zip-archive 0.5
* Reject absolute and drive-qualified entry paths on extraction, even
when OptDestination is given. `checkPath` now throws UnsafePath for
absolute (and, on Windows, drive-qualified) paths regardless of options.
* Validate symbolic link entry paths on extraction.
* Decode file names per the UTF-8 flag, with CP437 fallback.
Previously file names were unconditionally decoded as UTF-8, so archives
with non-UTF-8 names raised an impure UnicodeException.
Now the general purpose bit flag is consulted: when bit 11 (language
encoding flag) is set, names are decoded as UTF-8 leniently (invalid
bytes become U+FFFD); otherwise they are decoded as code page 437 as
specified in APPNOTE.TXT appendix D.
* Return Nothing from `decryptData` on truncated encrypted data, instead
of raising an exception.
* Raise Zip64NotSupported instead of silently truncating on overflow.
The library does not support ZIP64, but nothing prevented writing
archives that would need it: toEntry truncated sizes of entries of 4GB
or more to Word32, and putArchive truncated entry counts (Word16) and
central directory offsets (Word32), silently producing corrupt output.
Add a Zip64NotSupported constructor to ZipException [API change]
and throw it (as a pure exception) from toEntry for oversized
entries and from putArchive for too many entries or too-large
archives. Local file offsets are now computed in Int64 so overflow
can actually be detected.
* Undo the local time zone shift when setting extracted file times.
`readEntry` stores `eLastModified` shifted by the local time
zone offset; ensure that this shift is reversed on extraction.
* Clamp DOS datetimes at the upper end of the representable range
(year 2108).
* Stop claiming maximum compression in the general purpose bit flag.
`compressData` uses zlib's default compression level, but the written
flag (0x802) had bit 1 set, which means the entry was deflated with
maximum compression. Write 0x800 (UTF-8 file names only) instead.
* Make `addFilesToArchive` near-linear in the number of files.
Large directory trees now dedupe via Data.Set on normalized paths,
preserving the previous semantics (first entry for a path wins,
new entries precede old ones).
* Avoid retaining the whole remaining archive in `getCompressedData`.
The raw deflate stream is fed to zlib's incremental `decompressST`
chunk by chunk, counting only the bytes actually consumed, so
memory use is proportional to the entry rather than to everything
after it.
* Use a CRC32 lookup table in the PKWARE key schedule.
* Stream extraction in `writeEntry` with an incremental CRC check.
writeEntry previously computed the CRC32 of the whole uncompressed
entry and then wrote it with B.writeFile; the reference to the
data across the CRC pass forced the entire entry to be retained in
memory. Now the entry is written chunk by chunk to a temporary
file in the target directory while the CRC is updated
incrementally, and the file is renamed into place only if the CRC
matches. As before, a pre-existing file at the target path is
left intact when the CRC check fails; the temporary file is
removed on mismatch or on any exception during writing.
* Encode each entry path once in `putLocalFile` and `putFileHeader`.
Previously both serializers normalized and UTF-8-encoded the entry
path twice: once for its length field and once for the path bytes.
* cabal: use extra-doc-files stanza.
* Add regression test for issue #55 (dotfile paths).
* Fix spelling errors (@kianmeng, #69).
* Remove stack.yaml.
* Change default-language to Haskell2010.
* Remove spurious dependencies (pretty, mtl).
* Depend on base >= 4.11.
zip-archive 0.4.3.2
* readEntry: Fix computation of modification time (#67).
It should be a UNIX time (seconds since UNIX epoch), but
computed relative to the local time zone, not UTC.
zip-archive 0.4.3.1
* Use streaming decompress to identify extent of compressed data (#66).
This fixes a problem that arises for local files with bit 3
of the general purpose bit flag set. In this case, we don't
get information up front about the size of the compressed
data. So how do we know where the compressed data ends?
Previously, we tried to determine this by looking for the
signature of the data descriptor. But the data descriptor doesn't
always HAVE a signature, and it is also possible for signatures to
occur accidentally in the compressed data itself (#65).
Instead, we now use the streaming decompression interface from
zlib's Internal module to identify where the compressed data
ends. Fixes both #65 and #25.
zip-archive 0.4.3
* Improve code for retrieving compressed data of unknown length (#63).
Do not assume we'll have the signature 0x08074b50 that is
sometimes used for the data description, because it is not
in the spec and is not always used.
* Make some record fields strict.
