our old IPFS branches.

Co-Authored-By: Matthew Bauer <mjbauer95@gmail.com>
Co-Authored-By: Carlo Nucera <carlo.nucera@protonmail.com>
Co-authored-by: Robert Hensing <roberth@users.noreply.github.com>

src/libexpr/primops/fetchTree.cc

@@ -30,9 +30,13 @@ void emitTreeAttrs(
 
     // FIXME: support arbitrary input attributes.
 
-    auto narHash = input.getNarHash();
-    assert(narHash);
-    attrs.alloc("narHash").mkString(narHash->to_string(SRI, true));
+    if (auto narHash = input.getNarHash()) {
+        attrs.alloc("narHash").mkString(narHash->to_string(SRI, true));
+    } else if (auto treeHash = input.getTreeHash()) {
+        attrs.alloc("treeHash").mkString(treeHash->to_string(SRI, true));
+    } else
+        /* Must have either tree hash or NAR hash */
+        assert(false);
 
     if (input.getType() == "git")
         attrs.alloc("submodules").mkBool(
@@ -50,6 +54,10 @@ void emitTreeAttrs(
             attrs.alloc("shortRev").mkString(emptyHash.gitShortRev());
         }
 
+        if (auto treeHash = input.getTreeHash()) {
+            attrs.alloc("treeHash").mkString(treeHash->gitRev());
+        }
+
         if (auto revCount = input.getRevCount())
             attrs.alloc("revCount").mkInt(*revCount);
         else if (emptyRevFallback)

src/libfetchers/fetchers.cc

@@ -25,6 +25,8 @@ static void fixupInput(Input & input)
     input.getRef();
     if (input.getRev())
         input.locked = true;
+    if (input.getTreeHash())
+        input.locked = true;
     input.getRevCount();
     input.getLastModified();
     if (input.getNarHash())
@@ -91,7 +93,7 @@ Attrs Input::toAttrs() const
 
 bool Input::hasAllInfo() const
 {
-    return getNarHash() && scheme && scheme->hasAllInfo(*this);
+    return scheme && scheme->hasAllInfo(*this);
 }
 
 bool Input::operator ==(const Input & other) const
@@ -215,14 +217,19 @@ std::string Input::getName() const
 
 StorePath Input::computeStorePath(Store & store) const
 {
-    auto narHash = getNarHash();
-    if (!narHash)
-        throw Error("cannot compute store path for unlocked input '%s'", to_string());
-    return store.makeFixedOutputPath(getName(), FixedOutputInfo {
-        .method = FileIngestionMethod::Recursive,
-        .hash = *narHash,
-        .references = {},
-    });
+    if (auto treeHash = getTreeHash())
+        return store.makeFixedOutputPath(getName(), FixedOutputInfo {
+            .method = FileIngestionMethod::Git,
+            .hash = *treeHash,
+            .references = {},
+        });
+    if (auto narHash = getNarHash())
+        return store.makeFixedOutputPath(getName(), FixedOutputInfo {
+            .method = FileIngestionMethod::Recursive,
+            .hash = *narHash,
+            .references = {},
+        });
+    throw Error("cannot compute store path for unlocked input '%s'", to_string());
 }
 
 std::string Input::getType() const
@@ -265,6 +272,15 @@ std::optional<Hash> Input::getRev() const
     return hash;
 }
 
+std::optional<Hash> Input::getTreeHash() const
+{
+    if (auto s = maybeGetStrAttr(attrs, "treeHash")) {
+        experimentalFeatureSettings.require(Xp::GitHashing);
+        return Hash::parseAny(*s, htSHA1);
+    }
+    return {};
+}
+
 std::optional<uint64_t> Input::getRevCount() const
 {
     if (auto n = maybeGetIntAttr(attrs, "revCount"))

src/libfetchers/fetchers.hh

@@ -119,6 +119,7 @@ public:
     std::optional<Hash> getNarHash() const;
     std::optional<std::string> getRef() const;
     std::optional<Hash> getRev() const;
+    std::optional<Hash> getTreeHash() const;
     std::optional<uint64_t> getRevCount() const;
     std::optional<time_t> getLastModified() const;
 };