blob: bb0b8da6a1cfc2fb92ef494791b3083609e76690 [file] [log] [blame]
/*
* Copyright (c) 1995, 2017, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
package java.util.zip;
import java.io.Closeable;
import java.io.InputStream;
import java.io.IOException;
import java.io.EOFException;
import java.io.File;
import java.io.RandomAccessFile;
import java.io.UncheckedIOException;
import java.lang.ref.Cleaner.Cleanable;
import java.nio.charset.Charset;
import java.nio.charset.StandardCharsets;
import java.nio.file.attribute.BasicFileAttributes;
import java.nio.file.Files;
import java.util.ArrayDeque;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collections;
import java.util.Deque;
import java.util.Enumeration;
import java.util.HashMap;
import java.util.Iterator;
import java.util.Objects;
import java.util.NoSuchElementException;
import java.util.Set;
import java.util.Spliterator;
import java.util.Spliterators;
import java.util.WeakHashMap;
import java.util.function.Consumer;
import java.util.function.Function;
import java.util.function.IntFunction;
import java.util.jar.JarEntry;
import java.util.stream.Stream;
import java.util.stream.StreamSupport;
import jdk.internal.misc.JavaUtilZipFileAccess;
import jdk.internal.misc.SharedSecrets;
import jdk.internal.misc.VM;
import jdk.internal.perf.PerfCounter;
import jdk.internal.ref.CleanerFactory;
import static java.util.zip.ZipConstants64.*;
import static java.util.zip.ZipUtils.*;
/**
* This class is used to read entries from a zip file.
*
* <p> Unless otherwise noted, passing a {@code null} argument to a constructor
* or method in this class will cause a {@link NullPointerException} to be
* thrown.
*
* @apiNote
* To release resources used by this {@code ZipFile}, the {@link #close()} method
* should be called explicitly or by try-with-resources. Subclasses are responsible
* for the cleanup of resources acquired by the subclass. Subclasses that override
* {@link #finalize()} in order to perform cleanup should be modified to use alternative
* cleanup mechanisms such as {@link java.lang.ref.Cleaner} and remove the overriding
* {@code finalize} method.
*
* @implSpec
* If this {@code ZipFile} has been subclassed and the {@code close} method has
* been overridden, the {@code close} method will be called by the finalization
* when {@code ZipFile} is unreachable. But the subclasses should not depend on
* this specific implementation; the finalization is not reliable and the
* {@code finalize} method is deprecated to be removed.
*
* @author David Connelly
* @since 1.1
*/
public
class ZipFile implements ZipConstants, Closeable {
private final String name; // zip file name
private volatile boolean closeRequested;
private ZipCoder zc;
// The "resource" used by this zip file that needs to be
// cleaned after use.
// a) the input streams that need to be closed
// b) the list of cached Inflater objects
// c) the "native" source of this zip file.
private final CleanableResource res;
private static final int STORED = ZipEntry.STORED;
private static final int DEFLATED = ZipEntry.DEFLATED;
/**
* Mode flag to open a zip file for reading.
*/
public static final int OPEN_READ = 0x1;
/**
* Mode flag to open a zip file and mark it for deletion. The file will be
* deleted some time between the moment that it is opened and the moment
* that it is closed, but its contents will remain accessible via the
* {@code ZipFile} object until either the close method is invoked or the
* virtual machine exits.
*/
public static final int OPEN_DELETE = 0x4;
/**
* Opens a zip file for reading.
*
* <p>First, if there is a security manager, its {@code checkRead}
* method is called with the {@code name} argument as its argument
* to ensure the read is allowed.
*
* <p>The UTF-8 {@link java.nio.charset.Charset charset} is used to
* decode the entry names and comments.
*
* @param name the name of the zip file
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
* @throws SecurityException if a security manager exists and its
* {@code checkRead} method doesn't allow read access to the file.
*
* @see SecurityManager#checkRead(java.lang.String)
*/
public ZipFile(String name) throws IOException {
this(new File(name), OPEN_READ);
}
/**
* Opens a new {@code ZipFile} to read from the specified
* {@code File} object in the specified mode. The mode argument
* must be either {@code OPEN_READ} or {@code OPEN_READ | OPEN_DELETE}.
*
* <p>First, if there is a security manager, its {@code checkRead}
* method is called with the {@code name} argument as its argument to
* ensure the read is allowed.
*
* <p>The UTF-8 {@link java.nio.charset.Charset charset} is used to
* decode the entry names and comments
*
* @param file the ZIP file to be opened for reading
* @param mode the mode in which the file is to be opened
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
* @throws SecurityException if a security manager exists and
* its {@code checkRead} method
* doesn't allow read access to the file,
* or its {@code checkDelete} method doesn't allow deleting
* the file when the {@code OPEN_DELETE} flag is set.
* @throws IllegalArgumentException if the {@code mode} argument is invalid
* @see SecurityManager#checkRead(java.lang.String)
* @since 1.3
*/
public ZipFile(File file, int mode) throws IOException {
this(file, mode, StandardCharsets.UTF_8);
}
/**
* Opens a ZIP file for reading given the specified File object.
*
* <p>The UTF-8 {@link java.nio.charset.Charset charset} is used to
* decode the entry names and comments.
*
* @param file the ZIP file to be opened for reading
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
*/
public ZipFile(File file) throws ZipException, IOException {
this(file, OPEN_READ);
}
/**
* Opens a new {@code ZipFile} to read from the specified
* {@code File} object in the specified mode. The mode argument
* must be either {@code OPEN_READ} or {@code OPEN_READ | OPEN_DELETE}.
*
* <p>First, if there is a security manager, its {@code checkRead}
* method is called with the {@code name} argument as its argument to
* ensure the read is allowed.
