Package org.eclipse.jgit.storage.dfs

Source Code of org.eclipse.jgit.storage.dfs.DfsPackFile$Delta

/*
* Copyright (C) 2008-2011, Google Inc.
* Copyright (C) 2007, Robin Rosenberg <robin.rosenberg@dewire.com>
* Copyright (C) 2006-2008, Shawn O. Pearce <spearce@spearce.org>
* and other copyright owners as documented in the project's IP log.
*
* This program and the accompanying materials are made available
* under the terms of the Eclipse Distribution License v1.0 which
* accompanies this distribution, is reproduced below, and is
* available at http://www.eclipse.org/org/documents/edl-v10.php
*
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above copyright
*   notice, this list of conditions and the following disclaimer.
*
* - Redistributions in binary form must reproduce the above
*   copyright notice, this list of conditions and the following
*   disclaimer in the documentation and/or other materials provided
*   with the distribution.
*
* - Neither the name of the Eclipse Foundation, Inc. nor the
*   names of its contributors may be used to endorse or promote
*   products derived from this software without specific prior
*   written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND
* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES,
* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
* CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF
* ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/

package org.eclipse.jgit.storage.dfs;

import java.io.BufferedInputStream;
import java.io.EOFException;
import java.io.IOException;
import java.io.InputStream;
import java.nio.channels.Channels;
import java.text.MessageFormat;
import java.util.Set;
import java.util.zip.CRC32;
import java.util.zip.DataFormatException;
import java.util.zip.Inflater;

import org.eclipse.jgit.JGitText;
import org.eclipse.jgit.errors.CorruptObjectException;
import org.eclipse.jgit.errors.LargeObjectException;
import org.eclipse.jgit.errors.MissingObjectException;
import org.eclipse.jgit.errors.PackInvalidException;
import org.eclipse.jgit.errors.StoredObjectRepresentationNotAvailableException;
import org.eclipse.jgit.lib.AbbreviatedObjectId;
import org.eclipse.jgit.lib.AnyObjectId;
import org.eclipse.jgit.lib.Constants;
import org.eclipse.jgit.lib.ObjectId;
import org.eclipse.jgit.lib.ObjectLoader;
import org.eclipse.jgit.storage.file.PackIndex;
import org.eclipse.jgit.storage.file.PackReverseIndex;
import org.eclipse.jgit.storage.pack.BinaryDelta;
import org.eclipse.jgit.storage.pack.PackOutputStream;
import org.eclipse.jgit.storage.pack.StoredObjectRepresentation;
import org.eclipse.jgit.util.IO;
import org.eclipse.jgit.util.LongList;

/**
* A Git version 2 pack file representation. A pack file contains Git objects in
* delta packed format yielding high compression of lots of object where some
* objects are similar.
*/
public final class DfsPackFile {
  /**
   * File offset used to cache {@link #index} in {@link DfsBlockCache}.
   * <p>
   * To better manage memory, the forward index is stored as a single block in
   * the block cache under this file position. A negative value is used
   * because it cannot occur in a normal pack file, and it is less likely to
   * collide with a valid data block from the file as the high bits will all
   * be set when treated as an unsigned long by the cache code.
   */
  private static final long POS_INDEX = -1;

  /** Offset used to cache {@link #reverseIndex}. See {@link #POS_INDEX}. */
  private static final long POS_REVERSE_INDEX = -2;

  /** Cache that owns this pack file and its data. */
  private final DfsBlockCache cache;

  /** Description of the pack file's storage. */
  private final DfsPackDescription packDesc;

  /** Unique identity of this pack while in-memory. */
  final DfsPackKey key;

  /**
   * Total number of bytes in this pack file.
   * <p>
   * This field initializes to -1 and gets populated when a block is loaded.
   */
  volatile long length;

  /**
   * Preferred alignment for loading blocks from the backing file.
   * <p>
   * It is initialized to 0 and filled in on the first read made from the
   * file. Block sizes may be odd, e.g. 4091, caused by the underling DFS
   * storing 4091 user bytes and 5 bytes block metadata into a lower level
   * 4096 byte block on disk.
   */
  private volatile int blockSize;

