| 1 | const Writer = @This(); |
| 2 | |
| 3 | const std = @import("std"); |
| 4 | const Io = std.Io; |
| 5 | const assert = std.debug.assert; |
| 6 | const testing = std.testing; |
| 7 | |
| 8 | const block_size = @sizeOf(Header); |
| 9 | |
| 10 | /// Options for writing file/dir/link. If left empty 0o664 is used for |
| 11 | /// file mode and current time for mtime. |
| 12 | pub const Options = struct { |
| 13 | /// File system permission mode. |
| 14 | mode: u32 = 0, |
| 15 | /// File system modification time. |
| 16 | mtime: u64 = 0, |
| 17 | }; |
| 18 | |
| 19 | underlying_writer: *Io.Writer, |
| 20 | /// Assumed to use `/` for any path separators. |
| 21 | prefix: []const u8 = "", |
| 22 | |
| 23 | const Error = error{ |
| 24 | WriteFailed, |
| 25 | OctalOverflow, |
| 26 | NameTooLong, |
| 27 | }; |
| 28 | |
| 29 | /// Sets prefix for all other write* method paths. |
| 30 | /// `root` is assumed to use `/` for any path separators. |
| 31 | pub fn setRoot(w: *Writer, root: []const u8) Error!void { |
| 32 | if (root.len > 0) |
| 33 | try w.writeDir(root, .{}); |
| 34 | |
| 35 | w.prefix = root; |
| 36 | } |
| 37 | |
| 38 | pub fn writeDir(w: *Writer, sub_path: []const u8, options: Options) Error!void { |
| 39 | try w.writeHeader(.directory, sub_path, "", 0, options); |
| 40 | } |
| 41 | |
| 42 | pub const WriteFileError = Io.Writer.FileError || Error || Io.File.Reader.SizeError; |
| 43 | |
| 44 | pub fn writeFileTimestamp( |
| 45 | w: *Writer, |
| 46 | /// Assumed to use `/` for any path separators. |
| 47 | sub_path: []const u8, |
| 48 | file_reader: *Io.File.Reader, |
| 49 | mtime: Io.Timestamp, |
| 50 | ) WriteFileError!void { |
| 51 | return writeFile(w, sub_path, file_reader, @intCast(mtime.toSeconds())); |
| 52 | } |
| 53 | |
| 54 | pub fn writeFile( |
| 55 | w: *Writer, |
| 56 | /// Assumed to use `/` for any path separators. |
| 57 | sub_path: []const u8, |
| 58 | file_reader: *Io.File.Reader, |
| 59 | /// If you want to match the file format's expectations, it wants number of |
| 60 | /// seconds since POSIX epoch. Zero is also a great option here to make |
| 61 | /// generated tarballs more reproducible. |
| 62 | mtime: u64, |
| 63 | ) WriteFileError!void { |
| 64 | const size = try file_reader.getSize(); |
| 65 | |
| 66 | var header: Header = .{}; |
| 67 | try w.setPath(&header, sub_path); |
| 68 | try header.setSize(size); |
| 69 | try header.setMtime(mtime); |
| 70 | try header.updateChecksum(); |
| 71 | |
| 72 | try w.underlying_writer.writeAll(@ptrCast((&header)[0..1])); |
| 73 | _ = try w.underlying_writer.sendFileAll(file_reader, .unlimited); |
| 74 | try w.writePadding64(size); |
| 75 | } |
| 76 | |
| 77 | pub const WriteFileStreamError = Error || Io.Reader.StreamError; |
| 78 | |
| 79 | /// Writes file reading file content from `reader`. Reads exactly `size` bytes |
| 80 | /// from `reader`, or returns `error.EndOfStream`. |
| 81 | pub fn writeFileStream( |
| 82 | w: *Writer, |
| 83 | /// Assumed to use `/` for any path separators. |
| 84 | sub_path: []const u8, |
| 85 | size: u64, |
| 86 | reader: *Io.Reader, |
| 87 | options: Options, |
| 88 | ) WriteFileStreamError!void { |
| 89 | try w.writeHeader(.regular, sub_path, "", size, options); |
| 90 | try reader.streamExact64(w.underlying_writer, size); |
