LibTar: Implement TarOutputStream

This output stream creates gnu format tar archives
and currently only supports files and directories.

Userland/Libraries/LibTar/Tar.h

@@ -27,7 +27,9 @@
 
 #pragma once
 
+#include <AK/String.h>
 #include <AK/StringView.h>
+#include <string.h>
 #include <sys/types.h>
 
 namespace Tar {
@@ -52,9 +54,8 @@ constexpr const char* gnu_version = " ";     // gnu format version
 constexpr const char* ustar_magic = "ustar"; // ustar format magic
 constexpr const char* ustar_version = "00";  // ustar format version
 
-class Header {
+class [[gnu::packed]] Header {
 public:
-    // FIXME: support ustar filename prefix
     const StringView file_name() const { return m_file_name; }
     mode_t mode() const { return get_tar_field(m_mode); }
     uid_t uid() const { return get_tar_field(m_uid); }
@@ -70,6 +71,26 @@ public:
     const StringView group_name() const { return m_group_name; }
     int major() const { return get_tar_field(m_major); }
     int minor() const { return get_tar_field(m_minor); }
+    // FIXME: support ustar filename prefix
+    const StringView prefix() const { return m_prefix; }
+
+    void set_file_name(const String& file_name) { VERIFY(file_name.copy_characters_to_buffer(m_file_name, sizeof(m_file_name))); }
+    void set_mode(mode_t mode) { VERIFY(String::formatted("{:o}", mode).copy_characters_to_buffer(m_mode, sizeof(m_mode))); }
+    void set_uid(uid_t uid) { VERIFY(String::formatted("{:o}", uid).copy_characters_to_buffer(m_uid, sizeof(m_uid))); }
+    void set_gid(gid_t gid) { VERIFY(String::formatted("{:o}", gid).copy_characters_to_buffer(m_gid, sizeof(m_gid))); }
+    void set_size(size_t size) { VERIFY(String::formatted("{:o}", size).copy_characters_to_buffer(m_size, sizeof(m_size))); }
+    void set_timestamp(time_t timestamp) { VERIFY(String::formatted("{:o}", timestamp).copy_characters_to_buffer(m_timestamp, sizeof(m_timestamp))); }
+    void set_type_flag(FileType type) { m_type_flag = type; }
+    void set_link_name(const String& link_name) { VERIFY(link_name.copy_characters_to_buffer(m_link_name, sizeof(m_link_name))); }
+    void set_magic(const char* magic) { memcpy(m_magic, magic, sizeof(m_magic)); }           // magic doesnt necessarily include a null byte
+    void set_version(const char* version) { memcpy(m_version, version, sizeof(m_version)); } // version doesnt necessarily include a null byte
+    void set_owner_name(const String& owner_name) { VERIFY(owner_name.copy_characters_to_buffer(m_owner_name, sizeof(m_owner_name))); }
+    void set_group_name(const String& group_name) { VERIFY(group_name.copy_characters_to_buffer(m_group_name, sizeof(m_group_name))); }
+    void set_major(int major) { VERIFY(String::formatted("{:o}", major).copy_characters_to_buffer(m_major, sizeof(m_major))); }
+    void set_minor(int minor) { VERIFY(String::formatted("{:o}", minor).copy_characters_to_buffer(m_minor, sizeof(m_minor))); }
+    void set_prefix(const String& prefix) { VERIFY(prefix.copy_characters_to_buffer(m_prefix, sizeof(m_prefix))); }
+
+    void calculate_checksum();
 
 private:
     char m_file_name[100];
@@ -78,7 +99,7 @@ private:
     char m_gid[8];
     char m_size[12];
     char m_timestamp[12];
-    char m_checksum[8];
+    char m_checksum[8]; // an uninitialized header's checksum is filled with spaces
     char m_type_flag;
     char m_link_name[100];
     char m_magic[6];
@@ -87,7 +108,7 @@ private:
     char m_group_name[32];
     char m_major[8];
     char m_minor[8];
-    char m_prefix[155];
+    char m_prefix[155]; // zero out the prefix for archiving
 
     template<size_t N>
     static size_t get_tar_field(const char (&field)[N]);
@@ -107,5 +128,14 @@ size_t Header::get_tar_field(const char (&field)[N])
     }
     return value;
 }
+void Header::calculate_checksum()
+{
+    memset(m_checksum, ' ', sizeof(m_checksum));
+    auto checksum = 0u;
+    for (auto i = 0u; i < sizeof(Header); ++i) {
+        checksum += ((unsigned char*)this)[i];
+    }
+    VERIFY(String::formatted("{:o}", checksum).copy_characters_to_buffer(m_checksum, sizeof(m_checksum)));
+}
 
 }

Userland/Libraries/LibTar/TarStream.cpp

@@ -26,9 +26,10 @@
  */
 
 #include <LibTar/TarStream.h>
+#include <string.h>
 
 namespace Tar {
-TarFileStream::TarFileStream(TarStream& tar_stream)
+TarFileStream::TarFileStream(TarInputStream& tar_stream)
     : m_tar_stream(tar_stream)
     , m_generation(tar_stream.m_generation)
 {
@@ -83,11 +84,12 @@ bool TarFileStream::discard_or_error(size_t count)
     return m_tar_stream.m_stream.discard_or_error(count);
 }
 