*
* @param file the ZIP file to be opened for reading
* @param mode the mode in which the file is to be opened
* @param charset
* the {@linkplain java.nio.charset.Charset charset} to
* be used to decode the ZIP entry name and comment that are not
* encoded by using UTF-8 encoding (indicated by entry's general
* purpose flag).
*
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
*
* @throws SecurityException
* if a security manager exists and its {@code checkRead}
* method doesn't allow read access to the file,or its
* {@code checkDelete} method doesn't allow deleting the
* file when the {@code OPEN_DELETE} flag is set
*
* @throws IllegalArgumentException if the {@code mode} argument is invalid
*
* @see SecurityManager#checkRead(java.lang.String)
*
* @since 1.7
*/
public ZipFile(File file, int mode, Charset charset) throws IOException
{
if (((mode & OPEN_READ) == 0) ||
((mode & ~(OPEN_READ | OPEN_DELETE)) != 0)) {
throw new IllegalArgumentException("Illegal mode: 0x"+
Integer.toHexString(mode));
}
String name = file.getPath();
SecurityManager sm = System.getSecurityManager();
if (sm != null) {
sm.checkRead(name);
if ((mode & OPEN_DELETE) != 0) {
sm.checkDelete(name);
}
}
Objects.requireNonNull(charset, "charset");
this.zc = ZipCoder.get(charset);
this.name = name;
long t0 = System.nanoTime();
this.res = CleanableResource.get(this, file, mode);
PerfCounter.getZipFileOpenTime().addElapsedTimeFrom(t0);
PerfCounter.getZipFileCount().increment();
}
/**
* Opens a zip file for reading.
*
* <p>First, if there is a security manager, its {@code checkRead}
* method is called with the {@code name} argument as its argument
* to ensure the read is allowed.
*
* @param name the name of the zip file
* @param charset
* the {@linkplain java.nio.charset.Charset charset} to
* be used to decode the ZIP entry name and comment that are not
* encoded by using UTF-8 encoding (indicated by entry's general
* purpose flag).
*
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
* @throws SecurityException
* if a security manager exists and its {@code checkRead}
* method doesn't allow read access to the file
*
* @see SecurityManager#checkRead(java.lang.String)
*
* @since 1.7
*/
public ZipFile(String name, Charset charset) throws IOException
{
this(new File(name), OPEN_READ, charset);
}
/**
* Opens a ZIP file for reading given the specified File object.
*
* @param file the ZIP file to be opened for reading
* @param charset
* The {@linkplain java.nio.charset.Charset charset} to be
* used to decode the ZIP entry name and comment (ignored if
* the <a href="package-summary.html#lang_encoding"> language
* encoding bit</a> of the ZIP entry's general purpose bit
* flag is set).
*
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
*
* @since 1.7
*/
public ZipFile(File file, Charset charset) throws IOException
{
this(file, OPEN_READ, charset);
}
/**
* Returns the zip file comment, or null if none.
*
* @return the comment string for the zip file, or null if none
*
* @throws IllegalStateException if the zip file has been closed
*
* @since 1.7
*/
public String getComment() {
synchronized (this) {
ensureOpen();
if (res.zsrc.comment == null) {
return null;
}
return zc.toString(res.zsrc.comment);
}
}
/**
* Returns the zip file entry for the specified name, or null
* if not found.
*
* @param name the name of the entry
* @return the zip file entry, or null if not found
* @throws IllegalStateException if the zip file has been closed
*/
public ZipEntry getEntry(String name) {
return getEntry(name, ZipEntry::new);
}
/*
* Returns the zip file entry for the specified name, or null
* if not found.
*
* @param name the name of the entry
* @param func the function that creates the returned entry
*
* @return the zip file entry, or null if not found
* @throws IllegalStateException if the zip file has been closed
*/
private ZipEntry getEntry(String name, Function<String, ? extends ZipEntry> func) {
Objects.requireNonNull(name, "name");
synchronized (this) {
ensureOpen();
byte[] bname = zc.getBytes(name);
int pos = res.zsrc.getEntryPos(bname, true);
if (pos != -1) {
return getZipEntry(name, bname, pos, func);
}
}
return null;
}
/**
* Returns an input stream for reading the contents of the specified
* zip file entry.
* <p>
* Closing this ZIP file will, in turn, close all input streams that
* have been returned by invocations of this method.
*
* @param entry the zip file entry
* @return the input stream for reading the contents of the specified
* zip file entry.
* @throws ZipException if a ZIP format error has occurred
* @throws IOException if an I/O error has occurred
* @throws IllegalStateException if the zip file has been closed
*/
public InputStream getInputStream(ZipEntry entry) throws IOException {
Objects.requireNonNull(entry, "entry");
int pos = -1;
ZipFileInputStream in = null;
Source zsrc = res.zsrc;
Set<InputStream> istreams = res.istreams;
synchronized (this) {
ensureOpen();
if (Objects.equals(lastEntryName, entry.name)) {
pos = lastEntryPos;
} else if (!zc.isUTF8() && (entry.flag & EFS) != 0) {
pos = zsrc.getEntryPos(zc.getBytesUTF8(entry.name), false);
} else {
pos = zsrc.getEntryPos(zc.getBytes(entry.name), false);
}
if (pos == -1) {
return null;
}
in = new ZipFileInputStream(zsrc.cen, pos);
switch (CENHOW(zsrc.cen, pos)) {
case STORED:
synchronized (istreams) {
istreams.add(in);
}
return in;
case DEFLATED:
// Inflater likes a bit of slack
// MORE: Compute good size for inflater stream:
long size = CENLEN(zsrc.cen, pos) + 2;
if (size > 65536) {
size = 8192;
}
if (size <= 0) {
size = 4096;
}
InputStream is = new ZipFileInflaterInputStream(in, res, (int)size);
synchronized (istreams) {
istreams.add(is);
}
return is;
default:
throw new ZipException("invalid compression method");
}
}
}
private static class InflaterCleanupAction implements Runnable {
private final Inflater inf;
private final CleanableResource res;
InflaterCleanupAction(Inflater inf, CleanableResource res) {
this.inf = inf;
this.res = res;
}
@Override
public void run() {
res.releaseInflater(inf);
}
}
private class ZipFileInflaterInputStream extends InflaterInputStream {
private volatile boolean closeRequested;
private boolean eof = false;
private final Cleanable cleanable;
ZipFileInflaterInputStream(ZipFileInputStream zfin,
CleanableResource res, int size) {
this(zfin, res, res.getInflater(), size);
}
private ZipFileInflaterInputStream(ZipFileInputStream zfin,
CleanableResource res,
Inflater inf, int size) {
super(zfin, inf, size);
this.cleanable = CleanerFactory.cleaner().register(this,
new InflaterCleanupAction(inf, res));
}
public void close() throws IOException {
if (closeRequested)
return;
closeRequested = true;
super.close();
synchronized (res.istreams) {
res.istreams.remove(this);
}
cleanable.clean();
}
// Override fill() method to provide an extra "dummy" byte
// at the end of the input stream. This is required when
// using the "nowrap" Inflater option.
