archive changelog | Dart package

4.6.0 #

BREAKING CHANGE: All decoders (XZDecoder, TarDecoder, ZipDecoder, GZipDecoder, ZLibDecoder, BZip2Decoder and ZstdDecoder) and their wrappers like extractFileToDisk, extractArchiveToDisk, extractArchiveToDiskSync now share the same contract: by default no damaged data errors are thrown anymore. If error is needed, specify throwOnError: true to decodeBytes or decodeStream. If checksum verification is required, specify verify: true to decodeBytes or decodeStream: throwOnError defaults to verify, so a mismatch or damage throws. With verify: true, throwOnError: false it is reported like no flags do, decodeStream returns false and decodeBytes returns partial data. All throwOnError are reported as ArchiveException and checksum exceptions with ArchiveChecksumException that extends ArchiveException. throwOnError is now bool? in these signatures, so a subclass that overrides them has to follow. Without flags, a damaged gzip or zlib stream from dart:io now decodes without an error, where 4.5.x threw because dart:io always checks the trailer. Errors such as RangeError from damaged data are reported the same way as other damage, where 4.5.x rethrew some of them as ArchiveException. Read README for migration tips and exceptions that are still reported with original type (everything unrelated to damaged data and package code).
Unlike decoders mentioned above, all new converters APIs mentioned beneath use both verify and throwOnError by default due to the specification of their format so that they completely coincide with dart zlib and gzip transformers.
BREAKING CHANGE: All decoders now properly handle verify: true param and do not ignore it anymore. GZipDecoder and ZLibDecoder that rely on dart:io implementation still may not catch the checksum issue and report an error with throwOnError for some damaged archives: a gzip whose damaged data decodes longer than its declared size passes as a concatenated gzip, a checksum error inside a concatenated gzip is reported as ArchiveException, and a truncated zlib stream is found only with verify, and thrown as ArchiveException, as a truncated gzip is. GZipDecoderWeb and ZLibDecoderWeb report all of these. Cut raw deflate stream (raw: true) is not found by any flags there, it has no checksum.
BREAKING CHANGE: Added writeRange and reserve to OutputStream, and readInto and viewBytes to InputStream. All four have a default body, so a class that extends them needs no change. A class that implements them has to add these methods.
Fixed extractFileToDisk detecting files only by extension. It now also detects them by their header bytes as all platform decoders do (like MacOS, Windows, Linux). If detection by headers fails it uses old file extension format instead.
Fixed ZipDecoder and extractFileToDisk throwing Exception for an incorrect password, or returning wrong data for ZipCrypto. They now throw ArchivePasswordException that extends ArchiveException. This exception overrules throwOnError and is thrown always when password does not match. AES MAC is checked now, a mismatch throws ArchiveMacException that extends ArchiveChecksumException regardless of flags.
extractArchiveToDisk and extractArchiveToDiskSync throw with ArchivePasswordException as well for bad password case.
Added ArchiveLimitException that extends ArchiveException. It is thrown regardless of verify and throwOnError when decoded data is larger than maxOutputSize, when extracted files are larger than maxSize, when a zstd window is larger than maxWindowSize, when an xz block declares a dictionary larger than maxDictionarySize, when a tar long name, PAX header or sparse map is larger than TarFile.maxMetadataSize, and when a compressed zip entry decodes to more than its declared uncompressed size.
Fixed listTarFiles and extractTarFiles silently using only the readable part of a damaged archive. They now throw an error for it.
BREAKING CHANGE: lastModTime of zip entries is now in seconds since epoch, as in tar, instead of the DOS date and time format. The DOS value it held before, date << 16 | time, is now in ArchiveFile.lastModDosTime.
BREAKING CHANGE: extractFileToDisk now accepts verify and throwOnError, extractArchiveToDisk and extractArchiveToDiskSync accept throwOnError. Without them a damaged or cut archive no longer throws: Complete entries are extracted, and damaged ones leave no partial files on disk. With throwOnError the function throws an exception upon encountering the first corrupted entry, and verify adds checksum checks to it.
BREAKING CHANGE: CompressionType has new values, lzma and xz, so a switch over it needs a case for each of them.
Fixed TarEncoder writing a tar hard link read by TarDecoder as a symlink, which pointed to a missing file. It now writes it back as a hard link.
