Merge pull request #9485 from edolstra/tarball-cache

Add a Git-based content-addressed tarball cache

src/libfetchers/attrs.cc

@@ -104,4 +104,9 @@ std::map<std::string, std::string> attrsToQuery(const Attrs & attrs)
     return query;
 }
 
+Hash getRevAttr(const Attrs & attrs, const std::string & name)
+{
+    return Hash::parseAny(getStrAttr(attrs, name), HashAlgorithm::SHA1);
+}
+
 }

src/libfetchers/attrs.hh

@@ -39,4 +39,6 @@ bool getBoolAttr(const Attrs & attrs, const std::string & name);
 
 std::map<std::string, std::string> attrsToQuery(const Attrs & attrs);
 
+Hash getRevAttr(const Attrs & attrs, const std::string & name);
+
 }

src/libfetchers/git-utils.cc

@@ -6,8 +6,8 @@
 #include "finally.hh"
 #include "processes.hh"
 #include "signals.hh"
-
+#include "users.hh"
-#include <boost/core/span.hpp>
+#include "fs-sink.hh"
 
 #include <git2/attr.h>
 #include <git2/blob.h>
@@ -28,6 +28,7 @@
 #include <unordered_set>
 #include <queue>
 #include <regex>
+#include <span>
 
 namespace std {
 
@@ -356,6 +357,8 @@ struct GitRepoImpl : GitRepo, std::enable_shared_from_this<GitRepoImpl>
 
     ref<InputAccessor> getAccessor(const WorkdirInfo & wd, bool exportIgnore, MakeNotAllowedError e) override;
 
+    ref<GitFileSystemObjectSink> getFileSystemObjectSink() override;
+
     static int sidebandProgressCallback(const char * str, int len, void * payload)
     {
         auto act = (Activity *) payload;
@@ -770,6 +773,154 @@ struct GitExportIgnoreInputAccessor : CachingFilteringInputAccessor {
 
 };
 
