status: never assume empty blobs are unencrypted

See comment in source code for rationale.

.github/workflows/release-linux.yml

@@ -1,46 +0,0 @@
-on:
-  release:
-    types: [published]
-name: Build Release Binary (Linux)
-jobs:
-  build:
-    name: Build Release Binary
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v2
-      - name: Install dependencies
-        run: sudo apt install libssl-dev
-      - name: Build binary
-        run: make
-      - name: Upload release artifact
-        uses: actions/upload-artifact@v3
-        with:
-          name: git-crypt-artifacts
-          path: git-crypt
-  upload:
-    name: Upload Release Binary
-    runs-on: ubuntu-latest
-    needs: build
-    permissions:
-      contents: write
-    steps:
-      - name: Download release artifact
-        uses: actions/download-artifact@v3
-        with:
-          name: git-crypt-artifacts
-      - name: Upload release asset
-        uses: actions/github-script@v3
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          script: |
-            const fs = require("fs").promises;
-            const { repo: { owner, repo }, sha } = context;
-            await github.repos.uploadReleaseAsset({
-              owner, repo,
-              release_id: ${{ github.event.release.id }},
-              name: 'git-crypt-${{ github.event.release.name }}-linux-x86_64',
-              data: await fs.readFile('git-crypt'),
-            });

.github/workflows/release-windows.yml

@@ -1,56 +0,0 @@
-on:
-  release:
-    types: [published]
-name: Build Release Binary (Windows)
-jobs:
-  build:
-    name: Build Release Binary
-    runs-on: windows-2022
-    permissions:
-      contents: read
-    steps:
-      - name: Checkout repository
-        uses: actions/checkout@v2
-      - name: Setup msys2
-        uses: msys2/setup-msys2@v2
-        with:
-          msystem: MINGW64
-          update: true
-          install: >-
-            base-devel
-            msys2-devel
-            mingw-w64-x86_64-toolchain
-            mingw-w64-x86_64-openssl
-            openssl-devel
-      - name: Build binary
-        shell: msys2 {0}
-        run: make LDFLAGS="-static-libstdc++ -static -lcrypto -lws2_32"
-      - name: Upload release artifact
-        uses: actions/upload-artifact@v3
-        with:
-          name: git-crypt-artifacts
-          path: git-crypt.exe
-  upload:
-    name: Upload Release Binary
-    runs-on: ubuntu-latest
-    needs: build
-    permissions:
-      contents: write
-    steps:
-      - name: Download release artifact
-        uses: actions/download-artifact@v3
-        with:
-          name: git-crypt-artifacts
-      - name: Upload release asset
-        uses: actions/github-script@v3
-        with:
-          github-token: ${{ secrets.GITHUB_TOKEN }}
-          script: |
-            const fs = require("fs").promises;
-            const { repo: { owner, repo }, sha } = context;
-            await github.repos.uploadReleaseAsset({
-              owner, repo,
-              release_id: ${{ github.event.release.id }},
-              name: 'git-crypt-${{ github.event.release.name }}-x86_64.exe',
-              data: await fs.readFile('git-crypt.exe'),
-            });

CONTRIBUTING.md

@@ -4,7 +4,8 @@ documentation, bug reports, or anything else that improves git-crypt.
 
 When contributing code, please consider the following guidelines:
 
- * You are encouraged to open an issue on GitHub to discuss any non-trivial
+ * You are encouraged to open an issue on GitHub or send mail to
+   git-crypt-discuss@lists.cloudmutt.com to discuss any non-trivial
    changes before you start coding.
 
  * Please mimic the existing code style as much as possible.  In
@@ -14,7 +15,8 @@ When contributing code, please consider the following guidelines:
  * To minimize merge commits, please rebase your changes before opening
    a pull request.
 
- * To submit your patch, open a pull request on GitHub.
+ * To submit your patch, open a pull request on GitHub or send a
+   properly-formatted patch to git-crypt-discuss@lists.cloudmutt.com.
 
 Finally, be aware that since git-crypt is security-sensitive software,
 the bar for contributions is higher than average.  Please don't be

NEWS

@@ -1,8 +1,3 @@
-v0.7.0 (2022-04-21)
-  * Avoid "argument list too long" errors on macOS.
-  * Fix handling of "-" arguments.
-  * Minor documentation improvements.
-
 v0.6.0 (2017-11-26)
   * Add support for OpenSSL 1.1 (still works with OpenSSL 1.0).
   * Switch to C++11 (gcc 4.9 or higher now required to build).

