Merge pull request #11077 from hercules-ci/support-hardlinks-in-tarballs

Support hardlinks in tarballs

Author: Ericson2314

src/libfetchers/git-utils.cc

@@ -115,10 +115,10 @@ git_oid hashToOID(const Hash & hash)
     return oid;
 }
 
-Object lookupObject(git_repository * repo, const git_oid & oid)
+Object lookupObject(git_repository * repo, const git_oid & oid, git_object_t type = GIT_OBJECT_ANY)
 {
     Object obj;
-    if (git_object_lookup(Setter(obj), repo, &oid, GIT_OBJECT_ANY)) {
+    if (git_object_lookup(Setter(obj), repo, &oid, type)) {
         auto err = git_error_last();
         throw Error("getting Git object '%s': %s", oid, err->message);
     }
@@ -909,6 +909,61 @@ struct GitFileSystemObjectSinkImpl : GitFileSystemObjectSink
         addToTree(*pathComponents.rbegin(), oid, GIT_FILEMODE_LINK);
     }
 
+    void createHardlink(const CanonPath & path, const CanonPath & target) override
+    {
+        std::vector<std::string> pathComponents;
+        for (auto & c : path)
+            pathComponents.emplace_back(c);
+
+        if (!prepareDirs(pathComponents, false)) return;
+
+        // We can't just look up the path from the start of the root, since
+        // some parent directories may not have finished yet, so we compute
+        // a relative path that helps us find the right git_tree_builder or object.
+        auto relTarget = CanonPath(path).parent()->makeRelative(target);
+
+        auto dir = pendingDirs.rbegin();
+
+        // For each ../ component at the start, go up one directory.
+        // CanonPath::makeRelative() always puts all .. elements at the start,
+        // so they're all handled by this loop:
+        std::string_view relTargetLeft(relTarget);
+        while (hasPrefix(relTargetLeft, "../")) {
+            if (dir == pendingDirs.rend())
+                throw Error("invalid hard link target '%s' for path '%s'", target, path);
+            ++dir;
+            relTargetLeft = relTargetLeft.substr(3);
+        }
+        if (dir == pendingDirs.rend())
+            throw Error("invalid hard link target '%s' for path '%s'", target, path);
+
+        // Look up the remainder of the target, starting at the
+        // top-most `git_treebuilder`.
+        std::variant<git_treebuilder *, git_oid> curDir{dir->builder.get()};
+        Object tree; // needed to keep `entry` alive
+        const git_tree_entry * entry = nullptr;
+
+        for (auto & c : CanonPath(relTargetLeft)) {
+            if (auto builder = std::get_if<git_treebuilder *>(&curDir)) {
+                assert(*builder);
+                if (!(entry = git_treebuilder_get(*builder, std::string(c).c_str())))
+                    throw Error("cannot find hard link target '%s' for path '%s'", target, path);
+                curDir = *git_tree_entry_id(entry);
+            } else if (auto oid = std::get_if<git_oid>(&curDir)) {
+                tree = lookupObject(*repo, *oid, GIT_OBJECT_TREE);
+                if (!(entry = git_tree_entry_byname((const git_tree *) &*tree, std::string(c).c_str())))
+                    throw Error("cannot find hard link target '%s' for path '%s'", target, path);
+                curDir = *git_tree_entry_id(entry);
+            }
+        }
+
+        assert(entry);
+
+        addToTree(*pathComponents.rbegin(),
+            *git_tree_entry_id(entry),
+            git_tree_entry_filemode(entry));
+    }
+
     Hash sync() override {
         updateBuilders({});
 

src/libfetchers/git-utils.hh

@@ -7,7 +7,7 @@ namespace nix {
 
 namespace fetchers { struct PublicKey; }
 
-struct GitFileSystemObjectSink : FileSystemObjectSink
+struct GitFileSystemObjectSink : ExtendedFileSystemObjectSink
 {
     /**
      * Flush builder and return a final Git hash.

src/libutil/fs-sink.hh

@@ -41,6 +41,19 @@ struct FileSystemObjectSink
     virtual void createSymlink(const CanonPath & path, const std::string & target) = 0;
 };
 
+/**
+ * An extension of `FileSystemObjectSink` that supports file types
+ * that are not supported by Nix's FSO model.
+ */
+struct ExtendedFileSystemObjectSink : virtual FileSystemObjectSink
+{
+    /**
+     * Create a hard link. The target must be the path of a previously
+     * encountered file relative to the root of the FSO.
+     */
+    virtual void createHardlink(const CanonPath & path, const CanonPath & target) = 0;
+};
+
 /**
  * Recursively copy file system objects from the source into the sink.
  */

src/libutil/tarfile.cc

@@ -174,7 +174,7 @@ void unpackTarfile(const Path & tarFile, const Path & destDir)
     extract_archive(archive, destDir);
 }
 