protected void fill() throws IOException {
if (eof) {
throw new EOFException("Unexpected end of ZLIB input stream");
}
len = in.read(buf, 0, buf.length);
if (len == -1) {
buf[0] = 0;
len = 1;
eof = true;
}
inf.setInput(buf, 0, len);
}
public int available() throws IOException {
if (closeRequested)
return 0;
long avail = ((ZipFileInputStream)in).size() - inf.getBytesWritten();
return (avail > (long) Integer.MAX_VALUE ?
Integer.MAX_VALUE : (int) avail);
}
}
/**
* Returns the path name of the ZIP file.
* @return the path name of the ZIP file
*/
public String getName() {
return name;
}
private class ZipEntryIterator<T extends ZipEntry>
implements Enumeration<T>, Iterator<T> {
private int i = 0;
private final int entryCount;
private final Function<String, T> gen;
public ZipEntryIterator(int entryCount, Function<String, T> gen) {
this.entryCount = entryCount;
this.gen = gen;
}
@Override
public boolean hasMoreElements() {
return hasNext();
}
@Override
public boolean hasNext() {
return i < entryCount;
}
@Override
public T nextElement() {
return next();
}
@Override
@SuppressWarnings("unchecked")
public T next() {
synchronized (ZipFile.this) {
ensureOpen();
if (!hasNext()) {
throw new NoSuchElementException();
}
// each "entry" has 3 ints in table entries
return (T)getZipEntry(null, null, res.zsrc.getEntryPos(i++ * 3), gen);
}
}
@Override
public Iterator<T> asIterator() {
return this;
}
}
/**
* Returns an enumeration of the ZIP file entries.
* @return an enumeration of the ZIP file entries
* @throws IllegalStateException if the zip file has been closed
*/
public Enumeration<? extends ZipEntry> entries() {
synchronized (this) {
ensureOpen();
return new ZipEntryIterator<ZipEntry>(res.zsrc.total, ZipEntry::new);
}
}
private Enumeration<JarEntry> entries(Function<String, JarEntry> func) {
synchronized (this) {
ensureOpen();
return new ZipEntryIterator<JarEntry>(res.zsrc.total, func);
}
}
private class EntrySpliterator<T> extends Spliterators.AbstractSpliterator<T> {
private int index;
private final int fence;
private final IntFunction<T> gen;
EntrySpliterator(int index, int fence, IntFunction<T> gen) {
super((long)fence,
Spliterator.ORDERED | Spliterator.DISTINCT | Spliterator.IMMUTABLE |
Spliterator.NONNULL);
this.index = index;
this.fence = fence;
this.gen = gen;
}
@Override
public boolean tryAdvance(Consumer<? super T> action) {
if (action == null)
throw new NullPointerException();
if (index >= 0 && index < fence) {
synchronized (ZipFile.this) {
ensureOpen();
action.accept(gen.apply(res.zsrc.getEntryPos(index++ * 3)));
}
return true;
}
return false;
}
}
/**
* Returns an ordered {@code Stream} over the ZIP file entries.
*
* Entries appear in the {@code Stream} in the order they appear in
* the central directory of the ZIP file.
*
* @return an ordered {@code Stream} of entries in this ZIP file
* @throws IllegalStateException if the zip file has been closed
* @since 1.8
*/
public Stream<? extends ZipEntry> stream() {
synchronized (this) {
ensureOpen();
return StreamSupport.stream(new EntrySpliterator<>(0, res.zsrc.total,
pos -> getZipEntry(null, null, pos, ZipEntry::new)), false);
}
}
private String getEntryName(int pos) {
byte[] cen = res.zsrc.cen;
int nlen = CENNAM(cen, pos);
int clen = CENCOM(cen, pos);
int flag = CENFLG(cen, pos);
if (!zc.isUTF8() && (flag & EFS) != 0) {
return zc.toStringUTF8(cen, pos + CENHDR, nlen);
} else {
return zc.toString(cen, pos + CENHDR, nlen);
}
}
/*
* Returns an ordered {@code Stream} over the zip file entry names.
*
* Entry names appear in the {@code Stream} in the order they appear in
* the central directory of the ZIP file.
*
* @return an ordered {@code Stream} of entry names in this zip file
* @throws IllegalStateException if the zip file has been closed
* @since 10
*/
private Stream<String> entryNameStream() {
synchronized (this) {
ensureOpen();
return StreamSupport.stream(
new EntrySpliterator<>(0, res.zsrc.total, this::getEntryName), false);
}
}
/*
* Returns an ordered {@code Stream} over the zip file entries.