* Require binary >= 0.7.2, remove some CPP
zip-archive 0.4.2.2
* Use `command -v` before trying `which` in the test suite (#62).
`command` is a bash builtin, but for busybox we'll need `which`.
zip-archive 0.4.2.1
* Fix Windows build regression (#61).
zip-archive 0.4.2
* Fix problem with files with colon (#89).
* Remove build-tools. This was used to indicate that the 'unzip'
executable was needed for testing, but it was never intended to be used
this way and now the field is deprecated. The current test suite
simply skips the test using the unzip executable (with a warning) if
'unzip' is not in the path.
* Remove existing symlinks when extracting zip files with symlinks (#60,
Vikrem). Previously, writeEntry would raise an error if it tried to
create a symlink and a symlink already existed at that path. This
behavior was inconsistent with its behavior for regular files, which
it overwrote without comment. This commit causes symlinks to be replaced
by writeEntry instead of an error being raised.
* Remove binary < 0.6 CPP. It's no longer needed because we don't support
binary < 0.6. Also use manySig instead of many, to get better error
messages.
* Add type annotation for printf.
* Better checking for unsafe paths (#55). This method allows things like
`foo/bar/../../baz`.
* Require base >= 4.5 (#56)
* Add GitHub CI.
zip-archive 0.4.1
* writEntry behavior change: Improve raising of UnsafePath error (#55).
Previously we raised this error spuriously when archives were unpacked
outside the working directory. Now we raise it if eRelativePath contains
".." as a path component, or eRelativePath path is an absolute path and
there is no separate destination directory. (Note that `/foo/bar` is fine
as a path as long as a destination directory, e.g. `/usr/local`, is
specified.)
zip-archive 0.4
* Implement read-only support for PKWARE encryption (Sergii Rudchenko).
The "traditional" PKWARE encryption is a symmetric encryption
algorithm described in zip format specification in section 6.1.
This change allows to extract basic "password-protected" entries from
ZIP files. Note that the standard file extraction function
extractFilesFromArchive does not decrypt entries (it will raise
an exception if it encounters an encrypted entry). To handle
archives with encrypted entries, use the new function
fromEncryptedEntry.
API changes:
+ Add eEncryptionMethod field to Entry.
+ Add EncryptionMethod type.
+ Add function isEncryptedEntry.
+ Add function fromEncryptedEntry.
* Add CannotWriteEncryptedEntry constructor to ZipException.
* Add UnsafePath to ZipException (#50).
* writeEntry: raise UnsafePath exception for unsafe paths (#50).
This prevents malicious zip files from overwriting paths
above the working directory.
* Add Paths_zip_archive to autogen-modules.
* Clarify README and cabal description.
* Specify cabal-version: 2.0. Otherwise we get an unknown build
tool error using `build-depends` without a custom Setup.hs.
* Change build-type to simple. Retain 'build-tools: unzip' in
test stanza, though now it doesn't do anything except give a
hint to external tools. If unzip is not found in the path,
the test suite prints a message and counts the test that
requires unzip as succeeding (see #51).
zip-archive 0.3.3
* Remove dependency on old-time (typedrat).
* Drop splitBase flag and support for base versions < 3.
zip-archive 0.3.2.5
* Move 'build-tools: unzip' from library stanza to test stanza.
unzip should only be required for testing, not for regular
builds of the library.
zip-archive 0.3.2.4
* Make build-tools stanza conditional on non-windows. Closes #44.
zip-archive 0.3.2.3
* Use custom-setup stanza and specify build-tools. Closes #41.
zip-archive 0.3.2.2
* Use createSymbolicLink instead of createFileLink in tests. This allows
us to lower the directory lower bound (#40).
zip-archive 0.3.2.1
* Fixes for handling of symbolic links (#39, Tommaso Piazza).
* Fixes for symbolic link tests, and additional tests.
zip-archive 0.3.2
* Add ZipOption to preserve symbolic links (#37, Tommaso Piazza).
Add OptPreserveSymbolicLinks constructor to ZipOption. If this option
is set, symbolic links will be preserved. Symbolic links are not
supported on Windows.
* Require binary >= 0.6 (#36).
* Improve exit handling in zip-archive program.
zip-archive 0.3.1.1
* readEntry: Read file as a strict ByteString. This avoids
problems on Windows, where the file handle wasn't being closed.
* Added appveyor.yml to do continuous testing on Windows.
* Test suite: remove need for external zip program (#35).
Instead of creating an archive with zip, we now store
a small externally created zip archive to use for testing.
zip-archive 0.3.1
* Don't use a custom build (#28).