| 91 | try w.writePadding64(size); |
| 92 | } |
| 93 | |
| 94 | /// Writes file using bytes buffer `content` for size and file content. |
| 95 | pub fn writeFileBytes( |
| 96 | w: *Writer, |
| 97 | /// Assumed to use `/` for all path separators. |
| 98 | sub_path: []const u8, |
| 99 | content: []const u8, |
| 100 | options: Options, |
| 101 | ) Error!void { |
| 102 | try w.writeHeader(.regular, sub_path, "", content.len, options); |
| 103 | try w.underlying_writer.writeAll(content); |
| 104 | try w.writePadding(content.len); |
| 105 | } |
| 106 | |
| 107 | pub fn writeLink(w: *Writer, sub_path: []const u8, link_name: []const u8, options: Options) Error!void { |
| 108 | try w.writeHeader(.symbolic_link, sub_path, link_name, 0, options); |
| 109 | } |
| 110 | |
| 111 | fn writeHeader( |
| 112 | w: *Writer, |
| 113 | typeflag: Header.FileType, |
| 114 | sub_path: []const u8, |
| 115 | link_name: []const u8, |
| 116 | size: u64, |
| 117 | options: Options, |
| 118 | ) Error!void { |
| 119 | var header = Header.init(typeflag); |
| 120 | try w.setPath(&header, sub_path); |
| 121 | try header.setSize(size); |
| 122 | try header.setMtime(options.mtime); |
| 123 | if (options.mode != 0) |
| 124 | try header.setMode(options.mode); |
| 125 | if (typeflag == .symbolic_link) |
| 126 | header.setLinkname(link_name) catch |err| switch (err) { |
| 127 | error.NameTooLong => try w.writeExtendedHeader(.gnu_long_link, &.{link_name}), |
| 128 | else => |e| return e, |
| 129 | }; |
| 130 | try header.write(w.underlying_writer); |
| 131 | } |
| 132 | |
| 133 | /// Writes path in posix header, if don't fit (in name+prefix; 100+155 |
| 134 | /// bytes) writes it in gnu extended header. |
| 135 | fn setPath(w: *Writer, header: *Header, sub_path: []const u8) Error!void { |
| 136 | header.setPath(w.prefix, sub_path) catch |err| switch (err) { |
| 137 | error.NameTooLong => { |
| 138 | // write extended header |
| 139 | const buffers: []const []const u8 = if (w.prefix.len == 0) |
| 140 | &.{sub_path} |
| 141 | else |
| 142 | &.{ w.prefix, "/", sub_path }; |
| 143 | try w.writeExtendedHeader(.gnu_long_name, buffers); |
| 144 | }, |
| 145 | else => |e| return e, |
| 146 | }; |
| 147 | } |
| 148 | |
| 149 | /// Writes gnu extended header: gnu_long_name or gnu_long_link. |
| 150 | fn writeExtendedHeader(w: *Writer, typeflag: Header.FileType, buffers: []const []const u8) Error!void { |
| 151 | var len: usize = 0; |
| 152 | for (buffers) |buf| len += buf.len; |
| 153 | |
| 154 | var header: Header = .init(typeflag); |
| 155 | try header.setSize(len); |
| 156 | try header.write(w.underlying_writer); |
| 157 | for (buffers) |buf| |
| 158 | try w.underlying_writer.writeAll(buf); |
| 159 | try w.writePadding(len); |
| 160 | } |
| 161 | |
| 162 | fn writePadding(w: *Writer, bytes: usize) Io.Writer.Error!void { |
| 163 | return writePaddingPos(w, bytes % block_size); |
| 164 | } |
| 165 | |
| 166 | fn writePadding64(w: *Writer, bytes: u64) Io.Writer.Error!void { |
| 167 | return writePaddingPos(w, @intCast(bytes % block_size)); |
| 168 | } |
| 169 | |
| 170 | fn writePaddingPos(w: *Writer, pos: usize) Io.Writer.Error!void { |
| 171 | if (pos == 0) return; |
| 172 | try w.underlying_writer.splatByteAll(0, block_size - pos); |
| 173 | } |
| 174 | |
