This version is still in development and is not considered stable yet. For the latest stable version, please use Compress4J 5.x!

Migrating from 4.x to 5.0

Each table lists the 4.x API on the left and its 5.0 replacement on the right. See Security for the new defaults in depth.

Safer defaults

4.x 5.0

4.x	5.0
`escapingSymlinkPolicy` defaulted to `ALLOW`	The default is `DISALLOW`: escaping symlinks throw `UnsafeEntryException`. Pass `escapingSymlinkPolicy(ALLOW)` for archives you produced yourself.
No default limits	`ExtractionLimits.defaults()`: 1,000,000 entries, expansion ratio 100, no size limits. `limits(ExtractionLimits.noLimits())` restores 4.x behaviour for trusted input.
`ArchiveExtractor.UNLIMITED`	`ExtractionLimits.UNLIMITED`
Any negative `maxEntries`, `maxEntrySize` or `maxTotalSize` meant unlimited	Only `-1` (`ExtractionLimits.UNLIMITED`) means unlimited; other negative values throw `IllegalArgumentException`. `maxRatio` rejects `0` and negatives other than `-1`.
`maxRatio` did not exist	Every reader enforces `maxRatio` (default 100) once it has produced 1 MiB: tar, zip (also `streaming`), 7z, ar, cpio, arj, dump and `Decompressor`. A breach throws `LimitExceededException` with limit `RATIO`; raise it with `maxRatio(…)` or `limits(…)`.
Path traversal and escaping symlinks threw a plain `IOException`, which the `errorHandler` could `SKIP`	`UnsafeEntryException`, a subclass of the sealed `UnsafeInputException`. No `errorHandler` can suppress it.
`ArchiveLimitExceededException`, which the `errorHandler` never saw	`LimitExceededException`, also an `UnsafeInputException`, with `limit()`, `maximum()` and `entryName()`; its message names the builder method that raises the limit.

escapingSymlinkPolicy defaulted to ALLOW

The default is DISALLOW: escaping symlinks throw UnsafeEntryException. Pass escapingSymlinkPolicy(ALLOW) for archives you produced yourself.

No default limits

ExtractionLimits.defaults(): 1,000,000 entries, expansion ratio 100, no size limits. limits(ExtractionLimits.noLimits()) restores 4.x behaviour for trusted input.

ArchiveExtractor.UNLIMITED

ExtractionLimits.UNLIMITED

Any negative maxEntries, maxEntrySize or maxTotalSize meant unlimited

Only -1 (ExtractionLimits.UNLIMITED) means unlimited; other negative values throw IllegalArgumentException. maxRatio rejects 0 and negatives other than -1.

maxRatio did not exist

Every reader enforces maxRatio (default 100) once it has produced 1 MiB: tar, zip (also streaming), 7z, ar, cpio, arj, dump and Decompressor. A breach throws LimitExceededException with limit RATIO; raise it with maxRatio(…) or limits(…).

Path traversal and escaping symlinks threw a plain IOException, which the errorHandler could SKIP

UnsafeEntryException, a subclass of the sealed UnsafeInputException. No errorHandler can suppress it.

ArchiveLimitExceededException, which the errorHandler never saw

LimitExceededException, also an UnsafeInputException, with limit(), maximum() and entryName(); its message names the builder method that raises the limit.

Behaviour change for 4.x users. An errorHandler that returned SKIP or SKIP_ALL used to continue past a traversal or escaping-symlink entry. In 5.0 that entry aborts the extraction with UnsafeEntryException. Code that relied on skipping such entries must stop doing so or filter them out before extraction.

DISALLOW, the default, re-checks every symlink an extraction created once it finishes. Links that escape through later entries are removed and extraction fails with UnsafeEntryException.
RELATIVIZE_ABSOLUTE rewrites absolute targets under the output directory, then rejects any target that still escapes. 4.x accepted relative escapes.
maxEntries counts entries after stripComponents and the filter. Skipped unsupported or filtered entries are not counted, and the unsupportedEntryHandler runs once per skipped entry without a bound.
A highly compressible but legitimate archive, such as a log archive, can exceed ratio 100. The exception names maxRatio.
Reading .xz and .lzma without memoryLimitKiB is now capped at 256 MiB, in Decompressor and TarArchiveExtractor. memoryLimitKiB(Integer.MAX_VALUE) lifts the cap.
The public stream-taking constructors of 4.x are gone, so every extractor is built through a builder and enforces the limits.