  /** True once corruption has been detected that cannot be worked around. */
  private volatile boolean invalid;

  /**
   * Lock for initialization of {@link #index} and {@link #corruptObjects}.
   * <p>
   * This lock ensures only one thread can perform the initialization work.
   */
  private final Object initLock = new Object();

  /** Index mapping {@link ObjectId} to position within the pack stream. */
  private volatile DfsBlockCache.Ref<PackIndex> index;

  /** Reverse version of {@link #index} mapping position to {@link ObjectId}. */
  private volatile DfsBlockCache.Ref<PackReverseIndex> reverseIndex;

  /**
   * Objects we have tried to read, and discovered to be corrupt.
   * <p>
   * The list is allocated after the first corruption is found, and filled in
   * as more entries are discovered. Typically this list is never used, as
   * pack files do not usually contain corrupt objects.
   */
  private volatile LongList corruptObjects;

  /**
   * Construct a reader for an existing, packfile.
   *
   * @param cache
   *            cache that owns the pack data.
   * @param desc
   *            description of the pack within the DFS.
   * @param key
   *            interned key used to identify blocks in the block cache.
   */
  DfsPackFile(DfsBlockCache cache, DfsPackDescription desc, DfsPackKey key) {
    this.cache = cache;
    this.packDesc = desc;
    this.key = key;

    length = desc.getPackSize();
    if (length <= 0)
      length = -1;
  }

  /** @return description that was originally used to configure this pack file. */
  public DfsPackDescription getPackDescription() {
    return packDesc;
  }

  /** @return bytes cached in memory for this pack, excluding the index. */
  public long getCachedSize() {
    return key.cachedSize.get();
  }

  private String getPackName() {
    return packDesc.getPackName();
  }

  void setBlockSize(int newSize) {
    blockSize = newSize;
  }

  void setPackIndex(PackIndex idx) {
    long objCnt = idx.getObjectCount();
    int recSize = Constants.OBJECT_ID_LENGTH + 8;
    int sz = (int) Math.min(objCnt * recSize, Integer.MAX_VALUE);
    index = cache.put(key, POS_INDEX, sz, idx);
  }

  PackIndex getPackIndex(DfsReader ctx) throws IOException {
    return idx(ctx);
  }

  private PackIndex idx(DfsReader ctx) throws IOException {
    DfsBlockCache.Ref<PackIndex> idxref = index;
    if (idxref != null) {
      PackIndex idx = idxref.get();
      if (idx != null)
        return idx;
    }

    if (invalid)
      throw new PackInvalidException(getPackName());

    synchronized (initLock) {
      idxref = index;
      if (idxref != null) {
        PackIndex idx = idxref.get();
        if (idx != null)
          return idx;
      }

      PackIndex idx;
      try {
        ReadableChannel rc = ctx.db.openPackIndex(packDesc);
        try {
          InputStream in = Channels.newInputStream(rc);
          int wantSize = 8192;
          int bs = rc.blockSize();
          if (0 < bs && bs < wantSize)
            bs = (wantSize / bs) * bs;
          else if (bs <= 0)
            bs = wantSize;
          in = new BufferedInputStream(in, bs);
          idx = PackIndex.read(in);
        } finally {
          rc.close();
        }
      } catch (EOFException e) {
        invalid = true;
        IOException e2 = new IOException(MessageFormat.format(
            DfsText.get().shortReadOfIndex, packDesc.getIndexName()));
        e2.initCause(e);
        throw e2;
      } catch (IOException e) {
        invalid = true;
        IOException e2 = new IOException(MessageFormat.format(
            DfsText.get().cannotReadIndex, packDesc.getIndexName()));
        e2.initCause(e);
        throw e2;
      }

      setPackIndex(idx);
      return idx;
    }
  }

  private PackReverseIndex getReverseIdx(DfsReader ctx) throws IOException {
    DfsBlockCache.Ref<PackReverseIndex> revref = reverseIndex;
    if (revref != null) {
      PackReverseIndex revidx = revref.get();
      if (revidx != null)
        return revidx;
    }

    synchronized (initLock) {
      revref = reverseIndex;
      if (revref != null) {
        PackReverseIndex revidx = revref.get();
        if (revidx != null)
          return revidx;
      }