| 175 | /// According to the specification, tar should finish with two zero blocks, but |
| 176 | /// "reasonable system must not assume that such a block exists when reading an |
| 177 | /// archive". Therefore, the Zig standard library recommends to not call this |
| 178 | /// function. |
| 179 | pub fn finishPedantically(w: *Writer) Io.Writer.Error!void { |
| 180 | try w.underlying_writer.splatByteAll(0, block_size * 2); |
| 181 | } |
| 182 | |
| 183 | /// A struct that is exactly 512 bytes and matches tar file format. This is |
| 184 | /// intended to be used for outputting tar files; for parsing there is |
| 185 | /// `std.tar.Header`. |
| 186 | pub const Header = extern struct { |
| 187 | // This struct was originally copied from |
| 188 | // https://github.com/mattnite/tar/blob/main/src/main.zig which is MIT |
| 189 | // licensed. |
| 190 | // |
| 191 | // The name, linkname, magic, uname, and gname are null-terminated character |
| 192 | // strings. All other fields are zero-filled octal numbers in ASCII. Each |
| 193 | // numeric field of width w contains w minus 1 digits, and a null. |
| 194 | // Reference: https://www.gnu.org/software/tar/manual/html_node/Standard.html |
| 195 | // POSIX header: byte offset |
| 196 | name: [100]u8 = @splat(0), // 0 |
| 197 | mode: [7:0]u8 = default_mode.file, // 100 |
| 198 | uid: [7:0]u8 = @splat(0), // unused 108 |
| 199 | gid: [7:0]u8 = @splat(0), // unused 116 |
| 200 | size: [11:0]u8 = @splat('0'), // 124 |
| 201 | mtime: [11:0]u8 = @splat('0'), // 136 |
| 202 | checksum: [7:0]u8 = @splat(' '), // 148 |
| 203 | typeflag: FileType = .regular, // 156 |
| 204 | linkname: [100]u8 = @splat(0), // 157 |
| 205 | magic: [6]u8 = .{ 'u', 's', 't', 'a', 'r', 0 }, // 257 |
| 206 | version: [2]u8 = .{ '0', '0' }, // 263 |
| 207 | uname: [32]u8 = @splat(0), // unused 265 |
| 208 | gname: [32]u8 = @splat(0), // unused 297 |
| 209 | devmajor: [7:0]u8 = @splat(0), // unused 329 |
| 210 | devminor: [7:0]u8 = @splat(0), // unused 337 |
| 211 | prefix: [155]u8 = @splat(0), // 345 |
| 212 | pad: [12]u8 = @splat(0), // unused 500 |
| 213 | |
| 214 | pub const FileType = enum(u8) { |
| 215 | regular = '0', |
| 216 | symbolic_link = '2', |
| 217 | directory = '5', |
| 218 | gnu_long_name = 'L', |
| 219 | gnu_long_link = 'K', |
| 220 | }; |
| 221 | |
| 222 | const default_mode = struct { |
| 223 | const file = [_:0]u8{ '0', '0', '0', '0', '6', '6', '4' }; // 0o664 |
| 224 | const dir = [_:0]u8{ '0', '0', '0', '0', '7', '7', '5' }; // 0o775 |
| 225 | const sym_link = [_:0]u8{ '0', '0', '0', '0', '7', '7', '7' }; // 0o777 |
| 226 | const other = [_:0]u8{ '0', '0', '0', '0', '0', '0', '0' }; // 0o000 |
| 227 | }; |
| 228 | |
| 229 | pub fn init(typeflag: FileType) Header { |
| 230 | return .{ |
| 231 | .typeflag = typeflag, |
| 232 | .mode = switch (typeflag) { |
| 233 | .directory => default_mode.dir, |
| 234 | .symbolic_link => default_mode.sym_link, |
| 235 | .regular => default_mode.file, |
| 236 | else => default_mode.other, |
| 237 | }, |
| 238 | }; |
| 239 | } |
| 240 | |
| 241 | pub fn setSize(w: *Header, size: u64) error{OctalOverflow}!void { |
| 242 | try octal(&w.size, size); |
| 243 | } |
| 244 | |
| 245 | fn octal(buf: []u8, value: u64) error{OctalOverflow}!void { |
| 246 | var remainder: u64 = value; |
| 247 | var pos: usize = buf.len; |