Errors

4.x 5.0

4.x	5.0
`ErrorHandlerChoice.RETRY`	Removed. Retry outside the extractor: the entry stream is already consumed.
`ErrorHandlerChoice.BAIL_OUT`	`ABORT`, which now has `BAIL_OUT’s old meaning: stop and rethrow.
`ABORT` returning normally after a failure	`ABORT` rethrows the failure. The default handler aborts.
`errorHandler(null)`	Throws `NullPointerException`, as do `filter(null)` and `postProcessor(null)`.
Parser `RuntimeException` on corrupt input	`IOException` with the original as cause.
`InvalidPathException` for an entry name or link target with a NUL character	`IOException`, which the error handler can skip.

ErrorHandlerChoice.RETRY

Removed. Retry outside the extractor: the entry stream is already consumed.

ErrorHandlerChoice.BAIL_OUT

ABORT, which now has `BAIL_OUT’s old meaning: stop and rethrow.

ABORT returning normally after a failure

ABORT rethrows the failure. The default handler aborts.

errorHandler(null)

Throws NullPointerException, as do filter(null) and postProcessor(null).

Parser RuntimeException on corrupt input

IOException with the original as cause.

InvalidPathException for an entry name or link target with a NUL character

IOException, which the error handler can skip.

Corrupt archives and corrupt compressed data throw IOException. Callbacks (filter, error handler, post-processor, unsupported-entry handler) and MissingArchiveDependencyException still propagate unchanged.
A RuntimeException from a caller-supplied source stream is reported as corrupt input.
The error handler is consulted for parser failures while reading entry content. In 4.x extraction aborted.
stream() surfaces parser failures as UncheckedIOException wrapping the IOException.
A .Z header demanding a code table above 64 MiB fails with IOException.

Reading archives

4.x 5.0

4.x	5.0
`ArchiveExtractor.Entry`, `ArchiveExtractor.ErrorHandlerChoice`, `ArchiveExtractor.EscapingSymlinkPolicy`	Top-level `Entry`, `ErrorHandlerChoice` and `EscapingSymlinkPolicy` in `com.hominux.compress4j.archivers`.
`for (Entry e : extractor)` / `extractor.iterator()`	`extractor.stream()`; use `item.entry()`
`extractor.stream()` returning `Stream<Entry>`	`extractor.stream().map(ArchiveItem::entry)`
`extractor.nextEntry()` + `openEntryStream(entry)`	`extractor.stream()`; read `item.content()` while the item is current
`extractor.setStripComponents(1)` and other setters	The matching builder option, e.g. `.stripComponents(1)`
`setPostProcessor(Consumer<Path>)`	`postProcessor(BiConsumer<Entry, Path>)` builder option
`entry.linkTarget()` returning a nullable `String`	`entry.linkTarget()` returning `Optional<String>`
Record patterns over `Entry`, which was `(String, Type, int, String)`	`Entry` is `(String, Type, int, Optional<String>, Optional<FileTime>, OptionalLong)`. Build one with `new Entry(name, type, mode).withLinkTarget(target)`.
`filter` seeing the archive name	`filter` sees the name after `stripComponents`
Devices, FIFOs, sockets and hard links extracted as empty or regular files, or dropped silently (tar)	Skipped and reported to `unsupportedEntryHandler` as `UnsupportedEntry(name, kind)`. arj symbolic links go there too; dump no longer sends unsupported entries to the error handler.