Deprecated AesDecrypt, use the exported AesCtr instead. AesDecrypt is now a wrapper over AesCtr and gives the same output as before. Sha1, HmacSha1, pbkdf2HmacSha1 and Sha256 are exported as well.
Fixed ZstdDecoder throwing ArgumentError for a malformed dictionary. It throws ZstdDictionaryException, which extends ArchiveException, since a dictionary can be damaged data as well as a wrong argument.
Fixed ZstdEncoder throwing ArgumentError for a level above 22 or below zstdMinLevel. It clamps the level to that range, as ZSTD_c_compressionLevel of libzstd does. zstdMinLevel is -131072 instead of -7, the value of ZSTD_minCLevel().
Fixed ZstdDecoder accepting a maxWindowSize below 1024 bytes, the smallest zstd window. It throws ArgumentError for it.
BREAKING CHANGE: ArchiveFile.lastModDateTime returns a local DateTime built from lastModTime in seconds since epoch, instead of a UTC DateTime built from DOS date and time fields. ZipDecoder takes the time from the UT extra field (0x5455) when the entry has one, so a zip written by ZipEncoder reads back the same time in every time zone.
Fixed ZipDecoder returning the content of the first of several entries with the same name with the metadata of the last one. It keeps the last one, as 3.x did.
BREAKING CHANGE: extractFileToDisk calls callback for a zip entry after the entry is written, as it does for tar. It used to call it for every zip entry while the central directory was read, before any file was written.
Fixed extractFileToDisk extracting junk entries from a gzip, bzip2, xz or zstd file that does not hold a tar, also when its name ends with .tar.gz or a similar extension. It throws ArgumentError for such a file.
Fixed BZip2Decoder returning only the first stream of concatenated bzip2 files, as pbzip2 writes them, without an error. It decodes every stream in sequence, as bzip2 -d does.
Fixed ZLibDecoderWeb appending the streams after the first of concatenated zlib streams, matching dart:io of Dart 3.5.0, which was reverted as a bug. It decodes only the first and ignores the bytes after it, as ZLibDecoder on dart:io does since Dart 3.5.2 (dart-lang/sdk#56481).
extractArchiveToDisk writes entries synchronously, as extractArchiveToDiskSync does, and still returns a Future.
Added allowAbsoluteSymlinks to extractFileToDisk, extractArchiveToDisk and extractArchiveToDiskSync. Absolute symbolic links are still skipped by default.
Extraction removes \\?\, \\.\ and \\?\UNC\ prefixes, drive letters and leading slashes from entry names, as bsdtar does. An entry whose name is an absolute path used to be skipped.
Extraction skips an entry whose path runs through a symbolic link that is already in the output directory and points outside it, and keeps a symbolic link from the archive instead of turning it into a directory when a later entry has a path through it.
TarDecoder with verify and tarCodec.decoder throw for a header whose size field is not a number.
ZipEncoder writes the CRC and sizes in the local header of a stored entry taken from another zip, which it used to write with a data descriptor (#421).
Added Dart async* StreamTransformer/ByteConversionSink/Converter support (xzCodec, zstdCodec, bzip2Codec, tarCodec and zipCodec) for decode and encode (decoders for all formats except for zip). xz, zstd and bzip2 are Converters, bytes to bytes; tar and zip are StreamTransformers: TarDecoderTransformer, TarEncoderTransformer and ZipEncoderTransformer.
Added CodecsRecognizer with ArchiveFormat to recognize compressed file types with their header bytes.
Made CRC32, SHA-256, CRC64 and Adler32 significantly faster and less RAM-consumable on all platforms (for verify: true), and SHA-1 with its HMAC and PBKDF2 and AES for AES zip.
ZipCrypto decryption is 6.6-8.6x faster now: its keys are ints that stay exact on web instead of BigInt.
Added ProgressOutputStream to monitor progress during unpack.
Added usage examples of new apis is package readme file, shrinked amount of docs for XZDecoder.
Added XZMultithreadOptions.converter and ZstdMultithreadOptions.converter, the options a Converter takes for multithreaded decoding/encoding via converter.
Added ultra to ZstdEncoderConverter and ZstdCodec. Without contentSize levels 20-22 are reduced to 19 unless ultra: true is given, as zstd does without --ultra: level 22 of unknown size takes 655 MB even for 100 bytes of input. ZstdEncoder knows the input size and takes these levels without the flag.
CRC64 for XZDecoder now works both on io and web (dart2-js) (with new Crc64 class). Old api isCrc64Supported remained unchanged and still returns false and throws with getCrc64 on dart2-js.
Significantly reduced XZDecoder RAM usage for --x86, crc32, crc64, SHA-256 and default decoding with decodeStream.
Reduced RAM usage used by GZipDecoder.decodeBytes and ZLibDecoder.decodeBytes on dart:io
Slightly reduced amount of used RAM by ZLibDecoder.