      PackReverseIndex revidx = new PackReverseIndex(idx(ctx));
      reverseIndex = cache.put(key, POS_REVERSE_INDEX,
          packDesc.getReverseIndexSize(), revidx);
      return revidx;
    }
  }

  boolean hasObject(DfsReader ctx, AnyObjectId id) throws IOException {
    final long offset = idx(ctx).findOffset(id);
    return 0 < offset && !isCorrupt(offset);
  }

  /**
   * Get an object from this pack.
   *
   * @param ctx
   *            temporary working space associated with the calling thread.
   * @param id
   *            the object to obtain from the pack. Must not be null.
   * @return the object loader for the requested object if it is contained in
   *         this pack; null if the object was not found.
   * @throws IOException
   *             the pack file or the index could not be read.
   */
  ObjectLoader get(DfsReader ctx, AnyObjectId id)
      throws IOException {
    long offset = idx(ctx).findOffset(id);
    return 0 < offset && !isCorrupt(offset) ? load(ctx, offset) : null;
  }

  long findOffset(DfsReader ctx, AnyObjectId id) throws IOException {
    return idx(ctx).findOffset(id);
  }

  void resolve(DfsReader ctx, Set<ObjectId> matches, AbbreviatedObjectId id,
      int matchLimit) throws IOException {
    idx(ctx).resolve(matches, id, matchLimit);
  }

  /** Release all memory used by this DfsPackFile instance. */
  public void close() {
    cache.remove(this);
    index = null;
    reverseIndex = null;
  }

  /**
   * Obtain the total number of objects available in this pack. This method
   * relies on pack index, giving number of effectively available objects.
   *
   * @param ctx
   *            current reader for the calling thread.
   * @return number of objects in index of this pack, likewise in this pack
   * @throws IOException
   *             the index file cannot be loaded into memory.
   */
  long getObjectCount(DfsReader ctx) throws IOException {
    return idx(ctx).getObjectCount();
  }

  /**
   * Search for object id with the specified start offset in associated pack
   * (reverse) index.
   *
   * @param ctx
   *            current reader for the calling thread.
   * @param offset
   *            start offset of object to find
   * @return object id for this offset, or null if no object was found
   * @throws IOException
   *             the index file cannot be loaded into memory.
   */
  ObjectId findObjectForOffset(DfsReader ctx, long offset) throws IOException {
    return getReverseIdx(ctx).findObject(offset);
  }

  private byte[] decompress(long position, int sz, DfsReader ctx)
      throws IOException, DataFormatException {
    byte[] dstbuf;
    try {
      dstbuf = new byte[sz];
    } catch (OutOfMemoryError noMemory) {
      // The size may be larger than our heap allows, return null to
      // let the caller know allocation isn't possible and it should
      // use the large object streaming approach instead.
      //
      // For example, this can occur when sz is 640 MB, and JRE
      // maximum heap size is only 256 MB. Even if the JRE has
      // 200 MB free, it cannot allocate a 640 MB byte array.
      return null;
    }

    if (ctx.inflate(this, position, dstbuf, false) != sz)
      throw new EOFException(MessageFormat.format(
          JGitText.get().shortCompressedStreamAt,
          Long.valueOf(position)));
    return dstbuf;
  }

  void copyPackAsIs(PackOutputStream out, boolean validate, DfsReader ctx)
      throws IOException {
    // Pin the first window, this ensures the length is accurate.
    ctx.pin(this, 0);
    ctx.copyPackAsIs(this, length, validate, out);
  }

  void copyAsIs(PackOutputStream out, DfsObjectToPack src,
      boolean validate, DfsReader ctx) throws IOException,
      StoredObjectRepresentationNotAvailableException {
    final CRC32 crc1 = validate ? new CRC32() : null;
    final CRC32 crc2 = validate ? new CRC32() : null;
    final byte[] buf = out.getCopyBuffer();