ArchiveExtractor.Entry, ArchiveExtractor.ErrorHandlerChoice, ArchiveExtractor.EscapingSymlinkPolicy

Top-level Entry, ErrorHandlerChoice and EscapingSymlinkPolicy in com.hominux.compress4j.archivers.

for (Entry e : extractor) / extractor.iterator()

extractor.stream(); use item.entry()

extractor.stream() returning Stream<Entry>

extractor.stream().map(ArchiveItem::entry)

extractor.nextEntry() + openEntryStream(entry)

extractor.stream(); read item.content() while the item is current

extractor.setStripComponents(1) and other setters

The matching builder option, e.g. .stripComponents(1)

setPostProcessor(Consumer<Path>)

postProcessor(BiConsumer<Entry, Path>) builder option

entry.linkTarget() returning a nullable String

entry.linkTarget() returning Optional<String>

Record patterns over Entry, which was (String, Type, int, String)

Entry is (String, Type, int, Optional<String>, Optional<FileTime>, OptionalLong). Build one with new Entry(name, type, mode).withLinkTarget(target).

filter seeing the archive name

filter sees the name after stripComponents

Devices, FIFOs, sockets and hard links extracted as empty or regular files, or dropped silently (tar)

Skipped and reported to unsupportedEntryHandler as UnsupportedEntry(name, kind). arj symbolic links go there too; dump no longer sends unsupported entries to the error handler.

An extractor reads its archive once: call either stream() or extract(), once.

Zips written by 4.x still extract their symlinks as regular files. Only archives re-created with 5.0 contain real symlinks.

Writing archives

4.x 5.0

4.x	5.0
`addFile(name, Path)`	`add(EntrySource.file(name, path))`; `EntrySource.file` accepts regular files only.
`addFile(name, Path, modTime)`	`add(EntrySource.file(name, path).withLastModified(modTime))`
`addFile(name, byte[])`, `addFile(name, byte[], modTime)`	`add(EntrySource.file(name, bytes))`, with `.withLastModified(modTime)`
`addFile(name, InputStream)`	`add(EntrySource.file(name, stream, size))`, or `add(EntrySource.buffered(source, tempDir))` when the size is unknown
`addFile(name, InputStream, modTime)`	`add(EntrySource.file(name, stream, size))`, with `.withLastModified(modTime)`. 4.x took no size; 5.0 needs one, or `EntrySource.buffered(source, tempDir)` when it is unknown.
`addFile(name, path, attrs, modTime)`	`add(EntrySource.file(name, path).withLastModified(modTime))`
`addDirectory(name)`, `addDirectory(name, modTime)`	`add(EntrySource.directory(name))`, with `.withLastModified(modTime)`
`addDirectoryRecursively(Path, FileTime)`	`addDirectoryRecursively("", dir, time)`
A symlink or directory `Path` added under a chosen name	`EntrySource.of(base, path)` names it relative to `base`. To choose the name, construct `new EntrySource.Symlink(…)` or `new EntrySource.Directory(…)`.
`builder.filter((name, path) → …)`	`builder.filter(source → …)` using `source.name()`; the filter takes a `Predicate<EntrySource>`.
`creator.withFilter(…)`	`builder.filter(…)`
`protected accept(String, Path)` hook	`builder.filter(…)`

addFile(name, Path)

add(EntrySource.file(name, path)); EntrySource.file accepts regular files only.

addFile(name, Path, modTime)

add(EntrySource.file(name, path).withLastModified(modTime))

addFile(name, byte[]), addFile(name, byte[], modTime)

add(EntrySource.file(name, bytes)), with .withLastModified(modTime)

addFile(name, InputStream)

add(EntrySource.file(name, stream, size)), or add(EntrySource.buffered(source, tempDir)) when the size is unknown

addFile(name, InputStream, modTime)

add(EntrySource.file(name, stream, size)), with .withLastModified(modTime). 4.x took no size; 5.0 needs one, or EntrySource.buffered(source, tempDir) when it is unknown.

addFile(name, path, attrs, modTime)

add(EntrySource.file(name, path).withLastModified(modTime))

addDirectory(name), addDirectory(name, modTime)

add(EntrySource.directory(name)), with .withLastModified(modTime)

addDirectoryRecursively(Path, FileTime)

addDirectoryRecursively("", dir, time)

A symlink or directory Path added under a chosen name

EntrySource.of(base, path) names it relative to base. To choose the name, construct new EntrySource.Symlink(…) or new EntrySource.Directory(…).