Added blockSize100k to BZip2Encoder.encodeBytes and encodeStream
Added ArchiveFile.isHardLink, set by TarDecoder for tar hard links. symbolicLink is unchanged.
Fixed symlinks scopes for extractFileToDisk that could overwrite existing file on disk outside of output dir
Fixed Unix mode reading for ZipDecoder for archives created on Windows.
Fixed ZipDecoder reading a zip entry name that is not valid UTF-8 as Latin-1. It reads it as CP437, the encoding zip tools use when the UTF-8 flag is not set, so such names list differently than in 4.5.1. Pass ZipDecoder(filenameEncoding:) to read them in another encoding.
Fixed ZipDecoder ignoring the Info-ZIP Unicode Path extra field (0x7075). An entry that has one lists under the UTF-8 name from that field, so names differ from 4.5.1. The field is used only when its CRC matches the stored name.
Fixed CRC64 XZEncoder files created on dart-js failed to decode on native platforms.
Fixed ZLibDecoder on web.
TarEncoder now writes "ustar\0" magic in header for easier recognition.
TarEncoder now writes non-ASCII names and symlink targets as PAX path/linkpath records, so readers that take header bytes in a local code page, such as Windows tar.exe, get the right names.
Zip passwords are now encoded as UTF-8 to properly handle non-ASCII passwords. ZipDecoder tries UTF-8 first and then bytes that older versions of the package wrote.
Fixed ZipDecoder decrypting AES and ZipCrypto entries inside the buffer, what corrupted it.
Fixed(added) LZMA decoding for ZipDecoder and encoding for ZipEncoder with CompressionType.lzma. Encoder still falls back to DEFLATE, but decoder properly reads lzma data.
ZipDecoder now reads zstd (method 93) and xz (method 95) zip entries as CompressionType.zstd and CompressionType.xz, and ZipEncoder writes them with ZstdEncoder and XZEncoder. XZEncoder stores data uncompressed, so an xz entry is not smaller than its data.
TarDecoder and tarCodec.decoder now read sparse files the way libarchive does: old GNU sparse headers with extension blocks and PAX sparse formats 0.0, 0.1 and 1.0. Holes are expanded when the content is read.
Fixed mode 0 file rights for ArchiveFile from ZipDecoder created on Windows.
Fixed SHA-256 checks for XZDecoder with verify: true.
Fixed symlinks encoding in ZipEncoder and decoding non-ASCII names with macos unzip from archives of this package.
Fixed latin1 encoding for TarEncoder with non-ASCII file names.
Fixed ArchiveFile.directory mode (0755 instead of 0644) so that it can be opened after unpacking.
Fixed a few cases with accepting corrupted file in XZDecoder.
Fixed some 7z files failed to decode with XZDecoder.
Fixed invalid bzip2 archive returned true from decodeStream.
Fixed BZip2Decoder throwing RangeError on an archive cut inside a block when it read from an InputMemoryStream. Both memory and file streams return false now, without inconsistency.
Fixed a zip entry compressed with bzip2 handing back short data with no error when the entry is damaged. It throws ArchiveException now, the way a damaged deflate entry already did.
Fixed file names trimming in TAR.
ZipEncoder now applies CompressionType.none to entries with no content.
Replaced ChunkedConversionSink with ZLibOutputSink for gzip and zlib to forward every piece the codec produces instead of holding it to the end.
Fixed tar stream decoder allocating whatever a GNU long name or PAX header declared, before any of those bytes arrived. A 512 byte header claiming 2^40 was an out of memory kill; the buffer now grows with what actually arrives.
Fixed ZipEncoder with a password writing a stale MAC into entries that have no content, such as every directory. The local header declared 12 bytes it never wrote, which left every local header after it off by 2 and unreachable to a forward-only reader.
Added ZipEncoder(streamed: true) that writes entries as they are read, with low RAM usage on big files (4GB+, uses zip64 format from zip APPNOTE).
Fixed decoding zip64 with ZipDecoder.
Multithreaded XZDecoder.decodeStream now stops at the first broken block, like the single threaded one.
Fixed ZipDecoder returning an empty archive for a zip with extra bytes in front of it (self extracting zips).
Fixed ZipDecoder throwing RangeError instead of ArchiveException for some damaged zips.
Fixed LzmaDecoder throwing RangeError for truncated or damaged LZMA data, and XZDecoder reading more bytes than the input has for a stream cut inside a stream header, block header, LZMA2 chunk header, index or footer. They throw ArchiveException and report a truncated stream.
LzmaDecoder stops at the end marker of an LZMA stream of unknown size, as NSIS installers store them, instead of failing. endMarkerSeen tells whether it did. XZDecoder rejects an LZMA2 chunk that has one, which LZMA2 does not allow.