*
* Entries appear in the {@code Stream} in the order they appear in
* the central directory of the jar file.
*
* @param func the function that creates the returned entry
* @return an ordered {@code Stream} of entries in this zip file
* @throws IllegalStateException if the zip file has been closed
* @since 10
*/
private Stream<JarEntry> stream(Function<String, JarEntry> func) {
synchronized (this) {
ensureOpen();
return StreamSupport.stream(new EntrySpliterator<>(0, res.zsrc.total,
pos -> (JarEntry)getZipEntry(null, null, pos, func)), false);
}
}
private String lastEntryName;
private int lastEntryPos;
/* Checks ensureOpen() before invoke this method */
private ZipEntry getZipEntry(String name, byte[] bname, int pos,
Function<String, ? extends ZipEntry> func) {
byte[] cen = res.zsrc.cen;
int nlen = CENNAM(cen, pos);
int elen = CENEXT(cen, pos);
int clen = CENCOM(cen, pos);
int flag = CENFLG(cen, pos);
if (name == null || bname.length != nlen) {
// to use the entry name stored in cen, if the passed in name is
// (1) null, invoked from iterator, or
// (2) not equal to the name stored, a slash is appended during
// getEntryPos() search.
if (!zc.isUTF8() && (flag & EFS) != 0) {
name = zc.toStringUTF8(cen, pos + CENHDR, nlen);
} else {
name = zc.toString(cen, pos + CENHDR, nlen);
}
}
ZipEntry e = func.apply(name); //ZipEntry e = new ZipEntry(name);
e.flag = flag;
e.xdostime = CENTIM(cen, pos);
e.crc = CENCRC(cen, pos);
e.size = CENLEN(cen, pos);
e.csize = CENSIZ(cen, pos);
e.method = CENHOW(cen, pos);
if (elen != 0) {
int start = pos + CENHDR + nlen;
e.setExtra0(Arrays.copyOfRange(cen, start, start + elen), true);
}
if (clen != 0) {
int start = pos + CENHDR + nlen + elen;
if (!zc.isUTF8() && (flag & EFS) != 0) {
e.comment = zc.toStringUTF8(cen, start, clen);
} else {
e.comment = zc.toString(cen, start, clen);
}
}
lastEntryName = e.name;
lastEntryPos = pos;
return e;
}
/**
* Returns the number of entries in the ZIP file.
*
* @return the number of entries in the ZIP file
* @throws IllegalStateException if the zip file has been closed
*/
public int size() {
synchronized (this) {
ensureOpen();
return res.zsrc.total;
}
}
private static class CleanableResource implements Runnable {
// The outstanding inputstreams that need to be closed
final Set<InputStream> istreams;
// List of cached Inflater objects for decompression
Deque<Inflater> inflaterCache;
final Cleanable cleanable;
Source zsrc;
CleanableResource(ZipFile zf, File file, int mode) throws IOException {
this.cleanable = CleanerFactory.cleaner().register(zf, this);
this.istreams = Collections.newSetFromMap(new WeakHashMap<>());
this.inflaterCache = new ArrayDeque<>();
this.zsrc = Source.get(file, (mode & OPEN_DELETE) != 0);
}
void clean() {
cleanable.clean();
}
/*
* Gets an inflater from the list of available inflaters or allocates
* a new one.
*/
Inflater getInflater() {
Inflater inf;
synchronized (inflaterCache) {
if ((inf = inflaterCache.poll()) != null) {
return inf;
}
}
return new Inflater(true);
}
/*
* Releases the specified inflater to the list of available inflaters.
*/
void releaseInflater(Inflater inf) {
Deque<Inflater> inflaters = this.inflaterCache;
if (inflaters != null) {
synchronized (inflaters) {
// double checked!
if (inflaters == this.inflaterCache) {
inf.reset();
inflaters.add(inf);
return;
}
}
}
// inflaters cache already closed - just end it.
inf.end();
}
public void run() {
IOException ioe = null;
// Release cached inflaters and close the cache first
Deque<Inflater> inflaters = this.inflaterCache;
if (inflaters != null) {
synchronized (inflaters) {
// no need to double-check as only one thread gets a
// chance to execute run() (Cleaner guarantee)...
Inflater inf;
while ((inf = inflaters.poll()) != null) {
inf.end();
}
// close inflaters cache
this.inflaterCache = null;
}
}
// Close streams, release their inflaters
if (istreams != null) {
synchronized (istreams) {
if (!istreams.isEmpty()) {
InputStream[] copy = istreams.toArray(new InputStream[0]);
istreams.clear();
for (InputStream is : copy) {
try {
is.close();
} catch (IOException e) {
if (ioe == null) ioe = e;
else ioe.addSuppressed(e);
}
}
}
}
}
// Release zip src
if (zsrc != null) {
synchronized (zsrc) {
try {
Source.release(zsrc);
zsrc = null;
} catch (IOException e) {
if (ioe == null) ioe = e;
else ioe.addSuppressed(e);
}
}
}
if (ioe != null) {
throw new UncheckedIOException(ioe);
}
}
CleanableResource(File file, int mode)
throws IOException {
this.cleanable = null;
this.istreams = Collections.newSetFromMap(new WeakHashMap<>());
this.inflaterCache = new ArrayDeque<>();
this.zsrc = Source.get(file, (mode & OPEN_DELETE) != 0);
}
/*
* If {@code ZipFile} has been subclassed and the {@code close} method is
* overridden, uses the {@code finalizer} mechanism for resource cleanup.
* So {@code close} method can be called when the the {@code ZipFile} is
* unreachable. This mechanism will be removed when {@code finalize} method
* is removed from {@code ZipFile}.