builder.filter((name, path) → …)

builder.filter(source → …) using source.name(); the filter takes a Predicate<EntrySource>.

creator.withFilter(…)

builder.filter(…)

protected accept(String, Path) hook

builder.filter(…)

builder.filter(null) throws NullPointerException; it no longer resets the filter. The filter may run more than once per entry, so keep it free of side effects.
After a write fails, the creator is failed: further adds throw IllegalStateException. close() still releases the sink.
Every format enforces a file’s declared size. Too many or too few bytes throw an IOException naming the entry.
The creator strips leading slashes from entry names and converts \ to /. Names with a NUL, a drive prefix such as C: or a .. segment throw UnsafeEntryException. A name made only of slashes throws IllegalArgumentException. A walked file literally named ..\x aborts addDirectoryRecursively.
Sockets, FIFOs and devices are rejected as entry sources with IllegalArgumentException, including during addDirectoryRecursively.
addDirectoryRecursively stores directory permissions. On a file system without a POSIX view (Windows, zip FS), path modes stay unknown and the format default applies, instead of DOS flags.
ar and cpio symlink targets over 4096 bytes fail with IOException; zip caps symlink targets too.

Sources and sinks

4.x 5.0

4.x	5.0
`builder(Path)` opened the file and threw `IOException`	`builder(Path)` throws nothing and opens the file in `build()`, for every format, creator and extractor.
`builder(File)`	Removed; use `builder(Path)`.
`ZipArchiveExtractorBuilder.setSeekableByteChannel(ch)`	`ZipArchiveExtractor.builder(ch)`
`SevenZArchiveExtractorBuilder.setSeekableByteChannel(ch)`	`SevenZArchiveExtractor.builder(ch)`
Public constructors of the archive classes, including the stream-taking ones	Removed; every format class is `final`. Use the builders: `builder(Path)`, `builder(SeekableByteChannel)`, and `builder(InputStream)` or `builder(OutputStream)` where the format streams. Zip reads a stream with `ZipArchiveExtractor.streaming(InputStream)`.
Nested `XxxArchiveExtractorBuilder` and `XxxArchiveCreatorBuilder`	Nested `Builder` classes (`StreamingBuilder` for zip streaming) without public constructors.
`buildArchiveInputStream()`, `buildArchiveOutputStream()`	Removed. Commons Compress streams no longer appear in the API.
`ZipFileArchiveInputStream`, `SevenZFileArchiveInputStream`, `SevenZFileArchiveOutputStream`, and tar’s `TarArchiveInputStream` and `TarArchiveOutputStream` accessors	Removed.

builder(Path) opened the file and threw IOException

builder(Path) throws nothing and opens the file in build(), for every format, creator and extractor.

builder(File)

Removed; use builder(Path).

ZipArchiveExtractorBuilder.setSeekableByteChannel(ch)

ZipArchiveExtractor.builder(ch)

SevenZArchiveExtractorBuilder.setSeekableByteChannel(ch)

SevenZArchiveExtractor.builder(ch)

Public constructors of the archive classes, including the stream-taking ones

Removed; every format class is final. Use the builders: builder(Path), builder(SeekableByteChannel), and builder(InputStream) or builder(OutputStream) where the format streams. Zip reads a stream with ZipArchiveExtractor.streaming(InputStream).

Nested XxxArchiveExtractorBuilder and XxxArchiveCreatorBuilder

Nested Builder classes (StreamingBuilder for zip streaming) without public constructors.

buildArchiveInputStream(), buildArchiveOutputStream()

Removed. Commons Compress streams no longer appear in the API.

ZipFileArchiveInputStream, SevenZFileArchiveInputStream, SevenZFileArchiveOutputStream, and tar’s TarArchiveInputStream and TarArchiveOutputStream accessors

Removed.

New: builder(SeekableByteChannel) on every format, and ZipArchiveExtractor.streaming(InputStream). See format capabilities.
Zip streaming mode reports mode 0, extracts symlinks as regular files holding the target, and may leave an entry’s size() empty.
7z needs random access. To read a stream, copy it into a SeekableInMemoryByteChannel or a temp file; see 7z.
SevenZArchiveCreator.builder(Path) creates its file at build(), not when the builder is made. The channel creator writes from offset 0.
ArchiveExtractor and ArchiveCreator lose their type parameter. Their builders become ArchiveExtractor.Builder<B, E> and ArchiveCreator.Builder<B, C>.