With verify, a wrong CRC in an xz stream header, block header, index or footer, and a wrong gzip header checksum (FHCRC), throw ArchiveChecksumException instead of ArchiveException. GZipDecoderWeb checks the header checksum.
Fixed TarDecoder adding garbage folders to names in archives made by GNU tar --incremental.
Fixed TarDecoder treating some folder entries (GNU and very old tar formats) as files.
Fixed TarDecoder long names in archives made by older versions of this package.
Fixed BZip2Decoder rejecting some valid archives and data after the end of an archive, same as bzip2 1.0.8.
Fixed XZDecoder failing on archives larger than 4 GB on the web.
Fixed InputFileStream.toUint8List and readBytes returning wrong bytes when a file is read in parts.
GZipDecoderWeb and ZLibDecoderWeb are ~60x faster when writing to a file now.
Fixed ZipEncoder emptying ArchiveFile.stream entries, so they could not be read after encoding.
Fixed ZipEncoder with a password changing the original data of entries in memory.
Fixed extractFileToDisk never returning when the output drive does not exist.
Fixed multithreaded XZDecoder hanging on a truncated file and accepting some damaged archives that single threaded decoding refused.
Fixed BZip2Decoder crashing on an archive that declares block size 0.
Fixed ZipDecoder with verify: true rejecting AES zips that store no CRC.
Fixed ZipEncoder writing a wrong CRC when encoding again an archive decoded from an AES zip
Fixed a stored zip entry reading back empty after it was written to a file.
ArchiveFile.stream now takes the stream from its current position instead of from the start.
Fixed ZipEncoder failing to create on the web.
TarDecoder with verify: true now reports a file that is cut short, like other tar tools do.
Fixed ZipEncoder zip64 for small files that follow large ones and for entries over 4 GB.
Fixed InputFileStream ignoring its byteOrder for readUint16, readUint24, readUint32 and readUint64.
Fixed ZLibDecoderWeb and GZipDecoderWeb depending on the byte order of the input stream: decodeStream failed verify on a whole zlib stream from a little-endian stream, and gzip failed on a big-endian one.
Fixed not-restoring original InputStream endian for some decoders and encoders and decoding/encoding BigEndian files.
Fixed ZLibDecoderWeb dropping the decoded data when bytes follow the stream, and accepting an empty input.
Fixed ZipDecoder returning the raw compressed bytes as the content of an entry with an unsupported compression method. With throwOnError or verify reading that entry throws ArchiveException, without them it is empty, and extractFileToDisk skips it and extracts the rest.
Fixed an empty zip password encrypting and decrypting entries with an empty key.
Fixed ZipDecoder allocating size of a deflate entry over 500 MB, which threw OutOfMemoryError for a zip64 entry that claims terabytes.
Fixed listTarFiles listing a plain .tar as empty, and listTarFiles and extractTarFiles leaving open files and temporary folders uncleaned.
Fixed ZLibDecoderWeb and GZipDecoderWeb accepting malformed deflate: a zlib header with a method other than deflate or a window over 32 KB, a stream without its final block, and an oversubscribed Huffman table, which decoded to garbage before, are now reported with throwOnError and verify.
Fixed ZipFileEncoder.store writing entries as deflate level 0 and holding each whole entry in memory: a 1 GB file needed 2 GB of RAM. The store level of create, zipDirectory, addFile, addFileSync and addDirectory now writes uncompressed entries (method 0) streamed from the file. An ArchiveFile given to addArchiveFile is stored too, unless its compression is set.
Fixed ZipFileEncoder holding each whole deflated entry in memory: a 920 MB file needed 1042 MB of RAM at level 1 and 667 MB at level 6, now 19 MB at every level. ZipFileEncoder now uses ZipEncoder(streamed: true), so a deflated entry without a password has bit 3 set, zero CRC and sizes in its local header, and a data descriptor after its data. The archives decode the same, but their bytes differ from earlier versions.
Fixed OutputFileStream.writeStream moving the read position of its source to the end, unlike the other OutputStream types. The source position is now left unchanged.
ZipDirectory.read no longer reads every entry's local header. A ZipFileHeader reads its file when it is first asked for, and leaves the stream where it was. ZipDecoder reads each entry's file as before. Code that uses ZipDirectory directly has to keep the stream open until it has read file, and gets an error in a local header from file instead of from read.
Added ZipDirectory.headers, which reads the central directory one entry at a time as it is iterated, so a caller that walks a large archive holds one header at a time. ZipDirectory.read collects it into fileHeaders as before. With files: false the headers keep no reference to the stream, for a caller that reads entries itself.

