aboutsummaryrefslogtreecommitdiffstats
path: root/framework/src/ant/apache-ant-1.9.6/src/main/org/apache/tools/tar/TarOutputStream.java
diff options
context:
space:
mode:
Diffstat (limited to 'framework/src/ant/apache-ant-1.9.6/src/main/org/apache/tools/tar/TarOutputStream.java')
-rw-r--r--framework/src/ant/apache-ant-1.9.6/src/main/org/apache/tools/tar/TarOutputStream.java657
1 files changed, 657 insertions, 0 deletions
diff --git a/framework/src/ant/apache-ant-1.9.6/src/main/org/apache/tools/tar/TarOutputStream.java b/framework/src/ant/apache-ant-1.9.6/src/main/org/apache/tools/tar/TarOutputStream.java
new file mode 100644
index 00000000..032f7254
--- /dev/null
+++ b/framework/src/ant/apache-ant-1.9.6/src/main/org/apache/tools/tar/TarOutputStream.java
@@ -0,0 +1,657 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one or more
+ * contributor license agreements. See the NOTICE file distributed with
+ * this work for additional information regarding copyright ownership.
+ * The ASF licenses this file to You under the Apache License, Version 2.0
+ * (the "License"); you may not use this file except in compliance with
+ * the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ */
+
+/*
+ * This package is based on the work done by Timothy Gerard Endres
+ * (time@ice.com) to whom the Ant project is very grateful for his great code.
+ */
+
+package org.apache.tools.tar;
+
+import java.io.FilterOutputStream;
+import java.io.IOException;
+import java.io.OutputStream;
+import java.io.StringWriter;
+import java.nio.ByteBuffer;
+import java.util.Date;
+import java.util.HashMap;
+import java.util.Map;
+
+import org.apache.tools.zip.ZipEncoding;
+import org.apache.tools.zip.ZipEncodingHelper;
+
+/**
+ * The TarOutputStream writes a UNIX tar archive as an OutputStream.
+ * Methods are provided to put entries, and then write their contents
+ * by writing to this stream using write().
+ *
+ */
+public class TarOutputStream extends FilterOutputStream {
+ /** Fail if a long file name is required in the archive. */
+ public static final int LONGFILE_ERROR = 0;
+
+ /** Long paths will be truncated in the archive. */
+ public static final int LONGFILE_TRUNCATE = 1;
+
+ /** GNU tar extensions are used to store long file names in the archive. */
+ public static final int LONGFILE_GNU = 2;
+
+ /** POSIX/PAX extensions are used to store long file names in the archive. */
+ public static final int LONGFILE_POSIX = 3;
+
+ /** Fail if a big number (e.g. size > 8GiB) is required in the archive. */
+ public static final int BIGNUMBER_ERROR = 0;
+
+ /** star/GNU tar/BSD tar extensions are used to store big number in the archive. */
+ public static final int BIGNUMBER_STAR = 1;
+
+ /** POSIX/PAX extensions are used to store big numbers in the archive. */
+ public static final int BIGNUMBER_POSIX = 2;
+
+ // CheckStyle:VisibilityModifier OFF - bc
+ protected boolean debug;
+ protected long currSize;
+ protected String currName;
+ protected long currBytes;
+ protected byte[] oneBuf;
+ protected byte[] recordBuf;
+ protected int assemLen;
+ protected byte[] assemBuf;
+ protected TarBuffer buffer;
+ protected int longFileMode = LONGFILE_ERROR;
+ // CheckStyle:VisibilityModifier ON
+
+ private int bigNumberMode = BIGNUMBER_ERROR;
+
+ private boolean closed = false;
+
+ /** Indicates if putNextEntry has been called without closeEntry */
+ private boolean haveUnclosedEntry = false;
+
+ /** indicates if this archive is finished */
+ private boolean finished = false;
+
+ private final ZipEncoding encoding;
+
+ private boolean addPaxHeadersForNonAsciiNames = false;
+ private static final ZipEncoding ASCII =
+ ZipEncodingHelper.getZipEncoding("ASCII");
+
+ /**
+ * Constructor for TarInputStream.