*/
static CleanableResource get(ZipFile zf, File file, int mode)
throws IOException {
Class<?> clz = zf.getClass();
while (clz != ZipFile.class) {
try {
clz.getDeclaredMethod("close");
return new FinalizableResource(zf, file, mode);
} catch (NoSuchMethodException nsme) {}
clz = clz.getSuperclass();
}
return new CleanableResource(zf, file, mode);
}
static class FinalizableResource extends CleanableResource {
ZipFile zf;
FinalizableResource(ZipFile zf, File file, int mode)
throws IOException {
super(file, mode);
this.zf = zf;
}
@Override
void clean() {
run();
}
@Override
@SuppressWarnings("deprecation")
protected void finalize() throws IOException {
zf.close();
}
}
}
/**
* Closes the ZIP file.
*
* <p> Closing this ZIP file will close all of the input streams
* previously returned by invocations of the {@link #getInputStream
* getInputStream} method.
*
* @throws IOException if an I/O error has occurred
*/
public void close() throws IOException {
if (closeRequested) {
return;
}
closeRequested = true;
synchronized (this) {
// Close streams, release their inflaters, release cached inflaters
// and release zip source
try {
res.clean();
} catch (UncheckedIOException ioe) {
throw ioe.getCause();
}
}
}
/**
* Ensures that the system resources held by this ZipFile object are
* released when there are no more references to it.
*
* @deprecated The {@code finalize} method has been deprecated and will be
* removed. It is implemented as a no-op. Subclasses that override
* {@code finalize} in order to perform cleanup should be modified to
* use alternative cleanup mechanisms and to remove the overriding
* {@code finalize} method. The recommended cleanup for ZipFile object
* is to explicitly invoke {@code close} method when it is no longer in
* use, or use try-with-resources. If the {@code close} is not invoked
* explicitly the resources held by this object will be released when
* the instance becomes unreachable.
*
* @throws IOException if an I/O error has occurred
*/
@Deprecated(since="9", forRemoval=true)
protected void finalize() throws IOException {}
private void ensureOpen() {
if (closeRequested) {
throw new IllegalStateException("zip file closed");
}
if (res.zsrc == null) {
throw new IllegalStateException("The object is not initialized.");
}
}
private void ensureOpenOrZipException() throws IOException {
if (closeRequested) {
throw new ZipException("ZipFile closed");
}
}
/*
* Inner class implementing the input stream used to read a
* (possibly compressed) zip file entry.
*/
private class ZipFileInputStream extends InputStream {
private volatile boolean closeRequested;
private long pos; // current position within entry data
protected long rem; // number of remaining bytes within entry
protected long size; // uncompressed size of this entry
ZipFileInputStream(byte[] cen, int cenpos) {
rem = CENSIZ(cen, cenpos);
size = CENLEN(cen, cenpos);
pos = CENOFF(cen, cenpos);
// zip64
if (rem == ZIP64_MAGICVAL || size == ZIP64_MAGICVAL ||
pos == ZIP64_MAGICVAL) {
checkZIP64(cen, cenpos);
}
// negative for lazy initialization, see getDataOffset();
pos = - (pos + ZipFile.this.res.zsrc.locpos);
}
private void checkZIP64(byte[] cen, int cenpos) {
int off = cenpos + CENHDR + CENNAM(cen, cenpos);
int end = off + CENEXT(cen, cenpos);
while (off + 4 < end) {
int tag = get16(cen, off);
int sz = get16(cen, off + 2);
off += 4;
if (off + sz > end) // invalid data
break;
if (tag == EXTID_ZIP64) {
if (size == ZIP64_MAGICVAL) {
if (sz < 8 || (off + 8) > end)
break;
size = get64(cen, off);
sz -= 8;
off += 8;
}
if (rem == ZIP64_MAGICVAL) {
if (sz < 8 || (off + 8) > end)
break;
rem = get64(cen, off);
sz -= 8;
off += 8;
}
if (pos == ZIP64_MAGICVAL) {
if (sz < 8 || (off + 8) > end)
break;
pos = get64(cen, off);
sz -= 8;
off += 8;
}
break;
}
off += sz;
}
}
/* The Zip file spec explicitly allows the LOC extra data size to
* be different from the CEN extra data size. Since we cannot trust
* the CEN extra data size, we need to read the LOC to determine
* the entry data offset.
*/
private long initDataOffset() throws IOException {
if (pos <= 0) {
byte[] loc = new byte[LOCHDR];
pos = -pos;
int len = ZipFile.this.res.zsrc.readFullyAt(loc, 0, loc.length, pos);
if (len != LOCHDR) {
throw new ZipException("ZipFile error reading zip file");
}
if (LOCSIG(loc) != LOCSIG) {
throw new ZipException("ZipFile invalid LOC header (bad signature)");
}
pos += LOCHDR + LOCNAM(loc) + LOCEXT(loc);
}
return pos;
}
public int read(byte b[], int off, int len) throws IOException {
synchronized (ZipFile.this) {
ensureOpenOrZipException();
initDataOffset();
if (rem == 0) {
return -1;
}
if (len > rem) {
len = (int) rem;
}
if (len <= 0) {
return 0;
}
len = ZipFile.this.res.zsrc.readAt(b, off, len, pos);
if (len > 0) {
pos += len;
rem -= len;
}
}
if (rem == 0) {
close();
}
return len;
}
public int read() throws IOException {
byte[] b = new byte[1];
if (read(b, 0, 1) == 1) {
return b[0] & 0xff;
} else {
return -1;
}
}
public long skip(long n) throws IOException {
synchronized (ZipFile.this) {
initDataOffset();
if (n > rem) {
n = rem;
}
pos += n;
rem -= n;
}
if (rem == 0) {
close();
}
return n;
}
public int available() {
return rem > Integer.MAX_VALUE ? Integer.MAX_VALUE : (int) rem;
}
public long size() {
return size;
}
public void close() {
if (closeRequested) {
return;
}
closeRequested = true;
rem = 0;
synchronized (res.istreams) {
res.istreams.remove(this);
}
}
}
/**
* Returns the names of all non-directory entries that begin with
* "META-INF/" (case ignored). This method is used in JarFile, via
* SharedSecrets, as an optimization when looking up manifest and
* signature file entries. Returns null if no entries were found.