-time_t unpackTarfileToSink(TarArchive & archive, FileSystemObjectSink & parseSink)
+time_t unpackTarfileToSink(TarArchive & archive, ExtendedFileSystemObjectSink & parseSink)
 {
     time_t lastModified = 0;
 
@@ -195,7 +195,12 @@ time_t unpackTarfileToSink(TarArchive & archive, FileSystemObjectSink & parseSin
 
         lastModified = std::max(lastModified, archive_entry_mtime(entry));
 
-        switch (archive_entry_filetype(entry)) {
+        if (auto target = archive_entry_hardlink(entry)) {
+            parseSink.createHardlink(cpath, CanonPath(target));
+            continue;
+        }
+
+        switch (auto type = archive_entry_filetype(entry)) {
 
         case AE_IFDIR:
             parseSink.createDirectory(cpath);
@@ -232,7 +237,7 @@ time_t unpackTarfileToSink(TarArchive & archive, FileSystemObjectSink & parseSin
         }
 
         default:
-            throw Error("file '%s' in tarball has unsupported file type", path);
+            throw Error("file '%s' in tarball has unsupported file type %d", path, type);
         }
     }
 

src/libutil/tarfile.hh

@@ -41,6 +41,6 @@ void unpackTarfile(Source & source, const Path & destDir);
 
 void unpackTarfile(const Path & tarFile, const Path & destDir);
 
-time_t unpackTarfileToSink(TarArchive & archive, FileSystemObjectSink & parseSink);
+time_t unpackTarfileToSink(TarArchive & archive, ExtendedFileSystemObjectSink & parseSink);
 
 }

tests/functional/tarball.sh

@@ -71,3 +71,15 @@ test_tarball() {
 test_tarball '' cat
 test_tarball .xz xz
 test_tarball .gz gzip
+
+# Test hard links.
+# All entries in tree.tar.gz refer to the same file, and all have the same inode when unpacked by GNU tar.
+# We don't preserve the hard links, because that's an optimization we think is not worth the complexity,
+# so we only make sure that the contents are copied correctly.
+path="$(nix flake prefetch --json "tarball+file://$(pwd)/tree.tar.gz" | jq -r .storePath)"
+[[ $(cat "$path/a/b/foo") = bar ]]
+[[ $(cat "$path/a/b/xyzzy") = bar ]]
+[[ $(cat "$path/a/yyy") = bar ]]
+[[ $(cat "$path/a/zzz") = bar ]]
+[[ $(cat "$path/c/aap") = bar ]]
+[[ $(cat "$path/fnord") = bar ]]