* Renamed executable Zip -> zip-archive, added --debug option.
The --debug option prints the intermediate Haskell data structure.
zip-archive 0.3.0.7
* Fix check for unix file attributes (#34).
Previously attributes would not always be preserved
for files in zip archives.
zip-archive 0.3.0.6
* Bump bytestring lower bound so toStrict is guaranteed (Benjamin Landers).
zip-archive 0.3.0.5
* Fix bug in `OptLocation` handling (EugeneN). When using
`OptLocation folder False` (for adding files to an archive into a
folder without preserving full path hierarchy), original files'
names were ignored, resulting in all the files getting the same name.
zip-archive 0.3.0.4
* Fix `toArchive` so it doesn't use too much memory when a data
data descriptor holds the size (Michael Stahl, #29).
The size fields in the local file headers may not contain valid values,
in which case the sizes are stored in a "data descriptor" that follows
the file data. Previously handling this case required reading the
entire archive is a `[Word8]` list. With this change, `getWordsTilSig`
iteratively reads chunks as strict ByteStrings and converts them to
a lazy ByteString at the end.
zip-archive 0.3.0.3
* Test suite: use withTempDir to create temporary directory.
This should help fix problems some have encountered with the
test suite leaving a temporary directory behind.
zip-archive 0.3.0.2
* Fix test suite so it runs on Windows.
* Zip executable: get version from cabal `Paths_zip_archive` (#27).
zip-archive 0.3.0.1
* Set `eVersionMadeBy` to 0 (default) in `toEntry`, since we are
setting external attributes to 0. See jgm/pandoc#2822.
Only to `eVersionMadeBy` to UNIX if we actually read file
attributes on a UNIX system.
zip-archive 0.3
* Support preservation of file modes on Posix (Dan Aloni, #26).
* Add `eVersionMadeBy` field to `Entry` (API change).
* Export `ZipException` (API change).
* `fromEntry` no longer checks for CRC32 match. Previously, it issued
`error` if the match failed. CRC32 match is now checked in `writeEntry`
instead, and a `CRC32Exception` is raised if the checksum doesn't match.
* Test suite: return nonzero status if there are test failures.
Previously we mistakenly did this only on 'errors', not failures.
* Test suite: don't use -9 with zip as it isn't always available.
* Use .travis.yml that builds on both stack and cabal.
zip-archive 0.2.3.7
* Declared test suite's dependency on 'zip' using custom Setup.lhs (#21,#22).
zip-archive 0.2.3.6
* Removed hard-coded path to zip in test suite (#21).
* Removed misplaced build-depends in cabal file.
zip-archive 0.2.3.5
* Allow compilation with binary >= 0.5. Note that toArchiveOrFail
is not safe when compiled against binary < 0.7; it will never
return a Left value, and will raise an error if parsing fails,
just like toArchive. This is documented in the haddocks.
This is ugly, but justified by the need to have a version
of zip-archive that compiles against older versions of binary.
zip-archive 0.2.3.4
* Make sure all path comparisons compare normalized paths.
So, findEntryByPath "foo" will find something stored as "./foo"
in the zip container.
zip-archive 0.2.3.3
* Better normalization of file paths: "./foo/bar" and "foo/./bar"
are now treated the same, for example. Note that we do not
yet treat "foo/../bar" and "bar" as the same.
zip-archive 0.2.3.2
* Removed lower bound on directory (>= 1.2), which caused build
failures with GHC 7.4 and 7.6.
* Added travis script for automatic testing on 3 GHC versions.
zip-archive 0.2.3.1
* Require binary >= 0.7 and directory >= 1.2. The newer binary
is needed to provide toArchiveOrFail. The other change is
mainly for convenience, to avoid lots of ugly conditional
compilation.
zip-archive 0.2.3
* Export new function `toArchiveOrFail`. Closes #17.
* Set general purpose bit flag to use UTF8 in local file header.
Otherwise we get a mismatch between the flag in the central
directory and the flag in the local file header, which causes some
programs not to be able to extract the files. Closes #19.
zip-archive 0.2.2.1
* Fix a stack overflow in getWordsTillSig (Tristan Ravitch).
zip-archive 0.2.2
* Set bit 11 in the file header to ensure other programs
recognize UTF-8 encoded file names (Tobias Brandt).
zip-archive 0.2.1
* Added OptLocation, to specify the path to which a file
is to be added when readEntry is used (Stephen McIntosh).