*/
private String[] getMetaInfEntryNames() {
synchronized (this) {
ensureOpen();
Source zsrc = res.zsrc;
if (zsrc.metanames == null) {
return null;
}
String[] names = new String[zsrc.metanames.length];
byte[] cen = zsrc.cen;
for (int i = 0; i < names.length; i++) {
int pos = zsrc.metanames[i];
names[i] = new String(cen, pos + CENHDR, CENNAM(cen, pos),
StandardCharsets.UTF_8);
}
return names;
}
}
private static boolean isWindows;
static {
SharedSecrets.setJavaUtilZipFileAccess(
new JavaUtilZipFileAccess() {
@Override
public boolean startsWithLocHeader(ZipFile zip) {
return zip.res.zsrc.startsWithLoc;
}
@Override
public String[] getMetaInfEntryNames(ZipFile zip) {
return zip.getMetaInfEntryNames();
}
@Override
public JarEntry getEntry(ZipFile zip, String name,
Function<String, JarEntry> func) {
return (JarEntry)zip.getEntry(name, func);
}
@Override
public Enumeration<JarEntry> entries(ZipFile zip,
Function<String, JarEntry> func) {
return zip.entries(func);
}
@Override
public Stream<JarEntry> stream(ZipFile zip,
Function<String, JarEntry> func) {
return zip.stream(func);
}
@Override
public Stream<String> entryNameStream(ZipFile zip) {
return zip.entryNameStream();
}
}
);
isWindows = VM.getSavedProperty("os.name").contains("Windows");
}
private static class Source {
private final Key key; // the key in files
private int refs = 1;
private RandomAccessFile zfile; // zfile of the underlying zip file
private byte[] cen; // CEN & ENDHDR
private long locpos; // position of first LOC header (usually 0)
private byte[] comment; // zip file comment
// list of meta entries in META-INF dir
private int[] metanames;
private final boolean startsWithLoc; // true, if zip file starts with LOCSIG (usually true)
// A Hashmap for all entries.
//
// A cen entry of Zip/JAR file. As we have one for every entry in every active Zip/JAR,
// We might have a lot of these in a typical system. In order to save space we don't
// keep the name in memory, but merely remember a 32 bit {@code hash} value of the
// entry name and its offset {@code pos} in the central directory hdeader.
//
// private static class Entry {
// int hash; // 32 bit hashcode on name
// int next; // hash chain: index into entries
// int pos; // Offset of central directory file header
// }
// private Entry[] entries; // array of hashed cen entry
//
// To reduce the total size of entries further, we use a int[] here to store 3 "int"
// {@code hash}, {@code next and {@code "pos for each entry. The entry can then be
// referred by their index of their positions in the {@code entries}.
//
private int[] entries; // array of hashed cen entry
private int addEntry(int index, int hash, int next, int pos) {
entries[index++] = hash;
entries[index++] = next;
entries[index++] = pos;
return index;
}
private int getEntryHash(int index) { return entries[index]; }
private int getEntryNext(int index) { return entries[index + 1]; }
private int getEntryPos(int index) { return entries[index + 2]; }
private static final int ZIP_ENDCHAIN = -1;
private int total; // total number of entries
private int[] table; // Hash chain heads: indexes into entries
private int tablelen; // number of hash heads
private static class Key {
BasicFileAttributes attrs;
File file;
public Key(File file, BasicFileAttributes attrs) {
this.attrs = attrs;
this.file = file;
}
public int hashCode() {
long t = attrs.lastModifiedTime().toMillis();
return ((int)(t ^ (t >>> 32))) + file.hashCode();
}
public boolean equals(Object obj) {
if (obj instanceof Key) {
Key key = (Key)obj;
if (!attrs.lastModifiedTime().equals(key.attrs.lastModifiedTime())) {
return false;
}
Object fk = attrs.fileKey();
if (fk != null) {
return fk.equals(key.attrs.fileKey());
} else {
return file.equals(key.file);
}
}
return false;
}
}
private static final HashMap<Key, Source> files = new HashMap<>();
static Source get(File file, boolean toDelete) throws IOException {
Key key = new Key(file,
Files.readAttributes(file.toPath(), BasicFileAttributes.class));
Source src = null;
synchronized (files) {
src = files.get(key);
if (src != null) {
src.refs++;
return src;
}
}
src = new Source(key, toDelete);
synchronized (files) {
if (files.containsKey(key)) { // someone else put in first
src.close(); // close the newly created one
src = files.get(key);
src.refs++;
return src;
}
files.put(key, src);
return src;
}
}
static void release(Source src) throws IOException {
synchronized (files) {
if (src != null && --src.refs == 0) {
files.remove(src.key);
src.close();
}
}
}
private Source(Key key, boolean toDelete) throws IOException {
this.key = key;
if (toDelete) {
if (isWindows) {
this.zfile = SharedSecrets.getJavaIORandomAccessFileAccess()
.openAndDelete(key.file, "r");
} else {
this.zfile = new RandomAccessFile(key.file, "r");
key.file.delete();
}
} else {
this.zfile = new RandomAccessFile(key.file, "r");
}
try {
initCEN(-1);
byte[] buf = new byte[4];
readFullyAt(buf, 0, 4, 0);
this.startsWithLoc = (LOCSIG(buf) == LOCSIG);
} catch (IOException x) {
try {
this.zfile.close();
} catch (IOException xx) {}
throw x;
}
}
private void close() throws IOException {
zfile.close();
zfile = null;
cen = null;
entries = null;
table = null;
metanames = null;
}
private static final int BUF_SIZE = 8192;
private final int readFullyAt(byte[] buf, int off, int len, long pos)
throws IOException
{
synchronized(zfile) {
zfile.seek(pos);
int N = len;
while (N > 0) {
int n = Math.min(BUF_SIZE, N);
zfile.readFully(buf, off, n);
off += n;
N -= n;
}
return len;
}
}
private final int readAt(byte[] buf, int off, int len, long pos)
throws IOException
{
synchronized(zfile) {
zfile.seek(pos);
return zfile.read(buf, off, len);
}
}
private static final int hashN(byte[] a, int off, int len) {
int h = 1;
while (len-- > 0) {
h = 31 * h + a[off++];
}
return h;
}
private static final int hash_append(int hash, byte b) {
return hash * 31 + b;
}
private static class End {
int centot; // 4 bytes
long cenlen; // 4 bytes
long cenoff; // 4 bytes
long endpos; // 4 bytes
}
/*
* Searches for end of central directory (END) header. The contents of
* the END header will be read and placed in endbuf. Returns the file
* position of the END header, otherwise returns -1 if the END header
* was not found or an error occurred.