+struct GitFileSystemObjectSinkImpl : GitFileSystemObjectSink
+{
+    ref<GitRepoImpl> repo;
+
+    struct PendingDir
+    {
+        std::string name;
+        TreeBuilder builder;
+    };
+
+    std::vector<PendingDir> pendingDirs;
+
+    size_t componentsToStrip = 1;
+
+    void pushBuilder(std::string name)
+    {
+        git_treebuilder * b;
+        if (git_treebuilder_new(&b, *repo, nullptr))
+            throw Error("creating a tree builder: %s", git_error_last()->message);
+        pendingDirs.push_back({ .name = std::move(name), .builder = TreeBuilder(b) });
+    };
+
+    GitFileSystemObjectSinkImpl(ref<GitRepoImpl> repo) : repo(repo)
+    {
+        pushBuilder("");
+    }
+
+    std::pair<git_oid, std::string> popBuilder()
+    {
+        assert(!pendingDirs.empty());
+        auto pending = std::move(pendingDirs.back());
+        git_oid oid;
+        if (git_treebuilder_write(&oid, pending.builder.get()))
+            throw Error("creating a tree object: %s", git_error_last()->message);
+        pendingDirs.pop_back();
+        return {oid, pending.name};
+    };
+
+    void addToTree(const std::string & name, const git_oid & oid, git_filemode_t mode)
+    {
+        assert(!pendingDirs.empty());
+        auto & pending = pendingDirs.back();
+        if (git_treebuilder_insert(nullptr, pending.builder.get(), name.c_str(), &oid, mode))
+            throw Error("adding a file to a tree builder: %s", git_error_last()->message);
+    };
+
+    void updateBuilders(std::span<const std::string> names)
+    {
+        // Find the common prefix of pendingDirs and names.
+        size_t prefixLen = 0;
+        for (; prefixLen < names.size() && prefixLen + 1 < pendingDirs.size(); ++prefixLen)
+            if (names[prefixLen] != pendingDirs[prefixLen + 1].name)
+                break;
+
+        // Finish the builders that are not part of the common prefix.
+        for (auto n = pendingDirs.size(); n > prefixLen + 1; --n) {
+            auto [oid, name] = popBuilder();
+            addToTree(name, oid, GIT_FILEMODE_TREE);
+        }
+
+        // Create builders for the new directories.
+        for (auto n = prefixLen; n < names.size(); ++n)
+            pushBuilder(names[n]);
+    };
+
+    bool prepareDirs(const std::vector<std::string> & pathComponents, bool isDir)
+    {
+        std::span<const std::string> pathComponents2{pathComponents};
+
+        if (pathComponents2.size() <= componentsToStrip) return false;
+        pathComponents2 = pathComponents2.subspan(componentsToStrip);
+
+        updateBuilders(
+            isDir
+            ? pathComponents2
+            : pathComponents2.first(pathComponents2.size() - 1));
+
+        return true;
+    }
+
+    void createRegularFile(
+        const Path & path,
+        std::function<void(CreateRegularFileSink &)> func) override
+    {
+        auto pathComponents = tokenizeString<std::vector<std::string>>(path, "/");
+        if (!prepareDirs(pathComponents, false)) return;
+
+        git_writestream * stream = nullptr;
+        if (git_blob_create_from_stream(&stream, *repo, nullptr))
+            throw Error("creating a blob stream object: %s", git_error_last()->message);
+
+        struct CRF : CreateRegularFileSink {
+            const Path & path;
+            GitFileSystemObjectSinkImpl & back;
+            git_writestream * stream;
+            bool executable = false;
+            CRF(const Path & path, GitFileSystemObjectSinkImpl & back, git_writestream * stream)
+                : path(path), back(back), stream(stream)
+            {}
+            void operator () (std::string_view data) override
+            {
+                if (stream->write(stream, data.data(), data.size()))
+                    throw Error("writing a blob for tarball member '%s': %s", path, git_error_last()->message);
+            }
+            void isExecutable() override
+            {
+                executable = true;
+            }
+        } crf { path, *this, stream };
+        func(crf);
+
+        git_oid oid;
+        if (git_blob_create_from_stream_commit(&oid, stream))
+            throw Error("creating a blob object for tarball member '%s': %s", path, git_error_last()->message);
+
+        addToTree(*pathComponents.rbegin(), oid,
+            crf.executable
+            ? GIT_FILEMODE_BLOB_EXECUTABLE
+            : GIT_FILEMODE_BLOB);
+    }
+
+    void createDirectory(const Path & path) override
+    {
+        auto pathComponents = tokenizeString<std::vector<std::string>>(path, "/");
+        (void) prepareDirs(pathComponents, true);
+    }
+
+    void createSymlink(const Path & path, const std::string & target) override
+    {
+        auto pathComponents = tokenizeString<std::vector<std::string>>(path, "/");
+        if (!prepareDirs(pathComponents, false)) return;
+
+        git_oid oid;
+        if (git_blob_create_from_buffer(&oid, *repo, target.c_str(), target.size()))
+            throw Error("creating a blob object for tarball symlink member '%s': %s", path, git_error_last()->message);
+
+        addToTree(*pathComponents.rbegin(), oid, GIT_FILEMODE_LINK);
+    }
+
+    Hash sync() override {
+        updateBuilders({});
+
+        auto [oid, _name] = popBuilder();
+
+        return toHash(oid);
+    }
+};
+
 ref<GitInputAccessor> GitRepoImpl::getRawAccessor(const Hash & rev)
 {
     auto self = ref<GitRepoImpl>(shared_from_this());
@@ -804,6 +955,11 @@ ref<InputAccessor> GitRepoImpl::getAccessor(const WorkdirInfo & wd, bool exportI
     }
 }
 
+ref<GitFileSystemObjectSink> GitRepoImpl::getFileSystemObjectSink()
+{
+    return make_ref<GitFileSystemObjectSinkImpl>(ref<GitRepoImpl>(shared_from_this()));
+}
+
 std::vector<std::tuple<GitRepoImpl::Submodule, Hash>> GitRepoImpl::getSubmodules(const Hash & rev, bool exportIgnore)
 {
     /* Read the .gitmodules files from this revision. */
@@ -830,5 +986,11 @@ std::vector<std::tuple<GitRepoImpl::Submodule, Hash>> GitRepoImpl::getSubmodules
     return result;
 }
 
+ref<GitRepo> getTarballCache()
+{
+    static auto repoDir = std::filesystem::path(getCacheDir()) / "nix" / "tarball-cache";
+
+    return GitRepo::openRepo(repoDir, true, true);
+}
 