Compressors

4.x 5.0

4.x	5.0
`GzipCompressor.builder(p)` and the compressor of every other codec	`Compressor.builder(p, Compression.gzip())`; `Compression.bzip2()`, `xz()`, `lzma()`, `lz4Block()`, `lz4Framed()`, `zstd()`, `deflate()`, `snappyRaw()`, `snappyFramed()` and `pack200()` cover the rest.
`GzipDecompressor.builder(p)` and the decompressor of every other codec	`Decompressor.builder(p)` detects the codec, or `Decompressor.builder(p, Compression.gzip())`. Input with no detectable signature passes through unchanged, so name deflate, LZMA, LZ4 block and raw Snappy explicitly. Deflate64, Brotli and `.Z` read through `Compression.deflate64()`, `brotli()` and `unixZ()`.
Codec options on the codec builders	Options on the `Compression` value, e.g. `Compression.gzip().level(9)`; `DeflateCompressionLevel` is removed.
`compressor.write(File)`, `decompressor.write(File)`	`Path` only. `Decompressor.write` refuses an existing target unless `overwrite(true)`.
Packages `compressors.{snappy,lz4,xz,deflate,gzip,bzip2,pack200,zstd,lzma,z,deflate64,brotli}`	Removed, with the abstract bases that became `CodecCompressor` and `CodecDecompressor`.
No limits on decompression	`Decompressor` enforces `ExtractionLimits.defaults()` (ratio 100 after 1 MiB); set `limits(…)`, `maxTotalSize(…)` or `maxRatio(…)`. `maxEntries` and `maxEntrySize` do not apply.

GzipCompressor.builder(p) and the compressor of every other codec

Compressor.builder(p, Compression.gzip()); Compression.bzip2(), xz(), lzma(), lz4Block(), lz4Framed(), zstd(), deflate(), snappyRaw(), snappyFramed() and pack200() cover the rest.

GzipDecompressor.builder(p) and the decompressor of every other codec

Decompressor.builder(p) detects the codec, or Decompressor.builder(p, Compression.gzip()). Input with no detectable signature passes through unchanged, so name deflate, LZMA, LZ4 block and raw Snappy explicitly. Deflate64, Brotli and .Z read through Compression.deflate64(), brotli() and unixZ().

Codec options on the codec builders

Options on the Compression value, e.g. Compression.gzip().level(9); DeflateCompressionLevel is removed.

compressor.write(File), decompressor.write(File)

Path only. Decompressor.write refuses an existing target unless overwrite(true).

Packages compressors.{snappy,lz4,xz,deflate,gzip,bzip2,pack200,zstd,lzma,z,deflate64,brotli}

Removed, with the abstract bases that became CodecCompressor and CodecDecompressor.

No limits on decompression

Decompressor enforces ExtractionLimits.defaults() (ratio 100 after 1 MiB); set limits(…), maxTotalSize(…) or maxRatio(…). maxEntries and maxEntrySize do not apply.

Detection reads gzip, bzip2, xz, zstd, lz4-framed, .Z and snappy-framed by magic number. Concatenated gzip, bzip2, xz and framed LZ4 members are decompressed; zstd, framed Snappy and .Z are not verified to read them. LZMA has no magic number: pass Compression.lzma().
Decompressor never detects Pack200, and tar rejects Compression.pack200(). Pass Pack200 explicitly; it decodes eagerly in build(), so the limits do not bound it; use it for trusted input only.
Compressor and Decompressor open their path, channel or stream in build().