| 248 | while (remainder > 0 and pos > 0) { |
| 249 | pos -= 1; |
| 250 | const c: u8 = @as(u8, @intCast(remainder % 8)) + '0'; |
| 251 | buf[pos] = c; |
| 252 | remainder /= 8; |
| 253 | if (pos == 0 and remainder > 0) return error.OctalOverflow; |
| 254 | } |
| 255 | } |
| 256 | |
| 257 | pub fn setMode(w: *Header, mode: u32) error{OctalOverflow}!void { |
| 258 | try octal(&w.mode, mode); |
| 259 | } |
| 260 | |
| 261 | // Integer number of seconds since January 1, 1970, 00:00 Coordinated Universal Time. |
| 262 | pub fn setMtime(w: *Header, mtime: u64) error{OctalOverflow}!void { |
| 263 | try octal(&w.mtime, mtime); |
| 264 | } |
| 265 | |
| 266 | pub fn updateChecksum(w: *Header) !void { |
| 267 | var checksum: usize = ' '; // other 7 w.checksum bytes are initialized to ' ' |
| 268 | for (std.mem.asBytes(w)) |val| |
| 269 | checksum += val; |
| 270 | try octal(&w.checksum, checksum); |
| 271 | } |
| 272 | |
| 273 | pub fn write(h: *Header, bw: *Io.Writer) error{ OctalOverflow, WriteFailed }!void { |
| 274 | try h.updateChecksum(); |
| 275 | try bw.writeAll(std.mem.asBytes(h)); |
| 276 | } |
| 277 | |
| 278 | pub fn setLinkname(w: *Header, link: []const u8) !void { |
| 279 | if (link.len > w.linkname.len) return error.NameTooLong; |
| 280 | @memcpy(w.linkname[0..link.len], link); |
| 281 | } |
| 282 | |
| 283 | pub fn setPath(w: *Header, prefix: []const u8, sub_path: []const u8) !void { |
| 284 | const max_prefix = w.prefix.len; |
| 285 | const max_name = w.name.len; |
| 286 | const sep = std.fs.path.sep_posix; |
| 287 | |
| 288 | if (prefix.len + sub_path.len > max_name + max_prefix or prefix.len > max_prefix) |
| 289 | return error.NameTooLong; |
| 290 | |
| 291 | // both fit into name |
| 292 | if (prefix.len > 0 and prefix.len + sub_path.len < max_name) { |
| 293 | @memcpy(w.name[0..prefix.len], prefix); |
| 294 | w.name[prefix.len] = sep; |
| 295 | @memcpy(w.name[prefix.len + 1 ..][0..sub_path.len], sub_path); |
| 296 | return; |
| 297 | } |
| 298 | |
| 299 | // sub_path fits into name |
| 300 | // there is no prefix or prefix fits into prefix |
| 301 | if (sub_path.len <= max_name) { |
| 302 | @memcpy(w.name[0..sub_path.len], sub_path); |
| 303 | @memcpy(w.prefix[0..prefix.len], prefix); |
| 304 | return; |
| 305 | } |
| 306 | |
| 307 | if (prefix.len > 0) { |
| 308 | @memcpy(w.prefix[0..prefix.len], prefix); |
| 309 | w.prefix[prefix.len] = sep; |
| 310 | } |
| 311 | const prefix_pos = if (prefix.len > 0) prefix.len + 1 else 0; |
| 312 | |
| 313 | // add as much to prefix as you can, must split at / |
| 314 | const prefix_remaining = max_prefix - prefix_pos; |
| 315 | if (std.mem.findLast(u8, sub_path[0..@min(prefix_remaining, sub_path.len)], &.{'/'})) |sep_pos| { |
| 316 | @memcpy(w.prefix[prefix_pos..][0..sep_pos], sub_path[0..sep_pos]); |
| 317 | if ((sub_path.len - sep_pos - 1) > max_name) return error.NameTooLong; |
| 318 | @memcpy(w.name[0..][0 .. sub_path.len - sep_pos - 1], sub_path[sep_pos + 1 ..]); |
| 319 | return; |
| 320 | } |
| 321 | |
| 322 | return error.NameTooLong; |
| 323 | } |
| 324 | |
| 325 | comptime { |
| 326 | assert(@sizeOf(Header) == 512); |
| 327 | } |
| 328 | |
| 329 | test "setPath" { |
| 330 | const cases = [_]struct { |
| 331 | in: []const []const u8, |
| 332 | out: []const []const u8, |
| 333 | }{ |