*/
private End findEND() throws IOException {
long ziplen = zfile.length();
if (ziplen <= 0)
zerror("zip file is empty");
End end = new End();
byte[] buf = new byte[READBLOCKSZ];
long minHDR = (ziplen - END_MAXLEN) > 0 ? ziplen - END_MAXLEN : 0;
long minPos = minHDR - (buf.length - ENDHDR);
for (long pos = ziplen - buf.length; pos >= minPos; pos -= (buf.length - ENDHDR)) {
int off = 0;
if (pos < 0) {
// Pretend there are some NUL bytes before start of file
off = (int)-pos;
Arrays.fill(buf, 0, off, (byte)0);
}
int len = buf.length - off;
if (readFullyAt(buf, off, len, pos + off) != len ) {
zerror("zip END header not found");
}
// Now scan the block backwards for END header signature
for (int i = buf.length - ENDHDR; i >= 0; i--) {
if (buf[i+0] == (byte)'P' &&
buf[i+1] == (byte)'K' &&
buf[i+2] == (byte)'\005' &&
buf[i+3] == (byte)'\006') {
// Found ENDSIG header
byte[] endbuf = Arrays.copyOfRange(buf, i, i + ENDHDR);
end.centot = ENDTOT(endbuf);
end.cenlen = ENDSIZ(endbuf);
end.cenoff = ENDOFF(endbuf);
end.endpos = pos + i;
int comlen = ENDCOM(endbuf);
if (end.endpos + ENDHDR + comlen != ziplen) {
// ENDSIG matched, however the size of file comment in it does
// not match the real size. One "common" cause for this problem
// is some "extra" bytes are padded at the end of the zipfile.
// Let's do some extra verification, we don't care about the
// performance in this situation.
byte[] sbuf = new byte[4];
long cenpos = end.endpos - end.cenlen;
long locpos = cenpos - end.cenoff;
if (cenpos < 0 ||
locpos < 0 ||
readFullyAt(sbuf, 0, sbuf.length, cenpos) != 4 ||
GETSIG(sbuf) != CENSIG ||
readFullyAt(sbuf, 0, sbuf.length, locpos) != 4 ||
GETSIG(sbuf) != LOCSIG) {
continue;
}
}
if (comlen > 0) { // this zip file has comlen
comment = new byte[comlen];
if (readFullyAt(comment, 0, comlen, end.endpos + ENDHDR) != comlen) {
zerror("zip comment read failed");
}
}
// must check for a zip64 end record; it is always permitted to be present
try {
byte[] loc64 = new byte[ZIP64_LOCHDR];
if (end.endpos < ZIP64_LOCHDR ||
readFullyAt(loc64, 0, loc64.length, end.endpos - ZIP64_LOCHDR)
!= loc64.length || GETSIG(loc64) != ZIP64_LOCSIG) {
return end;
}
long end64pos = ZIP64_LOCOFF(loc64);
byte[] end64buf = new byte[ZIP64_ENDHDR];
if (readFullyAt(end64buf, 0, end64buf.length, end64pos)
!= end64buf.length || GETSIG(end64buf) != ZIP64_ENDSIG) {
return end;
}
// end64 candidate found,
long cenlen64 = ZIP64_ENDSIZ(end64buf);
long cenoff64 = ZIP64_ENDOFF(end64buf);
long centot64 = ZIP64_ENDTOT(end64buf);
// double-check
if (cenlen64 != end.cenlen && end.cenlen != ZIP64_MAGICVAL ||
cenoff64 != end.cenoff && end.cenoff != ZIP64_MAGICVAL ||
centot64 != end.centot && end.centot != ZIP64_MAGICCOUNT) {
return end;
}
// to use the end64 values
end.cenlen = cenlen64;
end.cenoff = cenoff64;
end.centot = (int)centot64; // assume total < 2g
end.endpos = end64pos;
} catch (IOException x) {} // no zip64 loc/end
return end;
}
}
}
zerror("zip END header not found");
return null; //make compiler happy
}
// Reads zip file central directory.
private void initCEN(int knownTotal) throws IOException {
if (knownTotal == -1) {
End end = findEND();
if (end.endpos == 0) {
locpos = 0;
total = 0;
entries = new int[0];
cen = null;
return; // only END header present
}
if (end.cenlen > end.endpos)
zerror("invalid END header (bad central directory size)");
long cenpos = end.endpos - end.cenlen; // position of CEN table
// Get position of first local file (LOC) header, taking into
// account that there may be a stub prefixed to the zip file.