4.5.2 #

Fixed ZipDecoder.decodeStream returning an empty archive when the input was not at position 0, such as after reading a magic number or decoding the same stream a second time (#405). decodeStream also leaves the input's read position where it was.
Added InputStream.totalLength, the size of the whole stream whatever the read position, where length is only the bytes after it.

4.5.1 #

Fixed zips written on the web, which decoded to nothing: the zlib encoder left the zip encoder's output big-endian, so every header after the first entry's data was written in the wrong byte order (#419).
ZipEncoder can be used under dart2js on Node, where creating its secure random source threw; it is now only created for a password.
Fixed zips written by ZipEncoder since 4.4.0, which 7-Zip reported as having header errors: the data descriptor flag was set in each entry's local header but not its central directory header. The UTF-8 flag also went into the central directory even for names that were not UTF-8.
Directory entries are stored rather than written as an empty deflate stream, which unzip rejected, and carry the MS-DOS directory attribute. In a zip with a password they are no longer marked encrypted.
Entries whose content is in memory, and stored entries that can be read twice, have their CRC and sizes in the local header again rather than in a data descriptor after the data. Java's ZipInputStream refuses a stored entry with a descriptor, and OpenDocument and EPUB files need their stored mimetype entry without one (#421).

4.5.0 #

Security: XZDecoder hung forever on a truncated .xz read from a file or a decode stream, including .tar.xz in extractFileToDisk.
Security: tar long name and pax header entries are limited to TarFile.maxMetadataSize (1 MB). A small .tar.gz could declare one large enough to take gigabytes of memory to read.
Security: decoding had no output limit, so a small archive could expand to fill memory or disk. Added maxOutputSize to decodeBytes and decodeStream on GZipDecoder, ZLibDecoder, BZip2Decoder and ZstdDecoder, and maxSize to extractFileToDisk, extractArchiveToDisk and extractArchiveToDiskSync.
A zip entry is no longer decoded past the uncompressed size the archive gives for it, so ArchiveFile.size can be checked before reading.
Security: ZstdDecoder sized its window and output from the frame header, so a 10 byte input committed hundreds of MB, even past maxWindowSize. The content size is now checked as each block decodes, and one that overflows is refused.
Security: XZDecoder limits the LZMA2 dictionary an archive may declare to maxDictionarySize (256 MB by default), which bounds the memory a streaming decode takes. It also takes maxOutputSize, and no longer preallocates gigabytes because a tiny archive's index claims them.
Multithreaded xz decoding checks each block against the size the index gives it. An index that misstated one used to make the ordered output hold everything after it in memory and report success with the output cut short. Blocks are no longer let run far ahead of the output.
Security: extraction creates symbolic links after every other entry and refuses one whose path runs through another link. A chain of links that each looked safe on its own could place a file outside the output directory.
Security: extractTarFiles did not keep entries inside the output directory. It now extracts as extractFileToDisk does, decoding as it reads rather than through a temp file it left behind.
Extraction reports an entry that fails to decode instead of leaving a partial file without a word, and an empty file no longer takes a 1 MB write buffer.
A tar hard link, which is extracted as a symbolic link, now points at its target from the link's own directory.
TarDecoder.decodeStream and decodeBytes take keepEntries. False passes entries only to the callback, so memory no longer grows with the entry count; extractFileToDisk uses it. A .tar.gz of a million empty entries now decodes in 39 MB rather than 326 MB.
Malformed zip headers, such as an odd length extra field on an encrypted entry or a zip64 locator pointing past the file, threw RangeError instead of ArchiveException. The AES extra field is also found when other extra fields come before it.
BZip2Decoder accepts a block declaring more than 18002 selectors and ignores the extra ones, as libbzip2 does, instead of reporting a truncated stream.
Reading a zip entry no longer preallocates its declared size when the compressed data could not expand that far.

4.4.0 #

Security: a crafted zip64 compressed size overflowed the stream bounds checks and could crash the decoder while reading an untrusted archive. Sizes are now clamped to the bytes actually present.
Added InputDecodeStream and decodeLazy on GZipDecoder, ZLibDecoder, BZip2Decoder, ZstdDecoder and XZDecoder, which decode a stream as it is read so a large archive can be piped into another decoder without a temp file.
extractFileToDisk extracts .tar.gz, .tar.bz2, .tar.xz and .tar.zst in a single pass, writing each entry as it is reached, with no temp file.
Decoding and encoding no longer buffer the whole file: gzip and zlib streaming, ZipEncoder, and encrypted zip entries (ZipCrypto and AES) now work in a bounded amount of memory whatever the file size.
Encrypted zip entries are decrypted as they are read; a wrong password now surfaces when the content is read.
Faster decoding: getCrc32 is about three times quicker on typed data, and InputFileStream reads in larger blocks.
Fixed XZEncoder, which produced archives other tools rejected for inputs of 128 bytes or more.