+ * @param os the output stream to use
+ */
+ public TarOutputStream(OutputStream os) {
+ this(os, TarBuffer.DEFAULT_BLKSIZE, TarBuffer.DEFAULT_RCDSIZE);
+ }
+
+ /**
+ * Constructor for TarInputStream.
+ * @param os the output stream to use
+ * @param encoding name of the encoding to use for file names
+ */
+ public TarOutputStream(OutputStream os, String encoding) {
+ this(os, TarBuffer.DEFAULT_BLKSIZE, TarBuffer.DEFAULT_RCDSIZE, encoding);
+ }
+
+ /**
+ * Constructor for TarInputStream.
+ * @param os the output stream to use
+ * @param blockSize the block size to use
+ */
+ public TarOutputStream(OutputStream os, int blockSize) {
+ this(os, blockSize, TarBuffer.DEFAULT_RCDSIZE);
+ }
+
+ /**
+ * Constructor for TarInputStream.
+ * @param os the output stream to use
+ * @param blockSize the block size to use
+ * @param encoding name of the encoding to use for file names
+ */
+ public TarOutputStream(OutputStream os, int blockSize, String encoding) {
+ this(os, blockSize, TarBuffer.DEFAULT_RCDSIZE, encoding);
+ }
+
+ /**
+ * Constructor for TarInputStream.
+ * @param os the output stream to use
+ * @param blockSize the block size to use
+ * @param recordSize the record size to use
+ */
+ public TarOutputStream(OutputStream os, int blockSize, int recordSize) {
+ this(os, blockSize, recordSize, null);
+ }
+
+ /**
+ * Constructor for TarInputStream.
+ * @param os the output stream to use
+ * @param blockSize the block size to use
+ * @param recordSize the record size to use
+ * @param encoding name of the encoding to use for file names
+ */
+ public TarOutputStream(OutputStream os, int blockSize, int recordSize,
+ String encoding) {
+ super(os);
+ this.encoding = ZipEncodingHelper.getZipEncoding(encoding);
+
+ this.buffer = new TarBuffer(os, blockSize, recordSize);
+ this.debug = false;
+ this.assemLen = 0;
+ this.assemBuf = new byte[recordSize];
+ this.recordBuf = new byte[recordSize];
+ this.oneBuf = new byte[1];
+ }
+
+ /**
+ * Set the long file mode.
+ * This can be LONGFILE_ERROR(0), LONGFILE_TRUNCATE(1) or LONGFILE_GNU(2).
+ * This specifies the treatment of long file names (names >= TarConstants.NAMELEN).
+ * Default is LONGFILE_ERROR.
+ * @param longFileMode the mode to use
+ */
+ public void setLongFileMode(int longFileMode) {
+ this.longFileMode = longFileMode;
+ }
+
+ /**
+ * Set the big number mode.
+ * This can be BIGNUMBER_ERROR(0), BIGNUMBER_POSIX(1) or BIGNUMBER_STAR(2).
+ * This specifies the treatment of big files (sizes > TarConstants.MAXSIZE) and other numeric values to big to fit into a traditional tar header.
+ * Default is BIGNUMBER_ERROR.
+ * @param bigNumberMode the mode to use
+ */
+ public void setBigNumberMode(int bigNumberMode) {
+ this.bigNumberMode = bigNumberMode;
+ }
+
+ /**
+ * Whether to add a PAX extension header for non-ASCII file names.
+ */
+ public void setAddPaxHeadersForNonAsciiNames(boolean b) {
+ addPaxHeadersForNonAsciiNames = b;
+ }
+
+ /**
+ * Sets the debugging flag.
+ *
+ * @param debugF True to turn on debugging.
+ */
+ public void setDebug(boolean debugF) {
+ this.debug = debugF;
+ }
+
+ /**
+ * Sets the debugging flag in this stream's TarBuffer.
+ *
+ * @param debug True to turn on debugging.
+ */
+ public void setBufferDebug(boolean debug) {
+ buffer.setDebug(debug);
+ }
+
+ /**
+ * Ends the TAR archive without closing the underlying OutputStream.