locpos = cenpos - end.cenoff;
if (locpos < 0) {
zerror("invalid END header (bad central directory offset)");
}
// read in the CEN and END
cen = new byte[(int)(end.cenlen + ENDHDR)];
if (readFullyAt(cen, 0, cen.length, cenpos) != end.cenlen + ENDHDR) {
zerror("read CEN tables failed");
}
total = end.centot;
} else {
total = knownTotal;
}
// hash table for entries
entries = new int[total * 3];
tablelen = ((total/2) | 1); // Odd -> fewer collisions
table = new int[tablelen];
Arrays.fill(table, ZIP_ENDCHAIN);
int idx = 0;
int hash = 0;
int next = -1;
// list for all meta entries
ArrayList<Integer> metanamesList = null;
// Iterate through the entries in the central directory
int i = 0;
int hsh = 0;
int pos = 0;
int limit = cen.length - ENDHDR;
while (pos + CENHDR <= limit) {
if (i >= total) {
// This will only happen if the zip file has an incorrect
// ENDTOT field, which usually means it contains more than
// 65535 entries.
initCEN(countCENHeaders(cen, limit));
return;
}
if (CENSIG(cen, pos) != CENSIG)
zerror("invalid CEN header (bad signature)");
int method = CENHOW(cen, pos);
int nlen = CENNAM(cen, pos);
int elen = CENEXT(cen, pos);
int clen = CENCOM(cen, pos);
if ((CENFLG(cen, pos) & 1) != 0)
zerror("invalid CEN header (encrypted entry)");
if (method != STORED && method != DEFLATED)
zerror("invalid CEN header (bad compression method: " + method + ")");
if (pos + CENHDR + nlen > limit)
zerror("invalid CEN header (bad header size)");
// Record the CEN offset and the name hash in our hash cell.
hash = hashN(cen, pos + CENHDR, nlen);
hsh = (hash & 0x7fffffff) % tablelen;
next = table[hsh];
table[hsh] = idx;
idx = addEntry(idx, hash, next, pos);
// Adds name to metanames.
if (isMetaName(cen, pos + CENHDR, nlen)) {
if (metanamesList == null)
metanamesList = new ArrayList<>(4);
metanamesList.add(pos);
}
// skip ext and comment
pos += (CENHDR + nlen + elen + clen);
i++;
}
total = i;
if (metanamesList != null) {
metanames = new int[metanamesList.size()];
for (int j = 0, len = metanames.length; j < len; j++) {
metanames[j] = metanamesList.get(j);
}
}
if (pos + ENDHDR != cen.length) {
zerror("invalid CEN header (bad header size)");
}
}
private static void zerror(String msg) throws ZipException {
throw new ZipException(msg);
}
/*
* Returns the {@code pos} of the zip cen entry corresponding to the
* specified entry name, or -1 if not found.
*/
private int getEntryPos(byte[] name, boolean addSlash) {
if (total == 0) {
return -1;
}
int hsh = hashN(name, 0, name.length);
int idx = table[(hsh & 0x7fffffff) % tablelen];
/*
* This while loop is an optimization where a double lookup
* for name and name+/ is being performed. The name char
* array has enough room at the end to try again with a
* slash appended if the first table lookup does not succeed.
*/
while(true) {
/*
* Search down the target hash chain for a entry whose
* 32 bit hash matches the hashed name.
*/
while (idx != ZIP_ENDCHAIN) {
if (getEntryHash(idx) == hsh) {
// The CEN name must match the specfied one
int pos = getEntryPos(idx);
if (name.length == CENNAM(cen, pos)) {
boolean matched = true;
int nameoff = pos + CENHDR;
for (int i = 0; i < name.length; i++) {
if (name[i] != cen[nameoff++]) {
matched = false;
break;
}
}
if (matched) {
return pos;
}
}
}
idx = getEntryNext(idx);
}
/* If not addSlash, or slash is already there, we are done */
if (!addSlash || name.length == 0 || name[name.length - 1] == '/') {
return -1;
}
/* Add slash and try once more */
name = Arrays.copyOf(name, name.length + 1);
name[name.length - 1] = '/';
hsh = hash_append(hsh, (byte)'/');
//idx = table[hsh % tablelen];
idx = table[(hsh & 0x7fffffff) % tablelen];
addSlash = false;
}
}
/**
* Returns true if the bytes represent a non-directory name
* beginning with "META-INF/", disregarding ASCII case.
*/
private static boolean isMetaName(byte[] name, int off, int len) {
// Use the "oldest ASCII trick in the book"
return len > 9 // "META-INF/".length()
&& name[off + len - 1] != '/' // non-directory
&& (name[off++] | 0x20) == 'm'
&& (name[off++] | 0x20) == 'e'
&& (name[off++] | 0x20) == 't'
&& (name[off++] | 0x20) == 'a'
&& (name[off++] ) == '-'
&& (name[off++] | 0x20) == 'i'
&& (name[off++] | 0x20) == 'n'
&& (name[off++] | 0x20) == 'f'
&& (name[off] ) == '/';
}
/**
* Returns the number of CEN headers in a central directory.
* Will not throw, even if the zip file is corrupt.
*
* @param cen copy of the bytes in a zip file's central directory
* @param size number of bytes in central directory
*/
private static int countCENHeaders(byte[] cen, int size) {
int count = 0;
for (int p = 0;
p + CENHDR <= size;
p += CENHDR + CENNAM(cen, p) + CENEXT(cen, p) + CENCOM(cen, p))
count++;
return count;
}
}
}