    // Rip apart the header so we can discover the size.
    //
    try {
      readFully(src.offset, buf, 0, 20, ctx);
    } catch (IOException ioError) {
      StoredObjectRepresentationNotAvailableException gone;
      gone = new StoredObjectRepresentationNotAvailableException(src);
      gone.initCause(ioError);
      throw gone;
    }
    int c = buf[0] & 0xff;
    final int typeCode = (c >> 4) & 7;
    long inflatedLength = c & 15;
    int shift = 4;
    int headerCnt = 1;
    while ((c & 0x80) != 0) {
      c = buf[headerCnt++] & 0xff;
      inflatedLength += (c & 0x7f) << shift;
      shift += 7;
    }

    if (typeCode == Constants.OBJ_OFS_DELTA) {
      do {
        c = buf[headerCnt++] & 0xff;
      } while ((c & 128) != 0);
      if (validate) {
        crc1.update(buf, 0, headerCnt);
        crc2.update(buf, 0, headerCnt);
      }
    } else if (typeCode == Constants.OBJ_REF_DELTA) {
      if (validate) {
        crc1.update(buf, 0, headerCnt);
        crc2.update(buf, 0, headerCnt);
      }

      readFully(src.offset + headerCnt, buf, 0, 20, ctx);
      if (validate) {
        crc1.update(buf, 0, 20);
        crc2.update(buf, 0, 20);
      }
      headerCnt += 20;
    } else if (validate) {
      crc1.update(buf, 0, headerCnt);
      crc2.update(buf, 0, headerCnt);
    }

    final long dataOffset = src.offset + headerCnt;
    final long dataLength = src.length;
    final long expectedCRC;
    final DfsBlock quickCopy;

    // Verify the object isn't corrupt before sending. If it is,
    // we report it missing instead.
    //
    try {
      quickCopy = ctx.quickCopy(this, dataOffset, dataLength);

      if (validate && idx(ctx).hasCRC32Support()) {
        // Index has the CRC32 code cached, validate the object.
        //
        expectedCRC = idx(ctx).findCRC32(src);
        if (quickCopy != null) {
          quickCopy.crc32(crc1, dataOffset, (int) dataLength);
        } else {
          long pos = dataOffset;
          long cnt = dataLength;
          while (cnt > 0) {
            final int n = (int) Math.min(cnt, buf.length);
            readFully(pos, buf, 0, n, ctx);
            crc1.update(buf, 0, n);
            pos += n;
            cnt -= n;
          }
        }
        if (crc1.getValue() != expectedCRC) {
          setCorrupt(src.offset);
          throw new CorruptObjectException(MessageFormat.format(
              JGitText.get().objectAtHasBadZlibStream,
              Long.valueOf(src.offset), getPackName()));
        }
      } else if (validate) {
        // We don't have a CRC32 code in the index, so compute it
        // now while inflating the raw data to get zlib to tell us
        // whether or not the data is safe.
        //
        Inflater inf = ctx.inflater();
        byte[] tmp = new byte[1024];
        if (quickCopy != null) {
          quickCopy.check(inf, tmp, dataOffset, (int) dataLength);
        } else {
          long pos = dataOffset;
          long cnt = dataLength;
          while (cnt > 0) {
            final int n = (int) Math.min(cnt, buf.length);
            readFully(pos, buf, 0, n, ctx);
            crc1.update(buf, 0, n);
            inf.setInput(buf, 0, n);
            while (inf.inflate(tmp, 0, tmp.length) > 0)
              continue;
            pos += n;
            cnt -= n;
          }
        }
        if (!inf.finished() || inf.getBytesRead() != dataLength) {
          setCorrupt(src.offset);
          throw new EOFException(MessageFormat.format(
              JGitText.get().shortCompressedStreamAt,
              Long.valueOf(src.offset)));
        }
        expectedCRC = crc1.getValue();
      } else {
        expectedCRC = -1;
      }
    } catch (DataFormatException dataFormat) {
      setCorrupt(src.offset);

      CorruptObjectException corruptObject = new CorruptObjectException(
          MessageFormat.format(
              JGitText.get().objectAtHasBadZlibStream,
              Long.valueOf(src.offset), getPackName()));
      corruptObject.initCause(dataFormat);