| 334 | .{ |
| 335 | .in = &.{ "", "123456789" }, |
| 336 | .out = &.{ "", "123456789" }, |
| 337 | }, |
| 338 | // can fit into name |
| 339 | .{ |
| 340 | .in = &.{ "prefix", "sub_path" }, |
| 341 | .out = &.{ "", "prefix/sub_path" }, |
| 342 | }, |
| 343 | // no more both fits into name |
| 344 | .{ |
| 345 | .in = &.{ "prefix", repeatString(8, "0123456789/") ++ "basename" }, |
| 346 | .out = &.{ "prefix", repeatString(8, "0123456789/") ++ "basename" }, |
| 347 | }, |
| 348 | // put as much as you can into prefix the rest goes into name |
| 349 | .{ |
| 350 | .in = &.{ "prefix", repeatString(10, "0123456789/") ++ "basename" }, |
| 351 | .out = &.{ "prefix/" ++ repeatString(9, "0123456789/") ++ "0123456789", "basename" }, |
| 352 | }, |
| 353 | |
| 354 | .{ |
| 355 | .in = &.{ "prefix", repeatString(15, "0123456789/") ++ "basename" }, |
| 356 | .out = &.{ "prefix/" ++ repeatString(12, "0123456789/") ++ "0123456789", "0123456789/0123456789/basename" }, |
| 357 | }, |
| 358 | .{ |
| 359 | .in = &.{ "prefix", repeatString(21, "0123456789/") ++ "basename" }, |
| 360 | .out = &.{ "prefix/" ++ repeatString(12, "0123456789/") ++ "0123456789", repeatString(8, "0123456789/") ++ "basename" }, |
| 361 | }, |
| 362 | .{ |
| 363 | .in = &.{ "", repeatString(10, "012345678/") ++ "foo" }, |
| 364 | .out = &.{ repeatString(9, "012345678/") ++ "012345678", "foo" }, |
| 365 | }, |
| 366 | }; |
| 367 | |
| 368 | for (cases) |case| { |
| 369 | var header = Header.init(.regular); |
| 370 | try header.setPath(case.in[0], case.in[1]); |
| 371 | try testing.expectEqualStrings(case.out[0], std.mem.sliceTo(&header.prefix, 0)); |
| 372 | try testing.expectEqualStrings(case.out[1], std.mem.sliceTo(&header.name, 0)); |
| 373 | } |
| 374 | |
| 375 | const error_cases = [_]struct { |
| 376 | in: []const []const u8, |
| 377 | }{ |
| 378 | // basename can't fit into name (106 characters) |
| 379 | .{ .in = &.{ "zig", "test/cases/compile_errors/regression_test_2980_base_type_u32_is_not_type_checked_properly_when_assigning_a_value_within_a_struct.zig" } }, |
| 380 | // cant fit into 255 + sep |
| 381 | .{ .in = &.{ "prefix", repeatString(22, "0123456789/") ++ "basename" } }, |
| 382 | // can fit but sub_path can't be split (there is no separator) |
| 383 | .{ .in = &.{ "prefix", repeatString(10, "0123456789") ++ "a" } }, |
| 384 | .{ .in = &.{ "prefix", repeatString(14, "0123456789") ++ "basename" } }, |
| 385 | }; |
| 386 | |
| 387 | for (error_cases) |case| { |
| 388 | var header = Header.init(.regular); |
| 389 | try testing.expectError( |
| 390 | error.NameTooLong, |
| 391 | header.setPath(case.in[0], case.in[1]), |
| 392 | ); |
| 393 | } |
| 394 | } |
| 395 | }; |
| 396 | |
| 397 | test { |
| 398 | _ = Header; |
| 399 | } |
| 400 | |
| 401 | test "write files" { |
| 402 | const files = [_]struct { |
| 403 | path: []const u8, |
| 404 | content: []const u8, |
| 405 | }{ |
| 406 | .{ .path = "foo", .content = "bar" }, |
| 407 | .{ .path = repeatString(10, "a12345678/") ++ "foo", .content = repeatString(511, "a") }, |
| 408 | .{ .path = repeatString(24, "b12345678/") ++ "foo", .content = repeatString(512, "b") }, |
| 409 | .{ .path = repeatString(25, "c12345678/") ++ "foo", .content = repeatString(513, "c") }, |
| 410 | .{ .path = repeatString(51, "d12345678/") ++ "foo", .content = repeatString(1025, "d") }, |