+ *
+ * An archive consists of a series of file entries terminated by an
+ * end-of-archive entry, which consists of two 512 blocks of zero bytes.
+ * POSIX.1 requires two EOF records, like some other implementations.
+ *
+ * @throws IOException on error
+ */
+ public void finish() throws IOException {
+ if (finished) {
+ throw new IOException("This archive has already been finished");
+ }
+
+ if (haveUnclosedEntry) {
+ throw new IOException("This archives contains unclosed entries.");
+ }
+ writeEOFRecord();
+ writeEOFRecord();
+ buffer.flushBlock();
+ finished = true;
+ }
+
+ /**
+ * Ends the TAR archive and closes the underlying OutputStream.
+ * This means that finish() is called followed by calling the
+ * TarBuffer's close().
+ * @throws IOException on error
+ */
+ @Override
+ public void close() throws IOException {
+ if(!finished) {
+ finish();
+ }
+
+ if (!closed) {
+ buffer.close();
+ out.close();
+ closed = true;
+ }
+ }
+
+ /**
+ * Get the record size being used by this stream's TarBuffer.
+ *
+ * @return The TarBuffer record size.
+ */
+ public int getRecordSize() {
+ return buffer.getRecordSize();
+ }
+
+ /**
+ * Put an entry on the output stream. This writes the entry's
+ * header record and positions the output stream for writing
+ * the contents of the entry. Once this method is called, the
+ * stream is ready for calls to write() to write the entry's
+ * contents. Once the contents are written, closeEntry()
+ * <B>MUST</B> be called to ensure that all buffered data
+ * is completely written to the output stream.
+ *
+ * @param entry The TarEntry to be written to the archive.
+ * @throws IOException on error
+ */
+ public void putNextEntry(TarEntry entry) throws IOException {
+ if(finished) {
+ throw new IOException("Stream has already been finished");
+ }
+ Map<String, String> paxHeaders = new HashMap<String, String>();
+ final String entryName = entry.getName();
+ boolean paxHeaderContainsPath = handleLongName(entry, entryName, paxHeaders, "path",
+ TarConstants.LF_GNUTYPE_LONGNAME, "file name");
+
+ final String linkName = entry.getLinkName();
+ boolean paxHeaderContainsLinkPath = linkName != null && linkName.length() > 0
+ && handleLongName(entry, linkName, paxHeaders, "linkpath",
+ TarConstants.LF_GNUTYPE_LONGLINK, "link name");
+
+ if (bigNumberMode == BIGNUMBER_POSIX) {
+ addPaxHeadersForBigNumbers(paxHeaders, entry);
+ } else if (bigNumberMode != BIGNUMBER_STAR) {
+ failForBigNumbers(entry);
+ }
+
+ if (addPaxHeadersForNonAsciiNames && !paxHeaderContainsPath
+ && !ASCII.canEncode(entryName)) {
+ paxHeaders.put("path", entryName);
+ }
+
+ if (addPaxHeadersForNonAsciiNames && !paxHeaderContainsLinkPath
+ && (entry.isLink() || entry.isSymbolicLink())
+ && !ASCII.canEncode(linkName)) {
+ paxHeaders.put("linkpath", linkName);
+ }
+
+ if (paxHeaders.size() > 0) {
+ writePaxHeaders(entry, entryName, paxHeaders);
+ }
+
+ entry.writeEntryHeader(recordBuf, encoding,
+ bigNumberMode == BIGNUMBER_STAR);
+ buffer.writeRecord(recordBuf);
+
+ currBytes = 0;
+
+ if (entry.isDirectory()) {
+ currSize = 0;
+ } else {
+ currSize = entry.getSize();
+ }
+ currName = entryName;
+ haveUnclosedEntry = true;
+ }
+
+ /**
+ * Close an entry. This method MUST be called for all file
+ * entries that contain data. The reason is that we must
+ * buffer data written to the stream in order to satisfy
+ * the buffer's record based writes. Thus, there may be
+ * data fragments still being assembled that must be written
+ * to the output stream before this entry is closed and the
+ * next entry written.