      StoredObjectRepresentationNotAvailableException gone;
      gone = new StoredObjectRepresentationNotAvailableException(src);
      gone.initCause(corruptObject);
      throw gone;

    } catch (IOException ioError) {
      StoredObjectRepresentationNotAvailableException gone;
      gone = new StoredObjectRepresentationNotAvailableException(src);
      gone.initCause(ioError);
      throw gone;
    }

    if (quickCopy != null) {
      // The entire object fits into a single byte array window slice,
      // and we have it pinned.  Write this out without copying.
      //
      out.writeHeader(src, inflatedLength);
      quickCopy.write(out, dataOffset, (int) dataLength, null);

    } else if (dataLength <= buf.length) {
      // Tiny optimization: Lots of objects are very small deltas or
      // deflated commits that are likely to fit in the copy buffer.
      //
      if (!validate) {
        long pos = dataOffset;
        long cnt = dataLength;
        while (cnt > 0) {
          final int n = (int) Math.min(cnt, buf.length);
          readFully(pos, buf, 0, n, ctx);
          pos += n;
          cnt -= n;
        }
      }
      out.writeHeader(src, inflatedLength);
      out.write(buf, 0, (int) dataLength);
    } else {
      // Now we are committed to sending the object. As we spool it out,
      // check its CRC32 code to make sure there wasn't corruption between
      // the verification we did above, and us actually outputting it.
      //
      out.writeHeader(src, inflatedLength);
      long pos = dataOffset;
      long cnt = dataLength;
      while (cnt > 0) {
        final int n = (int) Math.min(cnt, buf.length);
        readFully(pos, buf, 0, n, ctx);
        if (validate)
          crc2.update(buf, 0, n);
        out.write(buf, 0, n);
        pos += n;
        cnt -= n;
      }
      if (validate && crc2.getValue() != expectedCRC) {
        throw new CorruptObjectException(MessageFormat.format(
            JGitText.get().objectAtHasBadZlibStream,
            Long.valueOf(src.offset), getPackName()));
      }
    }
  }

  boolean invalid() {
    return invalid;
  }

  void setInvalid() {
    invalid = true;
  }

  private void readFully(long position, byte[] dstbuf, int dstoff, int cnt,
      DfsReader ctx) throws IOException {
    if (ctx.copy(this, position, dstbuf, dstoff, cnt) != cnt)
      throw new EOFException();
  }

  long alignToBlock(long pos) {
    int size = blockSize;
    if (size == 0)
      size = cache.getBlockSize();
    return (pos / size) * size;
  }

  DfsBlock getOrLoadBlock(long pos, DfsReader ctx) throws IOException {
    return cache.getOrLoad(this, pos, ctx);
  }

  DfsBlock readOneBlock(long pos, DfsReader ctx)
      throws IOException {
    if (invalid)
      throw new PackInvalidException(getPackName());

    boolean close = true;
    ReadableChannel rc = ctx.db.openPackFile(packDesc);
    try {
      // If the block alignment is not yet known, discover it. Prefer the
      // larger size from either the cache or the file itself.
      int size = blockSize;
      if (size == 0) {
        size = rc.blockSize();
        if (size <= 0)
          size = cache.getBlockSize();
        else if (size < cache.getBlockSize())
          size = (cache.getBlockSize() / size) * size;
        blockSize = size;
        pos = (pos / size) * size;
      }

      // If the size of the file is not yet known, try to discover it.
      // Channels may choose to return -1 to indicate they don't
      // know the length yet, in this case read up to the size unit
      // given by the caller, then recheck the length.
      long len = length;
      if (len < 0) {
        len = rc.size();
        if (0 <= len)
          length = len;
      }

      if (0 <= len && len < pos + size)
        size = (int) (len - pos);
      if (size <= 0)
        throw new EOFException(MessageFormat.format(
            DfsText.get().shortReadOfBlock, Long.valueOf(pos),
            getPackName(), Long.valueOf(0), Long.valueOf(0)));

      byte[] buf = new byte[size];
      rc.position(pos);
      int cnt = IO.read(rc, buf, 0, size);
      if (cnt != size) {
        if (0 <= len) {
          throw new EOFException(MessageFormat.format(
                DfsText.get().shortReadOfBlock,
                Long.valueOf(pos),
                getPackName(),
                Integer.valueOf(size),
                Integer.valueOf(cnt)));
        }