| 411 | .{ .path = repeatString(11, "e123456789"), .content = "e" }, |
| 412 | }; |
| 413 | |
| 414 | var file_name_buffer: [std.fs.max_path_bytes]u8 = undefined; |
| 415 | var link_name_buffer: [std.fs.max_path_bytes]u8 = undefined; |
| 416 | |
| 417 | // with root |
| 418 | { |
| 419 | const root = "root"; |
| 420 | |
| 421 | var output: Io.Writer.Allocating = .init(testing.allocator); |
| 422 | var w: Writer = .{ .underlying_writer = &output.writer }; |
| 423 | defer output.deinit(); |
| 424 | try w.setRoot(root); |
| 425 | for (files) |file| |
| 426 | try w.writeFileBytes(file.path, file.content, .{}); |
| 427 | |
| 428 | var input: Io.Reader = .fixed(output.written()); |
| 429 | var it: std.tar.Iterator = .init(&input, .{ |
| 430 | .file_name_buffer = &file_name_buffer, |
| 431 | .link_name_buffer = &link_name_buffer, |
| 432 | }); |
| 433 | |
| 434 | // first entry is directory with prefix |
| 435 | { |
| 436 | const actual = (try it.next()).?; |
| 437 | try testing.expectEqualStrings(root, actual.name); |
| 438 | try testing.expectEqual(std.tar.FileKind.directory, actual.kind); |
| 439 | } |
| 440 | |
| 441 | var i: usize = 0; |
| 442 | while (try it.next()) |actual| { |
| 443 | defer i += 1; |
| 444 | const expected = files[i]; |
| 445 | try testing.expectEqualStrings(root, actual.name[0..root.len]); |
| 446 | try testing.expectEqual('/', actual.name[root.len..][0]); |
| 447 | try testing.expectEqualStrings(expected.path, actual.name[root.len + 1 ..]); |
| 448 | |
| 449 | var content: Io.Writer.Allocating = .init(testing.allocator); |
| 450 | defer content.deinit(); |
| 451 | try it.streamRemaining(actual, &content.writer); |
| 452 | try testing.expectEqualSlices(u8, expected.content, content.written()); |
| 453 | } |
| 454 | } |
| 455 | // without root |
| 456 | { |
| 457 | var output: Io.Writer.Allocating = .init(testing.allocator); |
| 458 | var w: Writer = .{ .underlying_writer = &output.writer }; |
| 459 | defer output.deinit(); |
| 460 | for (files) |file| { |
| 461 | var content: Io.Reader = .fixed(file.content); |
| 462 | try w.writeFileStream(file.path, file.content.len, &content, .{}); |
| 463 | } |
| 464 | |
| 465 | var input: Io.Reader = .fixed(output.written()); |
| 466 | var it: std.tar.Iterator = .init(&input, .{ |
| 467 | .file_name_buffer = &file_name_buffer, |
| 468 | .link_name_buffer = &link_name_buffer, |
| 469 | }); |
| 470 | |
| 471 | var i: usize = 0; |
| 472 | while (try it.next()) |actual| { |
| 473 | defer i += 1; |
| 474 | const expected = files[i]; |
| 475 | try testing.expectEqualStrings(expected.path, actual.name); |
| 476 | |
| 477 | var content: Io.Writer.Allocating = .init(testing.allocator); |
| 478 | defer content.deinit(); |
| 479 | try it.streamRemaining(actual, &content.writer); |
| 480 | try testing.expectEqualSlices(u8, expected.content, content.written()); |
| 481 | } |
| 482 | try w.finishPedantically(); |
| 483 | } |
| 484 | } |
| 485 | |
| 486 | /// Marked `inline` to avoid unnecessary binary float, since arguments are always comptime-known. |
| 487 | inline fn repeatString(comptime n: usize, comptime str: []const u8) []const u8 { |
| 488 | const buf: [n][str.len]u8 = @splat(str[0..str.len].*); |
| 489 | return @ptrCast(&buf); |
| 490 | } |