Tar

4.x 5.0

4.x	5.0
`TarGzArchiveCreator.builder(p)` and the BZip2, Xz, Lzma, Lz4 and Zstd variants	`TarArchiveCreator.builder(p).compression(Compression.gzip())`
`TarGzArchiveExtractor.builder(p)` and the BZip2, Xz, Lzma, Lz4, Zstd and `TarZ` variants	`TarArchiveExtractor.builder(p)` detects the compression. `.tar.lzma` needs `.compression(Compression.lzma())`.
`BaseTarArchiveExtractor`, `BaseTarArchiveCreator`, `compressorOutputStreamBuilder`, `xzInputStream`, `lz4InputStream`, `lzmaInputStream`, `zstdInputStream`, `zInputStream`	Removed.
`encoding(String)`	`encoding(Charset)`
`longFileMode(int)`, `bigNumberMode(int)`	`longFileMode(TarLongFileMode)` and `bigNumberMode(TarBigNumberMode)`. Defaults are `POSIX` and `STAR` instead of the commons `ERROR` defaults, so long names and big numbers archive without configuration.
`blockSize(int)` accepting any value	`IllegalArgumentException` unless the value is -511 or a positive multiple of 512.
Bad header checksum, or 1 to 511 bytes of input, read as an empty archive	`IOException`
Tar entries on Windows reporting mode `0`	Archived modes on every host.

TarGzArchiveCreator.builder(p) and the BZip2, Xz, Lzma, Lz4 and Zstd variants

TarArchiveCreator.builder(p).compression(Compression.gzip())

TarGzArchiveExtractor.builder(p) and the BZip2, Xz, Lzma, Lz4, Zstd and TarZ variants

TarArchiveExtractor.builder(p) detects the compression. .tar.lzma needs .compression(Compression.lzma()).

BaseTarArchiveExtractor, BaseTarArchiveCreator, compressorOutputStreamBuilder, xzInputStream, lz4InputStream, lzmaInputStream, zstdInputStream, zInputStream

Removed.

encoding(String)

encoding(Charset)

longFileMode(int), bigNumberMode(int)

longFileMode(TarLongFileMode) and bigNumberMode(TarBigNumberMode). Defaults are POSIX and STAR instead of the commons ERROR defaults, so long names and big numbers archive without configuration.

blockSize(int) accepting any value

IllegalArgumentException unless the value is -511 or a positive multiple of 512.

Bad header checksum, or 1 to 511 bytes of input, read as an empty archive

IOException

Tar entries on Windows reporting mode 0

Archived modes on every host.

Other formats

4.x 5.0

4.x	5.0
`cpioInputStream()`, `cpioOutputStream()`, `CpioArchiveInputStreamBuilder`, `CpioArchiveOutputStreamBuilder`, `.and()`	`format(CpioFormat)` (`NEW`, `OLD_ASCII`, `OLD_BINARY`), `blockSize(int)` and `encoding(Charset)` on the builder. The `NEW_CRC` format is not offered, and `blockSize` below 1 throws `IllegalArgumentException`.
arj and dump `encoding(String)`	`encoding(Charset)`
Zip `compressionMethod(int)`	`compressionMethod(ZipCompressionMethod)`: `STORED` or `DEFLATED`
Zip `setUseZip64(Zip64Mode)`	`zip64(ZipZip64Mode)`
Zip `setCreateUnicodeExtraFields(UnicodeExtraFieldPolicy)`	`createUnicodeExtraFields(ZipUnicodeExtraFields)`
Zip `setComment`, `setFallbackToUTF8`, `setUseLanguageEncodingFlag`, `setEncoding(String)`	`comment`, `fallbackToUtf8`, `useLanguageEncodingFlag`, `encoding(Charset)`
Zip `setIgnoreLocalFileHeader`, `setMaxNumberOfDisks`, `setUseUnicodeExtraFields`, `setZstdInputStreamFactory`	`ignoreLocalFileHeader`, `maxNumberOfDisks`, `useUnicodeExtraFields`, `zstdInputStreamFactory`, now taking a `ZipArchiveExtractor.ZstdStreamFactory`

cpioInputStream(), cpioOutputStream(), CpioArchiveInputStreamBuilder, CpioArchiveOutputStreamBuilder, .and()

format(CpioFormat) (NEW, OLD_ASCII, OLD_BINARY), blockSize(int) and encoding(Charset) on the builder. The NEW_CRC format is not offered, and blockSize below 1 throws IllegalArgumentException.

arj and dump encoding(String)

encoding(Charset)