4.3.0 #

Added ZstdDecoder, a pure Dart decoder for the zstd (Zstandard) format.
Added ZstdEncoder, a pure Dart zstd encoder for the VM, JavaScript and wasm, at levels -7 to 22.
Zip entries compressed with zstd.
extractFileToDisk and the tar command accept .tar.zst and .tzst, and TarFileEncoder.tarDirectory can write .tar.zst with compression: TarFileEncoder.zstd.
Added multithreaded decoding to XZDecoder. Passing an XZMultithreadOptions to decodeBytes or decodeStream spreads the work over isolates, one xz block per job, and reports the result through its onDone callback. Both methods behave exactly as before when it is omitted. On a 1.1 GB archive of six blocks: 16.1 s single threaded, 8.0 s on the default three workers, 5.0 s on six.
decodeStream reading from an InputFileStream now lets each worker read its own block straight from disk, so the compressed archive never passes through the calling isolate. Decoding a 1.1 GB archive to an OutputFileStream peaks at 1.8 GB, below the 3.0 GB the single threaded path uses, while being twice as fast.
Multithreaded decoding falls back to the single threaded path on the web, where there are no isolates, and the isolate machinery is tree-shaken out of web builds entirely.
Added InputFileStream.fileBuffer, fileOffset and fileLength.
Added --x86 flag support to XZDecoder
Improved verify: true speed for XZDecoder
Improved overall decode speed for XZDecoder
Decreased RAM usage for XZDecoder
Added concatenated streams support for XZDecoder
Added crc64 wasm support (verify: true)
Added uncompressedSize getter for XZDecoder, returning the original file size before compression.
Fix: pb=4 flag range error in XZDecoder
Fix: padding for _streamStart in XZDecoder
Added throwOnError to XZDecoder.decodeBytes and decodeStream. Without it a malformed or truncated archive still returns the partial output with nothing to say it is not the whole file, which was the only decode with no way at all to report a failure. In multithreaded mode the exception is delivered to XZMultithreadOptions.onError, and setting throwOnError without an onError is now refused rather than losing the failure.
Fix: a corrupt xz block lost the part of itself that had already decoded when verify: true was used with an output that cannot be read back, such as an OutputFileStream or any multithreaded decode. The output now stops in the same place whichever way the decode was asked for.
Added XZDecoder.maxPreallocateSize, capping how large an output buffer is allocated from a size the archive itself declares. Defaults to xzDefaultMaxPreallocateSize, 2 GB natively and 256 MB on the web.
Added XZMultithreadOptions.fileReadBufferSize.
Fix: an xz --check=none archive with corrupt blocks passed verify: true.
Fix: the xz block header reader trusted its own fields, accepting an unterminated multibyte integer, a filter properties length past the end of the header, empty delta or LZMA2 properties, and a bad stream header or footer CRC.
Fix: the zip End of Central Directory record was missed when its signature straddled a chunk of the backwards search, or sat within the last 21 bytes of the file, so a valid archive could decode as empty.
TarDecoder with verify: true now checks each header's own checksum and throws ArchiveException on a mismatch.
Fix: GNU base-256 numeric tar header fields were read as octal, so sizes of 8 GB or more and large uid, gid and mtime values decoded as garbage.
Fix: tar PAX extended headers are walked by each record's declared length instead of split on newlines, so a record holding binary no longer corrupts the rest of the header.
Fix: a PAX size record was applied to the metadata headers following it rather than to the entry it describes.
Fix: a GNU ././@LongLink header of type K set the next entry's name instead of its symlink target.
Fix: storeData: false skipped the tar long name and PAX headers along with the file data, losing the names they carry.
Fix: a negative size in a tar header is refused instead of read as an entry.
Fix: with verify: true the end of a tar is a full 512 byte zero block, which two zero bytes could not be told apart from a damaged header. A tail shorter than a header block now ends the archive instead of decoding as junk.
Fix: tar base-256 fields wider than TarFile.maxNumericField are refused on read and write instead of overflowing silently.
Fix: decoded regular tar files were given an empty link name and re-encoded as symlinks.
Fix: long tar names and symlink targets were written without the L and K type flags, so other tar implementations truncated them to 100 bytes, and long symlink targets were not written at all.
Fix: the GNU long name record was sized in characters rather than encoded bytes and ignored filenameEncoding.
Fix: encoding a tar entry consumed the stream it was given, so a second encode of the same archive wrote empty entries.
Fix: tar values too wide for the octal header field are written in GNU base-256 form instead of producing a corrupt header.
TarEncoder copies a stored stream through in chunks rather than reading the whole entry into memory.
TarFile.content is now settable and reads back what was set.
Fix: GZipDecoder.decodeStream returned true for a truncated or corrupt archive, silently losing files.
Fix: the web gzip decoder never checked the member CRC32 despite verify: true.
Fix: the web gzip header reader trusted its own length fields, reading past the end on a damaged extra field, an unterminated name or comment, or a header under ten bytes.
Fix: a gzip input under 20 bytes had header bytes read as a trailer.
The native gzip decoder now reads in 8 KB chunks instead of 1 KB.
Fix: an incomplete Huffman code table entry consumed no bits and looped forever, and a code length repeat past the end of the table indexed out of range.
Fix: an Inflate back-reference distance reaching past everything written indexed behind the output buffer.
Fix: the web zlib decoder read its two byte header without checking two bytes were there.
Fix: a short read near the end of a file overwrote the buffer's size with the bytes read, so every later read missed the cache and went back to disk.
Fix: multi-byte FileBuffer reads refilled the cache one byte early, costing a re-read on every buffer-aligned access.
Fix: InputFileStream.subset did not clamp its length to what remains of the source, so peeking past the end returned stale bytes.
Deprecated the fileSize parameter of the FileBuffer read methods, which is ignored.