        // Assume the entire thing was read in a single shot, compact
        // the buffer to only the space required.
        byte[] n = new byte[cnt];
        System.arraycopy(buf, 0, n, 0, n.length);
        buf = n;
      } else if (len < 0) {
        // With no length at the start of the read, the channel should
        // have the length available at the end.
        length = len = rc.size();
      }

      DfsBlock v = new DfsBlock(key, pos, buf);
      if (v.end < len)
        close = !cache.readAhead(rc, key, size, v.end, len, ctx);
      return v;
    } finally {
      if (close)
        rc.close();
    }
  }

  ObjectLoader load(DfsReader ctx, long pos)
      throws IOException {
    try {
      final byte[] ib = ctx.tempId;
      Delta delta = null;
      byte[] data = null;
      int type = Constants.OBJ_BAD;
      boolean cached = false;

      SEARCH: for (;;) {
        readFully(pos, ib, 0, 20, ctx);
        int c = ib[0] & 0xff;
        final int typeCode = (c >> 4) & 7;
        long sz = c & 15;
        int shift = 4;
        int p = 1;
        while ((c & 0x80) != 0) {
          c = ib[p++] & 0xff;
          sz += (c & 0x7f) << shift;
          shift += 7;
        }

        switch (typeCode) {
        case Constants.OBJ_COMMIT:
        case Constants.OBJ_TREE:
        case Constants.OBJ_BLOB:
        case Constants.OBJ_TAG: {
          if (delta != null) {
            data = decompress(pos + p, (int) sz, ctx);
            type = typeCode;
            break SEARCH;
          }

          if (sz < ctx.getStreamFileThreshold()) {
            data = decompress(pos + p, (int) sz, ctx);
            if (data != null)
              return new ObjectLoader.SmallObject(typeCode, data);
          }
          return new LargePackedWholeObject(typeCode, sz, pos, p, this, ctx.db);
        }

        case Constants.OBJ_OFS_DELTA: {
          c = ib[p++] & 0xff;
          long base = c & 127;
          while ((c & 128) != 0) {
            base += 1;
            c = ib[p++] & 0xff;
            base <<= 7;
            base += (c & 127);
          }
          base = pos - base;
          delta = new Delta(delta, pos, (int) sz, p, base);
          if (sz != delta.deltaSize)
            break SEARCH;

          DeltaBaseCache.Entry e = ctx.getDeltaBaseCache().get(key, base);
          if (e != null) {
            type = e.type;
            data = e.data;
            cached = true;
            break SEARCH;
          }
          pos = base;
          continue SEARCH;
        }

        case Constants.OBJ_REF_DELTA: {
          readFully(pos + p, ib, 0, 20, ctx);
          long base = findDeltaBase(ctx, ObjectId.fromRaw(ib));
          delta = new Delta(delta, pos, (int) sz, p + 20, base);
          if (sz != delta.deltaSize)
            break SEARCH;

          DeltaBaseCache.Entry e = ctx.getDeltaBaseCache().get(key, base);
          if (e != null) {
            type = e.type;
            data = e.data;
            cached = true;
            break SEARCH;
          }
          pos = base;
          continue SEARCH;
        }

        default:
          throw new IOException(MessageFormat.format(
              JGitText.get().unknownObjectType, Integer.valueOf(typeCode)));
        }
      }

      // At this point there is at least one delta to apply to data.
      // (Whole objects with no deltas to apply return early above.)