Zip compressionMethod(int)

compressionMethod(ZipCompressionMethod): STORED or DEFLATED

Zip setUseZip64(Zip64Mode)

zip64(ZipZip64Mode)

Zip setCreateUnicodeExtraFields(UnicodeExtraFieldPolicy)

createUnicodeExtraFields(ZipUnicodeExtraFields)

Zip setComment, setFallbackToUTF8, setUseLanguageEncodingFlag, setEncoding(String)

comment, fallbackToUtf8, useLanguageEncodingFlag, encoding(Charset)

Zip setIgnoreLocalFileHeader, setMaxNumberOfDisks, setUseUnicodeExtraFields, setZstdInputStreamFactory

ignoreLocalFileHeader, maxNumberOfDisks, useUnicodeExtraFields, zstdInputStreamFactory, now taking a ZipArchiveExtractor.ZstdStreamFactory

New in 5.0: SevenZArchiveCreator.builder(…).contentCompression(SevenZMethod) with the SevenZMethod enum COPY, LZMA2, LZMA, BZIP2 and DEFLATE.

The zip creator’s compressionMethod, comment, zip64 and encoding reject null with NullPointerException; so do the 7z creator’s contentCompression and the 7z extractor’s password.

File modes

4.x 5.0

4.x	5.0
Windows extraction marked executable files read-only	A file is read-only only when its owner-write bit is missing. Hidden is never set and directories are never marked read-only.
Directory modes applied as each directory appeared	Applied after all entries, deepest first. After creation a directory keeps an interim mode of the archive’s permission bits plus owner-rwx, so a read-only directory cannot block its files. An aborted extraction leaves directories owner-writable; the group and other bits stay within the archive’s grant. A `postProcessor` that changes a directory’s mode may be overwritten by the archive mode.
Mode `0`	Files and directories keep the host defaults.
An entry path holding a symlink was overwritten through the link	The write fails and goes to the error handler.

Windows extraction marked executable files read-only

A file is read-only only when its owner-write bit is missing. Hidden is never set and directories are never marked read-only.

Directory modes applied as each directory appeared

Applied after all entries, deepest first. After creation a directory keeps an interim mode of the archive’s permission bits plus owner-rwx, so a read-only directory cannot block its files. An aborted extraction leaves directories owner-writable; the group and other bits stay within the archive’s grant. A postProcessor that changes a directory’s mode may be overwritten by the archive mode.

Mode 0

Files and directories keep the host defaults.

An entry path holding a symlink was overwritten through the link

The write fails and goes to the error handler.

Removed APIs and dependencies

ArchiveExtractor.setAttributes, ArchiveExtractor.isIsOsWindows, ArchiveCreator.mode, ArchiveCreator.isIsOsWindows and ArchiveCreator.sanitiseName.
FileUtils.DOS_READ_ONLY, DOS_HIDDEN and NO_MODE; archivers.Compressor, utils.StringUtil and utils.ExcludeFromJacocoGeneratedReport. The utils package is now the unexported internal.util.
Custom formats: the format classes are final, and ArchiveExtractor and ArchiveCreator cannot be subclassed outside the library. The reader and writer SPI (internal.archive) is not exported from the module. Their constructors take internal types, and nextEntry, openEntryStream, closeEntryStream, readEntryContent, writeDirectoryEntry and writeFileEntry are gone.
MissingArchiveDependencyException is final.
commons-compress and commons-io are not api dependencies and no module requires them transitively. They move to runtime scope in the published POM. Code that uses their types declares the dependency and requires org.apache.commons.compress or org.apache.commons.io itself. EntrySource.File content is an EntrySource.ContentSupplier and zstdInputStreamFactory takes a ZipArchiveExtractor.ZstdStreamFactory, where both took commons-io function types (IOSupplier<InputStream> and IOFunction<InputStream, InputStream>). Lambdas and method references compile unchanged; a variable typed with the commons-io type, or an IOSupplier passed from content(), must change to the new type.
jakarta.annotation-api is no longer a dependency. Compress4J requires the JSpecify module (org.jspecify) and every exported package is @NullMarked, so Kotlin and NullAway see non-null parameters and return types.