4.2.0 #

Optimize performance and issues with large files with XZDecoder.

4.1.0 #

Fix: RangeError when parsing ZIP extra fields with trailing bytes.
Optimize LZ77 decode
Fix: issue with file_buffer preventing last value
Fix: invalid path separator in zip content compression

4.0.9 #

Fix extractFileToDisk file extension handling, where foo.bar.zip would fail.

4.0.8 #

Remove dependency to crypto package
Removed Adler32 and Crc32 classes
Fix: extractArchiveToDisk extensions should be case insensitive.
Fix: extractFileToDisk for .tar.gz files
Add ZipFileEncoder.addDirectorySync
Always use posix separators for ZipFileEncoder

4.0.7 #

Change posix dependency to 6.0.2.

4.0.6 #

Fix zip decoding when the last file of the archive is also a zip.
Add lastModifiedDateTime to ArchiveFile
Fix Archive files and fileMap getting out of sync after calling removeFile

4.0.5 #

Improve performance of OutputFileStream.
Add ArchiveFile.noCompress, which had been removed from the 3.x to 4.x update.
GZipDecoder should fall back to ZLib if there is no GZip header.

4.0.4 #

Fix level argument for ZipEncoder.add method.
Fix ArchiveFile.compression to work for controlling compression method used encoding zips.
Add ArchiveFile.compression CompressionType.bzip2 compression mode support.
Add ArchiveFile.compressionLevel to work for controlling compression level used for encoding zips.

4.0.3 #

Fix potential infinite loop when parsing zip headers.
Update conditional imports to be compatible with WASM.
Add addFileSync to ZipFileEncoder for synchronous call to addFile.

4.0.2 #

Reduce SDK min version to 3.0.
Fix import error with js_interop.

4.0.1 #

Fix error with GZip encoder for web builds.

4.0.0 #

Major cleanup of the code, includes potential breaking changes.
- decodeBuffer has been renamed to decodeStream in the various decoder classes.
- InputStream has been renamed to InputMemoryStream.
- OutputStream has been renamed to OutputMemoryStream.

3.6.1 #

Fix ArchiveFile.rawContent returning null after decoding a zip.

3.6.0 #

Fix zip encoding when a file was previously decoded.
Fix decoding zips with password when using InputFileStream.
ZipEncoder.encode autoClose now defaults to false.

3.5.1 #

Re-add zipPath to ZipFileEncoder.

3.5.0 #

Remove dependency to pointycastle package
Use utf8 encoding for string data
Fixes for encrypted zip encoding
Async and sync versions of extractArchiveToDisk

3.4.10 #

Fix ZipCrypto decryption

3.4.9 #

Revert breaking change for extractArchiveToDisk becoming async; add extractArchiveToDiskAsync for the async version.

3.4.8 #

Improve zip decompression performance with dart:io by using native ZLib decompression when possible.

3.4.7 #

Improve performance by not using List.setRange for copying bytes, which turns out to be very slow.

3.4.6 #

Fix for Zip64 file size causing memory errors.

3.4.5 #

Rewrote InputFileStream to reduce overall memory by using a shared file cache.
Added DateTime lastModDateTime getter to ArchiveFile.
Add support for zip encryption.

3.4.4 #

Fix for new default buffer size for InputFileStream consuming too much memory for large archives.

3.4.3 #

Fix bug in InputFileStream that caused it to only have an 8-byte buffer, making file streaming slow.
Increase the default buffer size for file I/O streams to 1MB.
Update pubspec dependency versions.

3.4.2 #

Add bzip2 decompression for zip files.

3.4.1 #

Fix for decoding zip64 zip files that have multiple extra fields.

3.4.0 #

Add Zip64 support to ZipEncoder to allow it to create zip files > 4GB.

3.3.9 #

Fix for extractFileToDisk causing corrupt files by closing a file stream before it finished writing.

3.3.8 #

Fix for zip security issue with symlinks, https://github.com/brendan-duncan/archive/issues/265. https://osv.dev/vulnerability/GHSA-9v85-q87q-g4vg.
Fix for zip security issue with file paths, https://github.com/brendan-duncan/archive/issues/266. https://osv.dev/vulnerability/GHSA-r285-q736-9v95.
Add progress callback for decoding zip files.
Don't allow tar files to include absolute paths.
Fix error decoding AES-192.

3.3.7 #

Add Zip AES-256 decryption
Fix symlink encoding for tar files

3.3.6 #

Fix errors decoding XZ files.