NEWS.md

@@ -1,11 +1,6 @@
 News
 ====
 
-######v0.7.0 (2022-04-21)
-* Avoid "argument list too long" errors on macOS.
-* Fix handling of "-" arguments.
-* Minor documentation improvements.
-
 ######v0.6.0 (2017-11-26)
 * Add support for OpenSSL 1.1 (still works with OpenSSL 1.0).
 * Switch to C++11 (gcc 4.9 or higher now required to build).

README

@@ -70,7 +70,7 @@ encryption and decryption happen transparently.
 
 CURRENT STATUS
 
-The latest version of git-crypt is 0.7.0, released on 2022-04-21.
+The latest version of git-crypt is 0.6.0, released on 2017-11-26.
 git-crypt aims to be bug-free and reliable, meaning it shouldn't
 crash, malfunction, or expose your confidential data.  However,
 it has not yet reached maturity, meaning it is not as documented,
@@ -158,3 +158,12 @@ match it accidentally.  If necessary, you can exclude .gitattributes from
 encryption like this:
 
 	.gitattributes !filter !diff
+
+
+MAILING LISTS
+
+To stay abreast of, and provide input to, git-crypt development, consider
+subscribing to one or both of our mailing lists:
+
+Announcements: https://lists.cloudmutt.com/mailman/listinfo/git-crypt-announce
+Discussion:    https://lists.cloudmutt.com/mailman/listinfo/git-crypt-discuss

README.md

@@ -71,8 +71,8 @@ encryption and decryption happen transparently.
 Current Status
 --------------
 
-The latest version of git-crypt is [0.7.0](NEWS.md), released on
-2022-04-21.  git-crypt aims to be bug-free and reliable, meaning it
+The latest version of git-crypt is [0.6.0](NEWS.md), released on
+2017-11-26.  git-crypt aims to be bug-free and reliable, meaning it
 shouldn't crash, malfunction, or expose your confidential data.
 However, it has not yet reached maturity, meaning it is not as
 documented, featureful, or easy-to-use as it should be.  Additionally,
@@ -160,3 +160,12 @@ match it accidentally.  If necessary, you can exclude .gitattributes from
 encryption like this:
 
     .gitattributes !filter !diff
+
+Mailing Lists
+-------------
+
+To stay abreast of, and provide input to, git-crypt development,
+consider subscribing to one or both of our mailing lists:
+
+* [Announcements](https://lists.cloudmutt.com/mailman/listinfo/git-crypt-announce)
+* [Discussion](https://lists.cloudmutt.com/mailman/listinfo/git-crypt-discuss)

commands.cpp

@@ -461,6 +461,25 @@ static std::pair<std::string, std::string> get_file_attributes (const std::strin
 	return std::make_pair(filter_attr, diff_attr);
 }
 
+static bool check_if_blob_is_empty (const std::string& object_id)
+{
+	// git cat-file blob object_id
+
+	std::vector<std::string>	command;
+	command.push_back("git");
+	command.push_back("cat-file");
+	command.push_back("blob");
+	command.push_back(object_id);
+
+	// TODO: do this more efficiently - don't read entire command output into buffer, only read what we need
+	std::stringstream		output;
+	if (!successful_exit(exec_command(command, output))) {
+		throw Error("'git cat-file' failed - is this a Git repository?");
+	}
+
+	return output.get() == std::stringstream::traits_type::eof();
+}
+
 static bool check_if_blob_is_encrypted (const std::string& object_id)
 {
 	// git cat-file blob object_id
@@ -770,6 +789,10 @@ int clean (int argc, const char** argv)
 		return 1;
 	}
 
+	if (file_size == 0 && key_file.get_skip_empty()) {
+		return 0;
+	}
+
 	// We use an HMAC of the file as the encryption nonce (IV) for CTR mode.
 	// By using a hash of the file we ensure that the encryption is
 	// deterministic so git doesn't think the file has changed when it really
@@ -887,6 +910,11 @@ int smudge (int argc, const char** argv)
 	// Read the header to get the nonce and make sure it's actually encrypted
 	unsigned char		header[10 + Aes_ctr_decryptor::NONCE_LEN];
 	std::cin.read(reinterpret_cast<char*>(header), sizeof(header));
+
+	if (std::cin.gcount() == 0 && key_file.get_skip_empty()) {
+		return 0;
+	}
+
 	if (std::cin.gcount() != sizeof(header) || std::memcmp(header, "\0GITCRYPT\0", 10) != 0) {
 		// File not encrypted - just copy it out to stdout
 		std::clog << "git-crypt: Warning: file not encrypted" << std::endl;
@@ -991,6 +1019,7 @@ int init (int argc, const char** argv)
 	std::clog << "Generating key..." << std::endl;
 	Key_file		key_file;
 	key_file.set_key_name(key_name);
+	key_file.set_skip_empty(true);
 	key_file.generate();
 