-TarStream::TarStream(InputStream& stream)
+TarInputStream::TarInputStream(InputStream& stream)
     : m_stream(stream)
 {
     if (!m_stream.read_or_error(Bytes(&m_header, sizeof(m_header)))) {
         m_finished = true;
+        m_stream.handle_any_error(); // clear out errors so we dont assert
         return;
     }
     VERIFY(m_stream.discard_or_error(block_size - sizeof(Header)));
@@ -98,7 +100,7 @@ static constexpr unsigned long block_ceiling(unsigned long offset)
     return block_size * (1 + ((offset - 1) / block_size));
 }
 
-void TarStream::advance()
+void TarInputStream::advance()
 {
     if (m_finished)
         return;
@@ -119,17 +121,70 @@ void TarStream::advance()
     VERIFY(m_stream.discard_or_error(block_size - sizeof(Header)));
 }
 
-bool TarStream::valid() const
+bool TarInputStream::valid() const
 {
     auto& header_magic = header().magic();
     auto& header_version = header().version();
     return (header_magic == gnu_magic && header_version == gnu_version) || (header_magic == ustar_magic && header_version == ustar_version);
 }
 
-TarFileStream TarStream::file_contents()
+TarFileStream TarInputStream::file_contents()
 {
     VERIFY(!m_finished);
     return TarFileStream(*this);
 }
 
+TarOutputStream::TarOutputStream(OutputStream& stream)
+    : m_stream(stream)
+{
+}
+
+void TarOutputStream::add_directory(const String& path, mode_t mode)
+{
+    VERIFY(!m_finished);
+    Header header;
+    memset(&header, 0, sizeof(header));
+    header.set_size(0);
+    header.set_file_name(String::formatted("{}/", path)); // Old tar implementations assume directory names end with a /
+    header.set_type_flag(Directory);
+    header.set_mode(mode);
+    header.set_magic(gnu_magic);
+    header.set_version(gnu_version);
+    header.calculate_checksum();
+    VERIFY(m_stream.write_or_error(Bytes { &header, sizeof(header) }));
+    u8 padding[block_size] = { 0 };
+    VERIFY(m_stream.write_or_error(Bytes { &padding, block_size - sizeof(header) }));
+}
+
+void TarOutputStream::add_file(const String& path, mode_t mode, const ReadonlyBytes& bytes)
+{
+    VERIFY(!m_finished);
+    Header header;
+    memset(&header, 0, sizeof(header));
+    header.set_size(bytes.size());
+    header.set_file_name(path);
+    header.set_type_flag(NormalFile);
+    header.set_mode(mode);
+    header.set_magic(gnu_magic);
+    header.set_version(gnu_version);
+    header.calculate_checksum();
+    VERIFY(m_stream.write_or_error(Bytes { &header, sizeof(header) }));
+    u8 padding[block_size] = { 0 };
+    VERIFY(m_stream.write_or_error(Bytes { &padding, block_size - sizeof(header) }));
+    size_t n_written = 0;
+    while (n_written < bytes.size()) {
+        n_written += m_stream.write(bytes.slice(n_written, min(bytes.size() - n_written, block_size)));
+    }
+    VERIFY(m_stream.write_or_error(Bytes { &padding, block_size - (n_written % block_size) }));
+}
+
+void TarOutputStream::finish()
+{
+    VERIFY(!m_finished);
+    u8 padding[block_size] = { 0 };
+    m_stream.write_or_error(Bytes { &padding, block_size }); // 2 empty records that are used to signify the end of the archive
+    m_stream.write_or_error(Bytes { &padding, block_size });
+    m_finished = true;
+}
+
 }

Userland/Libraries/LibTar/TarStream.h

@@ -1,5 +1,6 @@
 /*
  * Copyright (c) 2020, Peter Elliott <pelliott@ualberta.ca>
+ * Copyright (c) 2021, Idan Horowitz <idan.horowitz@gmail.com>
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -32,7 +33,7 @@
 
 namespace Tar {
 
-class TarStream;
+class TarInputStream;
 
 class TarFileStream : public InputStream {
 public:
@@ -43,16 +44,16 @@ public:
     bool discard_or_error(size_t count) override;
 
 private:
-    TarFileStream(TarStream& stream);
+    TarFileStream(TarInputStream& stream);
-    TarStream& m_tar_stream;
+    TarInputStream& m_tar_stream;
     int m_generation;
 
-    friend class TarStream;
+    friend class TarInputStream;
 };
 
-class TarStream {
+class TarInputStream {
 public:
-    TarStream(InputStream&);
+    TarInputStream(InputStream&);
     void advance();
     bool finished() const { return m_finished; }
     bool valid() const;
@@ -69,4 +70,18 @@ private:
     friend class TarFileStream;
 };
 
+class TarOutputStream {
+public:
+    TarOutputStream(OutputStream&);
+    void add_file(const String& path, mode_t, const ReadonlyBytes&);
+    void add_directory(const String& path, mode_t);
+    void finish();
+
+private:
+    OutputStream& m_stream;
+    bool m_finished { false };
+
+    friend class TarFileStream;
+};
+
 }

Userland/Utilities/tar.cpp

@@ -78,7 +78,7 @@ int main(int argc, char** argv)
 
         InputStream& file_input_stream = file_stream;
         InputStream& gzip_input_stream = gzip_stream;
-        Tar::TarStream tar_stream((gzip) ? gzip_input_stream : file_input_stream);
+        Tar::TarInputStream tar_stream((gzip) ? gzip_input_stream : file_input_stream);
         if (!tar_stream.valid()) {
             warnln("the provided file is not a well-formatted ustar file");
             return 1;