3.3.5 #

Fix file content when decoding zips

3.3.4 #

Fix analysis errors.

3.3.3 #

Support symlinks in ZIP archives
Fix ZIP decryption for ZipCrypto format

3.3.2 #

Fix for UTF-8 file name caused problem on Windows.

3.3.1 #

Fix for Inflate crashing on some compressed files.

3.3.0 #

IO encoders (ZipFileEncoder, TarFileEncoder), will now include directories and empty directories.
Fix for ZipEncoder file lastModTime.
Fix for ArchiveFile.string.
Add PAX format to tar decoder.
Make more file operations async.

3.2.2 #

Re-add List
Add String and TypedData (Int32List, Float32List, etc) content data for ArchiveFile.

3.2.1 #

Added buffer to OutputFileStream to improve performance by reducing the number of file writes.

3.2.0 #

For non-web applications, use native 'inflate' decompression when decompressing zip files.
Add asyncWrite option to extractArchiveToDisk and extractFileToDisk, moving file write operations to be async.
ArchiveFile.writeContent will release its memory after the data has been written, reducing overall memory usage.
Add clear method to ArchiveFile, clearing any decompressed data memory it's storing.

3.1.11 #

Fix indexing bug in Archive.addFile.

3.1.10 #

Fix performance regression with Archive.

3.1.9 #

Fix FileInputStream to work with ZipDecoder.

3.1.8 #

Catch invalid UTF8 string decoding.

3.1.7 #

Fix for UTF8 filenames

3.1.6 #

Fix problem with non-terminating long filenames.
File modification dates were incorrectly stored in milliseconds instead of seconds.

3.1.5 #

Disable XZ format CRC64 for html builds to fix errors.

3.1.4 #

Changed LICENSE to MIT.

3.1.3 #

Cleaned up LICENSE, moving other licenses to LICENSE-other.md.

3.1.2 #

Added the ability to override the timestamp encoded in a Zip file.

3.1.1 #

Fix zip encoder so that zip files created on Windows will open correctly on Linux.

3.1.0-dev #

Added const constructors to ZLibDecoder, ZLibEncoder, and ZLibDecoderBase.

3.0.0 #

Stable release supporting null safety.

3.0.0-nullsafety.0 #

Migrate to null safety.

2.0.13 #

Switch to dart strong mode; refactor code to resolve all dartanalyzer warnings.

2.0.12 #

Fix dartanalyzer warnings

2.0.11 #

Set the default permission for ArchiveFile to something more reasonable (0644 -rw-r--r--)

2.0.10 #

Fix for decoding empty zip files.

2.0.9 #

Add isSymbolicLink and nameOfLinkedFile to ArchiveFile.
Fix for encoding empty files.

2.0.8 #

Fix zip isFile

2.0.7 #

Fix zip file attributes.

2.0.6 #

Support GNU tar long file names
Maintain unix file permissions in zip archives.

2.0.5 #

Use dart:io ZLibCodec when run from dart:io.

2.0.4 #

Fix InputStream when a Uint8ListView is used as input data.

2.0.3 #

Use Utf8 for reading strings in archive archive files, for filenames and comments.

2.0.2 #

Fixes for ZipFileEncoder.

2.0.1 #

Remove the use of part and part of in the main library.
Added ZipFileEncoder to encode files and directories using dart:io.
Added createArchiveFromDirectory function to create an Archive object from a dart:io Directory.

2.0.0 #

Moved version up for Dart 2 support.
Fixed an issue with file compression flags when decoding zip archives.
Fixed an issue with bzip2 decoding in production code.

1.0.33 #

Support the latest version of package:args.

1.0.30 #

Add archive_io sub-package for supporting file streaming rather than storing everything in memory. This is a work-in-progress and under development.

1.0.29 #

Fix issue with POSIX tar files.
Upgrade dependency on archive to >=1.0.0 <2.0.0

1.0.20 #

Improve performance decompressing large files in zip archives.

1.0.19 #

Disable CRC verification by default when decoding archives.

1.0.18 #

Add support for encoding uncompressed files in zip archives.

1.0.17 #

Fix a bug in InputStream.

1.0.16 #

Add stream support to Inflate decompression.

1.0.15 #

Improved performance when writing large blocks.

1.0.14 #

Misc updates and fixes.

1.0.13 #

Added BZip2 encoder.
BREAKING CHANGE: File was renamed to ArchiveFile, to avoid conflicts with dart:io.

1.0.12 #

Added BZip2 decoder.

1.0.11 #

Changed InputStream to work with typed_data instead of List<int>, should reduce memory and increase performance.

1.0.10 #

Renamed InputBuffer and OutputBuffer to InputStream and OutputStream, respectively.
Added readBits method to InputStream.

archive 4.6.0 archive: ^4.6.0 copied to clipboard

Metadata