+ * @throws IOException on error
+ */
+ public void closeEntry() throws IOException {
+ if (finished) {
+ throw new IOException("Stream has already been finished");
+ }
+ if (!haveUnclosedEntry){
+ throw new IOException("No current entry to close");
+ }
+ if (assemLen > 0) {
+ for (int i = assemLen; i < assemBuf.length; ++i) {
+ assemBuf[i] = 0;
+ }
+
+ buffer.writeRecord(assemBuf);
+
+ currBytes += assemLen;
+ assemLen = 0;
+ }
+
+ if (currBytes < currSize) {
+ throw new IOException("entry '" + currName + "' closed at '"
+ + currBytes
+ + "' before the '" + currSize
+ + "' bytes specified in the header were written");
+ }
+ haveUnclosedEntry = false;
+ }
+
+ /**
+ * Writes a byte to the current tar archive entry.
+ *
+ * This method simply calls read( byte[], int, int ).
+ *
+ * @param b The byte written.
+ * @throws IOException on error
+ */
+ @Override
+ public void write(int b) throws IOException {
+ oneBuf[0] = (byte) b;
+
+ write(oneBuf, 0, 1);
+ }
+
+ /**
+ * Writes bytes to the current tar archive entry.
+ *
+ * This method simply calls write( byte[], int, int ).
+ *
+ * @param wBuf The buffer to write to the archive.
+ * @throws IOException on error
+ */
+ @Override
+ public void write(byte[] wBuf) throws IOException {
+ write(wBuf, 0, wBuf.length);
+ }
+
+ /**
+ * Writes bytes to the current tar archive entry. This method
+ * is aware of the current entry and will throw an exception if
+ * you attempt to write bytes past the length specified for the
+ * current entry. The method is also (painfully) aware of the
+ * record buffering required by TarBuffer, and manages buffers
+ * that are not a multiple of recordsize in length, including
+ * assembling records from small buffers.
+ *
+ * @param wBuf The buffer to write to the archive.
+ * @param wOffset The offset in the buffer from which to get bytes.
+ * @param numToWrite The number of bytes to write.
+ * @throws IOException on error
+ */
+ @Override
+ public void write(byte[] wBuf, int wOffset, int numToWrite) throws IOException {
+ if ((currBytes + numToWrite) > currSize) {
+ throw new IOException("request to write '" + numToWrite
+ + "' bytes exceeds size in header of '"
+ + currSize + "' bytes for entry '"
+ + currName + "'");
+
+ //
+ // We have to deal with assembly!!!
+ // The programmer can be writing little 32 byte chunks for all
+ // we know, and we must assemble complete records for writing.
+ // REVIEW Maybe this should be in TarBuffer? Could that help to
+ // eliminate some of the buffer copying.
+ //
+ }
+
+ if (assemLen > 0) {
+ if ((assemLen + numToWrite) >= recordBuf.length) {
+ int aLen = recordBuf.length - assemLen;
+
+ System.arraycopy(assemBuf, 0, recordBuf, 0,
+ assemLen);
+ System.arraycopy(wBuf, wOffset, recordBuf,
+ assemLen, aLen);
+ buffer.writeRecord(recordBuf);
+
+ currBytes += recordBuf.length;
+ wOffset += aLen;
+ numToWrite -= aLen;
+ assemLen = 0;
+ } else {
+ System.arraycopy(wBuf, wOffset, assemBuf, assemLen,
+ numToWrite);
+
+ wOffset += numToWrite;
+ assemLen += numToWrite;
+ numToWrite = 0;
+ }
+ }
+
+ //
+ // When we get here we have EITHER:
+ // o An empty "assemble" buffer.
+ // o No bytes to write (numToWrite == 0)
+ //
+ while (numToWrite > 0) {
+ if (numToWrite < recordBuf.length) {
+ System.arraycopy(wBuf, wOffset, assemBuf, assemLen,
+ numToWrite);
+
+ assemLen += numToWrite;
+
+ break;
+ }
+
+ buffer.writeRecord(wBuf, wOffset);
+
+ int num = recordBuf.length;
+
+ currBytes += num;
+ numToWrite -= num;
+ wOffset += num;
+ }
+ }
+
+ /**
+ * Writes a PAX extended header with the given map as contents.