 }

src/libfetchers/git-utils.hh

@@ -2,11 +2,20 @@
 
 #include "filtering-input-accessor.hh"
 #include "input-accessor.hh"
+#include "fs-sink.hh"
 
 namespace nix {
 
 namespace fetchers { struct PublicKey; }
 
+struct GitFileSystemObjectSink : FileSystemObjectSink
+{
+    /**
+     * Flush builder and return a final Git hash.
+     */
+    virtual Hash sync() = 0;
+};
+
 struct GitRepo
 {
     virtual ~GitRepo()
@@ -64,18 +73,14 @@ struct GitRepo
         const std::string & url,
         const std::string & base) = 0;
 
-    struct TarballInfo
-    {
-        Hash treeHash;
-        time_t lastModified;
-    };
-
     virtual bool hasObject(const Hash & oid) = 0;
 
     virtual ref<InputAccessor> getAccessor(const Hash & rev, bool exportIgnore) = 0;
 
     virtual ref<InputAccessor> getAccessor(const WorkdirInfo & wd, bool exportIgnore, MakeNotAllowedError makeNotAllowedError) = 0;
 
+    virtual ref<GitFileSystemObjectSink> getFileSystemObjectSink() = 0;
+
     virtual void fetch(
         const std::string & url,
         const std::string & refspec,
@@ -90,4 +95,6 @@ struct GitRepo
         const std::vector<fetchers::PublicKey> & publicKeys) = 0;
 };
 
+ref<GitRepo> getTarballCache();
+
 }

src/libfetchers/github.cc

@@ -8,6 +8,8 @@
 #include "fetchers.hh"
 #include "fetch-settings.hh"
 #include "tarball.hh"
+#include "tarfile.hh"
+#include "git-utils.hh"
 
 #include <optional>
 #include <nlohmann/json.hpp>
@@ -180,49 +182,102 @@ struct GitArchiveInputScheme : InputScheme
         return headers;
     }
 
-    virtual Hash getRevFromRef(nix::ref<Store> store, const Input & input) const = 0;
+    struct RefInfo
+    {
+        Hash rev;
+        std::optional<Hash> treeHash;
+    };
+
+    virtual RefInfo getRevFromRef(nix::ref<Store> store, const Input & input) const = 0;
 
     virtual DownloadUrl getDownloadUrl(const Input & input) const = 0;
 
-    std::pair<StorePath, Input> fetch(ref<Store> store, const Input & _input) override
+    struct TarballInfo
     {
-        Input input(_input);
+        Hash treeHash;
+        time_t lastModified;
+    };
 
+    std::pair<Input, TarballInfo> downloadArchive(ref<Store> store, Input input) const
+    {
         if (!maybeGetStrAttr(input.attrs, "ref")) input.attrs.insert_or_assign("ref", "HEAD");
 
+        std::optional<Hash> upstreamTreeHash;
+
         auto rev = input.getRev();
-        if (!rev) rev = getRevFromRef(store, input);
+        if (!rev) {
+            auto refInfo = getRevFromRef(store, input);
+            rev = refInfo.rev;
+            upstreamTreeHash = refInfo.treeHash;
+            debug("HEAD revision for '%s' is %s", input.to_string(), refInfo.rev.gitRev());
+        }
 
         input.attrs.erase("ref");
         input.attrs.insert_or_assign("rev", rev->gitRev());
 
-        Attrs lockedAttrs({
+        auto cache = getCache();
-            {"type", "git-tarball"},
-            {"rev", rev->gitRev()},
-        });
 
-        if (auto res = getCache()->lookup(*store, lockedAttrs)) {
+        Attrs treeHashKey{{"_what", "gitRevToTreeHash"}, {"rev", rev->gitRev()}};
-            input.attrs.insert_or_assign("lastModified", getIntAttr(res->first, "lastModified"));
+        Attrs lastModifiedKey{{"_what", "gitRevToLastModified"}, {"rev", rev->gitRev()}};
-            return {std::move(res->second), input};
+
+        if (auto treeHashAttrs = cache->lookup(treeHashKey)) {
+            if (auto lastModifiedAttrs = cache->lookup(lastModifiedKey)) {
+                auto treeHash = getRevAttr(*treeHashAttrs, "treeHash");
+                auto lastModified = getIntAttr(*lastModifiedAttrs, "lastModified");
+                if (getTarballCache()->hasObject(treeHash))
+                    return {std::move(input), TarballInfo { .treeHash = treeHash, .lastModified = (time_t) lastModified }};
+                else
+                    debug("Git tree with hash '%s' has disappeared from the cache, refetching...", treeHash.gitRev());
+            }
         }
 