      if (data == null)
        throw new LargeObjectException();

      do {
        // Cache only the base immediately before desired object.
        if (cached)
          cached = false;
        else if (delta.next == null)
          ctx.getDeltaBaseCache().put(key, delta.basePos, type, data);

        pos = delta.deltaPos;

        byte[] cmds = decompress(pos + delta.hdrLen, delta.deltaSize, ctx);
        if (cmds == null) {
          data = null; // Discard base in case of OutOfMemoryError
          throw new LargeObjectException();
        }

        final long sz = BinaryDelta.getResultSize(cmds);
        if (Integer.MAX_VALUE <= sz)
          throw new LargeObjectException.ExceedsByteArrayLimit();

        final byte[] result;
        try {
          result = new byte[(int) sz];
        } catch (OutOfMemoryError tooBig) {
          data = null; // Discard base in case of OutOfMemoryError
          cmds = null;
          throw new LargeObjectException.OutOfMemory(tooBig);
        }

        BinaryDelta.apply(data, cmds, result);
        data = result;
        delta = delta.next;
      } while (delta != null);

      return new ObjectLoader.SmallObject(type, data);

    } catch (DataFormatException dfe) {
      CorruptObjectException coe = new CorruptObjectException(
          MessageFormat.format(
              JGitText.get().objectAtHasBadZlibStream, Long.valueOf(pos),
              getPackName()));
      coe.initCause(dfe);
      throw coe;
    }
  }

  private long findDeltaBase(DfsReader ctx, ObjectId baseId)
      throws IOException, MissingObjectException {
    long ofs = idx(ctx).findOffset(baseId);
    if (ofs < 0)
      throw new MissingObjectException(baseId,
          JGitText.get().missingDeltaBase);
    return ofs;
  }

  private static class Delta {
    /** Child that applies onto this object. */
    final Delta next;

    /** Offset of the delta object. */
    final long deltaPos;

    /** Size of the inflated delta stream. */
    final int deltaSize;

    /** Total size of the delta's pack entry header (including base). */
    final int hdrLen;

    /** Offset of the base object this delta applies onto. */
    final long basePos;

    Delta(Delta next, long ofs, int sz, int hdrLen, long baseOffset) {
      this.next = next;
      this.deltaPos = ofs;
      this.deltaSize = sz;
      this.hdrLen = hdrLen;
      this.basePos = baseOffset;
    }
  }

  byte[] getDeltaHeader(DfsReader wc, long pos)
      throws IOException, DataFormatException {
    // The delta stream starts as two variable length integers. If we
    // assume they are 64 bits each, we need 16 bytes to encode them,
    // plus 2 extra bytes for the variable length overhead. So 18 is
    // the longest delta instruction header.
    //
    final byte[] hdr = new byte[32];
    wc.inflate(this, pos, hdr, true /* header only */);
    return hdr;
  }

  int getObjectType(DfsReader ctx, long pos) throws IOException {
    final byte[] ib = ctx.tempId;
    for (;;) {
      readFully(pos, ib, 0, 20, ctx);
      int c = ib[0] & 0xff;
      final int type = (c >> 4) & 7;

      switch (type) {
      case Constants.OBJ_COMMIT:
      case Constants.OBJ_TREE:
      case Constants.OBJ_BLOB:
      case Constants.OBJ_TAG:
        return type;

      case Constants.OBJ_OFS_DELTA: {
        int p = 1;
        while ((c & 0x80) != 0)
          c = ib[p++] & 0xff;
        c = ib[p++] & 0xff;
        long ofs = c & 127;
        while ((c & 128) != 0) {
          ofs += 1;
          c = ib[p++] & 0xff;
          ofs <<= 7;
          ofs += (c & 127);
        }
        pos = pos - ofs;
        continue;
      }

      case Constants.OBJ_REF_DELTA: {
        int p = 1;
        while ((c & 0x80) != 0)
          c = ib[p++] & 0xff;
        readFully(pos + p, ib, 0, 20, ctx);
        pos = findDeltaBase(ctx, ObjectId.fromRaw(ib));
        continue;
      }

      default:
        throw new IOException(MessageFormat.format(
            JGitText.get().unknownObjectType, Integer.valueOf(type)));
      }
    }
  }

  long getObjectSize(DfsReader ctx, AnyObjectId id) throws IOException {
    final long offset = idx(ctx).findOffset(id);
    return 0 < offset ? getObjectSize(ctx, offset) : -1;
  }