+ */
+ void writePaxHeaders(TarEntry entry,
+ String entryName,
+ Map<String, String> headers) throws IOException {
+ String name = "./PaxHeaders.X/" + stripTo7Bits(entryName);
+ if (name.length() >= TarConstants.NAMELEN) {
+ name = name.substring(0, TarConstants.NAMELEN - 1);
+ }
+ while (name.endsWith("/")) {
+ // TarEntry's constructor would think this is a directory
+ // and not allow any data to be written
+ name = name.substring(0, name.length() - 1);
+ }
+ TarEntry pex = new TarEntry(name,
+ TarConstants.LF_PAX_EXTENDED_HEADER_LC);
+ transferModTime(entry, pex);
+
+ StringWriter w = new StringWriter();
+ for (Map.Entry<String, String> h : headers.entrySet()) {
+ String key = h.getKey();
+ String value = h.getValue();
+ int len = key.length() + value.length()
+ + 3 /* blank, equals and newline */
+ + 2 /* guess 9 < actual length < 100 */;
+ String line = len + " " + key + "=" + value + "\n";
+ int actualLength = line.getBytes("UTF-8").length;
+ while (len != actualLength) {
+ // Adjust for cases where length < 10 or > 100
+ // or where UTF-8 encoding isn't a single octet
+ // per character.
+ // Must be in loop as size may go from 99 to 100 in
+ // first pass so we'd need a second.
+ len = actualLength;
+ line = len + " " + key + "=" + value + "\n";
+ actualLength = line.getBytes("UTF-8").length;
+ }
+ w.write(line);
+ }
+ byte[] data = w.toString().getBytes("UTF-8");
+ pex.setSize(data.length);
+ putNextEntry(pex);
+ write(data);
+ closeEntry();
+ }
+
+ private String stripTo7Bits(String name) {
+ final int length = name.length();
+ StringBuilder result = new StringBuilder(length);
+ for (int i = 0; i < length; i++) {
+ char stripped = (char) (name.charAt(i) & 0x7F);
+ if (stripped != 0) { // would be read as Trailing null
+ result.append(stripped);
+ }
+ }
+ return result.toString();
+ }
+
+ /**
+ * Write an EOF (end of archive) record to the tar archive.
+ * An EOF record consists of a record of all zeros.
+ */
+ private void writeEOFRecord() throws IOException {
+ for (int i = 0; i < recordBuf.length; ++i) {
+ recordBuf[i] = 0;
+ }
+
+ buffer.writeRecord(recordBuf);
+ }
+
+ private void addPaxHeadersForBigNumbers(Map<String, String> paxHeaders,
+ TarEntry entry) {
+ addPaxHeaderForBigNumber(paxHeaders, "size", entry.getSize(),
+ TarConstants.MAXSIZE);
+ addPaxHeaderForBigNumber(paxHeaders, "gid", entry.getLongGroupId(),
+ TarConstants.MAXID);
+ addPaxHeaderForBigNumber(paxHeaders, "mtime",
+ entry.getModTime().getTime() / 1000,
+ TarConstants.MAXSIZE);
+ addPaxHeaderForBigNumber(paxHeaders, "uid", entry.getLongUserId(),
+ TarConstants.MAXID);
+ // star extensions by J\u00f6rg Schilling
+ addPaxHeaderForBigNumber(paxHeaders, "SCHILY.devmajor",
+ entry.getDevMajor(), TarConstants.MAXID);
+ addPaxHeaderForBigNumber(paxHeaders, "SCHILY.devminor",
+ entry.getDevMinor(), TarConstants.MAXID);
+ // there is no PAX header for file mode
+ failForBigNumber("mode", entry.getMode(), TarConstants.MAXID);
+ }
+
+ private void addPaxHeaderForBigNumber(Map<String, String> paxHeaders,
+ String header, long value,
+ long maxValue) {
+ if (value < 0 || value > maxValue) {
+ paxHeaders.put(header, String.valueOf(value));
+ }
+ }
+
+ private void failForBigNumbers(TarEntry entry) {
+ failForBigNumber("entry size", entry.getSize(), TarConstants.MAXSIZE);