+        /* Stream the tarball into the tarball cache. */
         auto url = getDownloadUrl(input);
 
-        auto result = downloadTarball(store, url.url, input.getName(), true, url.headers);
+        auto source = sinkToSource([&](Sink & sink) {
+            FileTransferRequest req(url.url);
+            req.headers = url.headers;
+            getFileTransfer()->download(std::move(req), sink);
+        });
 
-        input.attrs.insert_or_assign("lastModified", uint64_t(result.lastModified));
+        TarArchive archive { *source };
+        auto parseSink = getTarballCache()->getFileSystemObjectSink();
+        auto lastModified = unpackTarfileToSink(archive, *parseSink);
 
-        getCache()->add(
+        TarballInfo tarballInfo {
-            *store,
+            .treeHash = parseSink->sync(),
-            lockedAttrs,
+            .lastModified = lastModified
+        };
+
+        cache->upsert(treeHashKey, Attrs{{"treeHash", tarballInfo.treeHash.gitRev()}});
+        cache->upsert(lastModifiedKey, Attrs{{"lastModified", (uint64_t) tarballInfo.lastModified}});
+
+        #if 0
+        if (upstreamTreeHash != tarballInfo.treeHash)
+            warn(
+                "Git tree hash mismatch for revision '%s' of '%s': "
+                "expected '%s', got '%s'. "
+                "This can happen if the Git repository uses submodules.",
+                rev->gitRev(), input.to_string(), upstreamTreeHash->gitRev(), tarballInfo.treeHash.gitRev());
+        #endif
+
+        return {std::move(input), tarballInfo};
+    }
+
+    std::pair<ref<InputAccessor>, Input> getAccessor(ref<Store> store, const Input & _input) const override
     {
-                {"rev", rev->gitRev()},
+        auto [input, tarballInfo] = downloadArchive(store, _input);
-                {"lastModified", uint64_t(result.lastModified)}
-            },
-            result.storePath,
-            true);
 
-        return {result.storePath, input};
+        input.attrs.insert_or_assign("treeHash", tarballInfo.treeHash.gitRev());
+        input.attrs.insert_or_assign("lastModified", uint64_t(tarballInfo.lastModified));
+
+        auto accessor = getTarballCache()->getAccessor(tarballInfo.treeHash, false);
+
+        accessor->setPathDisplay("«" + input.to_string() + "»");
+
+        accessor->fingerprint = input.getFingerprint(store);
+
+        return {accessor, input};
     }
 
     std::optional<ExperimentalFeature> experimentalFeature() const override
@@ -269,7 +324,7 @@ struct GitHubInputScheme : GitArchiveInputScheme
         return getStrAttr(input.attrs, "repo");
     }
 
-    Hash getRevFromRef(nix::ref<Store> store, const Input & input) const override
+    RefInfo getRevFromRef(nix::ref<Store> store, const Input & input) const override
     {
         auto host = getHost(input);
         auto url = fmt(
@@ -284,9 +339,11 @@ struct GitHubInputScheme : GitArchiveInputScheme
             readFile(
                 store->toRealPath(
                     downloadFile(store, url, "source", false, headers).storePath)));
-        auto rev = Hash::parseAny(std::string { json["sha"] }, HashAlgorithm::SHA1);
+
-        debug("HEAD revision for '%s' is %s", url, rev.gitRev());
+        return RefInfo {
-        return rev;
+            .rev = Hash::parseAny(std::string { json["sha"] }, HashAlgorithm::SHA1),
+            .treeHash = Hash::parseAny(std::string { json["commit"]["tree"]["sha"] }, HashAlgorithm::SHA1)
+        };
     }
 
     DownloadUrl getDownloadUrl(const Input & input) const override
@@ -343,7 +400,7 @@ struct GitLabInputScheme : GitArchiveInputScheme
         return std::make_pair(token.substr(0,fldsplit), token.substr(fldsplit+1));
     }
 
-    Hash getRevFromRef(nix::ref<Store> store, const Input & input) const override
+    RefInfo getRevFromRef(nix::ref<Store> store, const Input & input) const override
     {
         auto host = maybeGetStrAttr(input.attrs, "host").value_or("gitlab.com");
         // See rate limiting note below
@@ -356,9 +413,10 @@ struct GitLabInputScheme : GitArchiveInputScheme
             readFile(
                 store->toRealPath(
                     downloadFile(store, url, "source", false, headers).storePath)));
-        auto rev = Hash::parseAny(std::string(json[0]["id"]), HashAlgorithm::SHA1);
+
-        debug("HEAD revision for '%s' is %s", url, rev.gitRev());
+        return RefInfo {
-        return rev;
+            .rev = Hash::parseAny(std::string(json[0]["id"]), HashAlgorithm::SHA1)
+        };
     }
 