  long getObjectSize(DfsReader ctx, long pos)
      throws IOException {
    final byte[] ib = ctx.tempId;
    readFully(pos, ib, 0, 20, ctx);
    int c = ib[0] & 0xff;
    final int type = (c >> 4) & 7;
    long sz = c & 15;
    int shift = 4;
    int p = 1;
    while ((c & 0x80) != 0) {
      c = ib[p++] & 0xff;
      sz += (c & 0x7f) << shift;
      shift += 7;
    }

    long deltaAt;
    switch (type) {
    case Constants.OBJ_COMMIT:
    case Constants.OBJ_TREE:
    case Constants.OBJ_BLOB:
    case Constants.OBJ_TAG:
      return sz;

    case Constants.OBJ_OFS_DELTA:
      c = ib[p++] & 0xff;
      while ((c & 128) != 0)
        c = ib[p++] & 0xff;
      deltaAt = pos + p;
      break;

    case Constants.OBJ_REF_DELTA:
      deltaAt = pos + p + 20;
      break;

    default:
      throw new IOException(MessageFormat.format(
          JGitText.get().unknownObjectType, Integer.valueOf(type)));
    }

    try {
      return BinaryDelta.getResultSize(getDeltaHeader(ctx, deltaAt));
    } catch (DataFormatException dfe) {
      CorruptObjectException coe = new CorruptObjectException(
          MessageFormat.format(
              JGitText.get().objectAtHasBadZlibStream, Long.valueOf(pos),
              getPackName()));
      coe.initCause(dfe);
      throw coe;
    }
  }

  void representation(DfsReader ctx, DfsObjectRepresentation r)
      throws IOException {
    final long pos = r.offset;
    final byte[] ib = ctx.tempId;
    readFully(pos, ib, 0, 20, ctx);
    int c = ib[0] & 0xff;
    int p = 1;
    final int typeCode = (c >> 4) & 7;
    while ((c & 0x80) != 0)
      c = ib[p++] & 0xff;

    long len = (getReverseIdx(ctx).findNextOffset(pos, length - 20) - pos);
    switch (typeCode) {
    case Constants.OBJ_COMMIT:
    case Constants.OBJ_TREE:
    case Constants.OBJ_BLOB:
    case Constants.OBJ_TAG:
      r.format = StoredObjectRepresentation.PACK_WHOLE;
      r.length = len - p;
      return;

    case Constants.OBJ_OFS_DELTA: {
      c = ib[p++] & 0xff;
      long ofs = c & 127;
      while ((c & 128) != 0) {
        ofs += 1;
        c = ib[p++] & 0xff;
        ofs <<= 7;
        ofs += (c & 127);
      }
      ofs = pos - ofs;
      r.format = StoredObjectRepresentation.PACK_DELTA;
      r.baseId = findObjectForOffset(ctx, ofs);
      r.length = len - p;
      return;
    }

    case Constants.OBJ_REF_DELTA: {
      len -= p;
      len -= Constants.OBJECT_ID_LENGTH;
      readFully(pos + p, ib, 0, 20, ctx);
      ObjectId id = ObjectId.fromRaw(ib);
      r.format = StoredObjectRepresentation.PACK_DELTA;
      r.baseId = id;
      r.length = len;
      return;
    }

    default:
      throw new IOException(MessageFormat.format(
          JGitText.get().unknownObjectType, Integer.valueOf(typeCode)));
    }
  }

  private boolean isCorrupt(long offset) {
    LongList list = corruptObjects;
    if (list == null)
      return false;
    synchronized (list) {
      return list.contains(offset);
    }
  }

  private void setCorrupt(long offset) {
    LongList list = corruptObjects;
    if (list == null) {
      synchronized (initLock) {
        list = corruptObjects;
        if (list == null) {
          list = new LongList();
          corruptObjects = list;
        }
      }
    }
    synchronized (list) {
      list.add(offset);
    }
  }
}
TOP

Related Classes of org.eclipse.jgit.storage.dfs.DfsPackFile$Delta

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.