+ failForBigNumberWithPosixMessage("group id", entry.getLongGroupId(), TarConstants.MAXID);
+ failForBigNumber("last modification time",
+ entry.getModTime().getTime() / 1000,
+ TarConstants.MAXSIZE);
+ failForBigNumber("user id", entry.getLongUserId(), TarConstants.MAXID);
+ failForBigNumber("mode", entry.getMode(), TarConstants.MAXID);
+ failForBigNumber("major device number", entry.getDevMajor(),
+ TarConstants.MAXID);
+ failForBigNumber("minor device number", entry.getDevMinor(),
+ TarConstants.MAXID);
+ }
+
+ private void failForBigNumber(String field, long value, long maxValue) {
+ failForBigNumber(field, value, maxValue, "");
+ }
+
+ private void failForBigNumberWithPosixMessage(String field, long value, long maxValue) {
+ failForBigNumber(field, value, maxValue, " Use STAR or POSIX extensions to overcome this limit");
+ }
+
+ private void failForBigNumber(String field, long value, long maxValue, String additionalMsg) {
+ if (value < 0 || value > maxValue) {
+ throw new RuntimeException(field + " '" + value
+ + "' is too big ( > "
+ + maxValue + " )");
+ }
+ }
+
+ /**
+ * Handles long file or link names according to the longFileMode setting.
+ *
+ * <p>I.e. if the given name is too long to be written to a plain
+ * tar header then
+ * <ul>
+ * <li>it creates a pax header who's name is given by the
+ * paxHeaderName parameter if longFileMode is POSIX</li>
+ * <li>it creates a GNU longlink entry who's type is given by
+ * the linkType parameter if longFileMode is GNU</li>
+ * <li>it throws an exception if longFileMode is ERROR</li>
+ * <li>it truncates the name if longFileMode is TRUNCATE</li>
+ * </ul></p>
+ *
+ * @param entry entry the name belongs to
+ * @param name the name to write
+ * @param paxHeaders current map of pax headers
+ * @param paxHeaderName name of the pax header to write
+ * @param linkType type of the GNU entry to write
+ * @param fieldName the name of the field
+ * @return whether a pax header has been written.
+ */
+ private boolean handleLongName(TarEntry entry , String name,
+ Map<String, String> paxHeaders,
+ String paxHeaderName, byte linkType, String fieldName)
+ throws IOException {
+ final ByteBuffer encodedName = encoding.encode(name);
+ final int len = encodedName.limit() - encodedName.position();
+ if (len >= TarConstants.NAMELEN) {
+
+ if (longFileMode == LONGFILE_POSIX) {
+ paxHeaders.put(paxHeaderName, name);
+ return true;
+ } else if (longFileMode == LONGFILE_GNU) {
+ // create a TarEntry for the LongLink, the contents
+ // of which are the link's name
+ TarEntry longLinkEntry =
+ new TarEntry(TarConstants.GNU_LONGLINK, linkType);
+
+ longLinkEntry.setSize(len + 1); // +1 for NUL
+ transferModTime(entry, longLinkEntry);
+ putNextEntry(longLinkEntry);
+ write(encodedName.array(), encodedName.arrayOffset(), len);
+ write(0); // NUL terminator
+ closeEntry();
+ } else if (longFileMode != LONGFILE_TRUNCATE) {
+ throw new RuntimeException(fieldName + " '" + name
+ + "' is too long ( > "
+ + TarConstants.NAMELEN + " bytes)");
+ }
+ }
+ return false;
+ }
+
+ private void transferModTime(TarEntry from, TarEntry to) {
+ Date fromModTime = from.getModTime();
+ long fromModTimeSeconds = fromModTime.getTime() / 1000;
+ if (fromModTimeSeconds < 0 || fromModTimeSeconds > TarConstants.MAXSIZE) {
+ fromModTime = new Date(0);
+ }
+ to.setModTime(fromModTime);
+ }
+}