tests/functional/tree.tar.gz

@@ -0,0 +1,112 @@
+#include "git-utils.hh"
+#include "file-system.hh"
+#include "gmock/gmock.h"
+#include <git2/global.h>
+#include <git2/repository.h>
+#include <git2/types.h>
+#include <gtest/gtest.h>
+#include "fs-sink.hh"
+#include "serialise.hh"
+
+namespace nix {
+
+class GitUtilsTest : public ::testing::Test
+{
+    // We use a single repository for all tests.
+    Path tmpDir;
+    std::unique_ptr<AutoDelete> delTmpDir;
+
+public:
+    void SetUp() override
+    {
+        tmpDir = createTempDir();
+        delTmpDir = std::make_unique<AutoDelete>(tmpDir, true);
+
+        // Create the repo with libgit2
+        git_libgit2_init();
+        git_repository * repo = nullptr;
+        auto r = git_repository_init(&repo, tmpDir.c_str(), 0);
+        ASSERT_EQ(r, 0);
+        git_repository_free(repo);
+    }
+
+    void TearDown() override
+    {
+        // Destroy the AutoDelete, triggering removal
+        // not AutoDelete::reset(), which would cancel the deletion.
+        delTmpDir.reset();
+    }
+
+    ref<GitRepo> openRepo()
+    {
+        return GitRepo::openRepo(tmpDir, true, false);
+    }
+};
+
+void writeString(CreateRegularFileSink & fileSink, std::string contents, bool executable)
+{
+    if (executable)
+        fileSink.isExecutable();
+    fileSink.preallocateContents(contents.size());
+    fileSink(contents);
+}
+
+TEST_F(GitUtilsTest, sink_basic)
+{
+    auto repo = openRepo();
+    auto sink = repo->getFileSystemObjectSink();
+
+    // TODO/Question: It seems a little odd that we use the tarball-like convention of requiring a top-level directory
+    // here
+    //                The sync method does not document this behavior, should probably renamed because it's not very
+    //                general, and I can't imagine that "non-conventional" archives or any other source to be handled by
+    //                this sink.
+
+    sink->createDirectory(CanonPath("foo-1.1"));
+
+    sink->createRegularFile(CanonPath("foo-1.1/hello"), [](CreateRegularFileSink & fileSink) {
+        writeString(fileSink, "hello world", false);
+    });
+    sink->createRegularFile(CanonPath("foo-1.1/bye"), [](CreateRegularFileSink & fileSink) {
+        writeString(fileSink, "thanks for all the fish", false);
+    });
+    sink->createSymlink(CanonPath("foo-1.1/bye-link"), "bye");
+    sink->createDirectory(CanonPath("foo-1.1/empty"));
+    sink->createDirectory(CanonPath("foo-1.1/links"));
+    sink->createHardlink(CanonPath("foo-1.1/links/foo"), CanonPath("foo-1.1/hello"));
+
+    // sink->createHardlink("foo-1.1/links/foo-2", CanonPath("foo-1.1/hello"));
+
+    auto result = sink->sync();
+    auto accessor = repo->getAccessor(result, false);
+    auto entries = accessor->readDirectory(CanonPath::root);
+    ASSERT_EQ(entries.size(), 5);
+    ASSERT_EQ(accessor->readFile(CanonPath("hello")), "hello world");
+    ASSERT_EQ(accessor->readFile(CanonPath("bye")), "thanks for all the fish");
+    ASSERT_EQ(accessor->readLink(CanonPath("bye-link")), "bye");
+    ASSERT_EQ(accessor->readDirectory(CanonPath("empty")).size(), 0);
+    ASSERT_EQ(accessor->readFile(CanonPath("links/foo")), "hello world");
+};
+
+TEST_F(GitUtilsTest, sink_hardlink)
+{
+    auto repo = openRepo();
+    auto sink = repo->getFileSystemObjectSink();
+
+    sink->createDirectory(CanonPath("foo-1.1"));
+
+    sink->createRegularFile(CanonPath("foo-1.1/hello"), [](CreateRegularFileSink & fileSink) {
+        writeString(fileSink, "hello world", false);
+    });
+
+    try {
+        sink->createHardlink(CanonPath("foo-1.1/link"), CanonPath("hello"));
+        FAIL() << "Expected an exception";
+    } catch (const nix::Error & e) {
+        ASSERT_THAT(e.msg(), testing::HasSubstr("invalid hard link target"));
+        ASSERT_THAT(e.msg(), testing::HasSubstr("/hello"));
+        ASSERT_THAT(e.msg(), testing::HasSubstr("foo-1.1/link"));
+    }
+};
+
+} // namespace nix

tests/unit/libfetchers/git-utils.cc

@@ -29,7 +29,7 @@ libfetchers-tests_LIBS = \
     libstore-test-support libutil-test-support \
     libfetchers libstore libutil
 
-libfetchers-tests_LDFLAGS := -lrapidcheck $(GTEST_LIBS)
+libfetchers-tests_LDFLAGS := -lrapidcheck $(GTEST_LIBS) $(LIBGIT2_LIBS)
 
 ifdef HOST_WINDOWS
   # Increase the default reserved stack size to 65 MB so Nix doesn't run out of space

tests/unit/libfetchers/local.mk

@@ -0,0 +1,34 @@
+#include <iostream>
+#include "tracing-file-system-object-sink.hh"
+
+namespace nix::test {
+
+void TracingFileSystemObjectSink::createDirectory(const CanonPath & path)
+{
+    std::cerr << "createDirectory(" << path << ")\n";
+    sink.createDirectory(path);
+}
+
+void TracingFileSystemObjectSink::createRegularFile(
+    const CanonPath & path, std::function<void(CreateRegularFileSink &)> fn)
+{
+    std::cerr << "createRegularFile(" << path << ")\n";
+    sink.createRegularFile(path, [&](CreateRegularFileSink & crf) {
+        // We could wrap this and trace about the chunks of data and such
+        fn(crf);
+    });
+}
+
+void TracingFileSystemObjectSink::createSymlink(const CanonPath & path, const std::string & target)
+{
+    std::cerr << "createSymlink(" << path << ", target: " << target << ")\n";
+    sink.createSymlink(path, target);
+}
+
+void TracingExtendedFileSystemObjectSink::createHardlink(const CanonPath & path, const CanonPath & target)
+{
+    std::cerr << "createHardlink(" << path << ", target: " << target << ")\n";
+    sink.createHardlink(path, target);
+}
+
+} // namespace nix::test