     DownloadUrl getDownloadUrl(const Input & input) const override
@@ -402,7 +460,7 @@ struct SourceHutInputScheme : GitArchiveInputScheme
         // Once it is implemented, however, should work as expected.
     }
 
-    Hash getRevFromRef(nix::ref<Store> store, const Input & input) const override
+    RefInfo getRevFromRef(nix::ref<Store> store, const Input & input) const override
     {
         // TODO: In the future, when the sourcehut graphql API is implemented for mercurial
         // and with anonymous access, this method should use it instead.
@@ -445,12 +503,12 @@ struct SourceHutInputScheme : GitArchiveInputScheme
                 id = parsedLine->target;
         }
 
-        if(!id)
+        if (!id)
             throw BadURL("in '%d', couldn't find ref '%d'", input.to_string(), ref);
 
-        auto rev = Hash::parseAny(*id, HashAlgorithm::SHA1);
+        return RefInfo {
-        debug("HEAD revision for '%s' is %s", fmt("%s/%s", base_url, ref), rev.gitRev());
+            .rev = Hash::parseAny(*id, HashAlgorithm::SHA1)
-        return rev;
+        };
     }
 
     DownloadUrl getDownloadUrl(const Input & input) const override

src/libutil/fs-sink.hh

@@ -26,6 +26,8 @@ struct CreateRegularFileSink : Sink
 
 struct FileSystemObjectSink
 {
+    virtual ~FileSystemObjectSink() = default;
+
     virtual void createDirectory(const Path & path) = 0;
 
     /**

src/libutil/tarfile.cc

@@ -132,4 +132,66 @@ void unpackTarfile(const Path & tarFile, const Path & destDir)
     extract_archive(archive, destDir);
 }
 
+time_t unpackTarfileToSink(TarArchive & archive, FileSystemObjectSink & parseSink)
+{
+    time_t lastModified = 0;
+
+    for (;;) {
+        // FIXME: merge with extract_archive
+        struct archive_entry * entry;
+        int r = archive_read_next_header(archive.archive, &entry);
+        if (r == ARCHIVE_EOF) break;
+        auto path = archive_entry_pathname(entry);
+        if (!path)
+            throw Error("cannot get archive member name: %s", archive_error_string(archive.archive));
+        if (r == ARCHIVE_WARN)
+            warn(archive_error_string(archive.archive));
+        else
+            archive.check(r);
+
+        lastModified = std::max(lastModified, archive_entry_mtime(entry));
+
+        switch (archive_entry_filetype(entry)) {
+
+        case AE_IFDIR:
+            parseSink.createDirectory(path);
+            break;
+
+        case AE_IFREG: {
+            parseSink.createRegularFile(path, [&](auto & crf) {
+                if (archive_entry_mode(entry) & S_IXUSR)
+                    crf.isExecutable();
+
+                while (true) {
+                    std::vector<unsigned char> buf(128 * 1024);
+                    auto n = archive_read_data(archive.archive, buf.data(), buf.size());
+                    if (n < 0)
+                        throw Error("cannot read file '%s' from tarball", path);
+                    if (n == 0) break;
+                    crf(std::string_view {
+                        (const char *) buf.data(),
+                        (size_t) n,
+                    });
+                }
+            });
+
+            break;
+        }
+
+        case AE_IFLNK: {
+            auto target = archive_entry_symlink(entry);
+
+            parseSink.createSymlink(path, target);
+
+            break;
+        }
+
+        default:
+            throw Error("file '%s' in tarball has unsupported file type", path);
+        }
+    }
+
+    return lastModified;
+}
+
 }

src/libutil/tarfile.hh

@@ -2,6 +2,7 @@
 ///@file
 
 #include "serialise.hh"
+#include "fs-sink.hh"
 #include <archive.h>
 
 namespace nix {
@@ -29,4 +30,6 @@ void unpackTarfile(Source & source, const Path & destDir);
 
 void unpackTarfile(const Path & tarFile, const Path & destDir);
 
+time_t unpackTarfileToSink(TarArchive & archive, FileSystemObjectSink & parseSink);
+
 }