 	mkdir_parent(internal_key_path);
@@ -1425,6 +1454,7 @@ int keygen (int argc, const char** argv)
 
 	std::clog << "Generating key..." << std::endl;
 	Key_file		key_file;
+	key_file.set_skip_empty(true);
 	key_file.generate();
 
 	if (std::strcmp(key_file_name, "-") == 0) {
@@ -1629,7 +1659,8 @@ int status (int argc, const char** argv)
 
 		if (file_attrs.first == "git-crypt" || std::strncmp(file_attrs.first.c_str(), "git-crypt-", 10) == 0) {
 			// File is encrypted
-			const bool	blob_is_unencrypted = !object_id.empty() && !check_if_blob_is_encrypted(object_id);
+			// If the file is empty, don't consider it unencrypted, because in newly-initialized repos (specifically those with keys with skip_empty set) we don't encrypt empty files. Unfortunately, we can't easily determine here if the key has skip_empty set, so just act like it is. This means we won't notice if an old repo has an empty unencrypted file that should be encrypted. Fortunately, this isn't really a big deal because empty files obviously don't contain anything sensitive in them.
+			const bool	blob_is_unencrypted = !object_id.empty() && !check_if_blob_is_encrypted(object_id) && !check_if_blob_is_empty(object_id);
 
 			if (fix_problems && blob_is_unencrypted) {
 				if (access(filename.c_str(), F_OK) != 0) {

git-crypt.hpp

@@ -31,7 +31,7 @@
 #ifndef GIT_CRYPT_GIT_CRYPT_HPP
 #define GIT_CRYPT_GIT_CRYPT_HPP
 
-#define VERSION "0.7.0"
+#define VERSION "0.6.0"
 
 extern const char*	argv0;	// initialized in main() to argv[0]
 

key.cpp

@@ -232,6 +232,11 @@ void		Key_file::load_header (std::istream& in)
 				key_name.clear();
 				throw Malformed();
 			}
+		} else if (field_id == HEADER_FIELD_SKIP_EMPTY) {
+			if (field_len != 0) {
+				throw Malformed();
+			}
+			skip_empty = true;
 		} else if (field_id & 1) { // unknown critical field
 			throw Incompatible();
 		} else {
@@ -256,6 +261,10 @@ void		Key_file::store (std::ostream& out) const
 		write_be32(out, key_name.size());
 		out.write(key_name.data(), key_name.size());
 	}
+	if (skip_empty) {
+		write_be32(out, HEADER_FIELD_SKIP_EMPTY);
+		write_be32(out, 0);
+	}
 	write_be32(out, HEADER_FIELD_END);
 	for (Map::const_iterator it(entries.begin()); it != entries.end(); ++it) {
 		it->second.store(out);

key.hpp

@@ -83,18 +83,23 @@ public:
 
 	void				set_key_name (const char* k) { key_name = k ? k : ""; }
 	const char*			get_key_name () const { return key_name.empty() ? 0 : key_name.c_str(); }
+
+	void				set_skip_empty (bool v) { skip_empty = v; }
+	bool				get_skip_empty () const { return skip_empty; }
 private:
 	typedef std::map<uint32_t, Entry, std::greater<uint32_t> > Map;
 	enum { FORMAT_VERSION = 2 };
 
 	Map				entries;
 	std::string			key_name;
+	bool				skip_empty = false;
 
 	void				load_header (std::istream&);
 
 	enum {
 		HEADER_FIELD_END	= 0,
-		HEADER_FIELD_KEY_NAME	= 1
+		HEADER_FIELD_KEY_NAME	= 1,
+		HEADER_FIELD_SKIP_EMPTY	= 3	// If this field is present, empty files are left unencrypted (see issue #53)
 	};
 	enum {
 		KEY_FIELD_END		= 0,

man/git-crypt.xml

@@ -7,8 +7,8 @@
 	-->
 	<refentryinfo>
 		<title>git-crypt</title>
-		<date>2022-04-21</date>
-		<productname>git-crypt 0.7.0</productname>
+		<date>2017-11-26</date>
+		<productname>git-crypt 0.6.0</productname>
 
 		<author>
 			<othername>Andrew Ayer</othername>