@Test
public void testEqualsWithWhitespace() {
final RawText a =
new RawText(Constants.encodeASCII("foo-a\n \n a b c\na \n foo\na b c\n"));
final RawText b =
new RawText(Constants.encodeASCII("foo-a b\n\nab c\na\nfoo\na b c \n"));
// "foo-a" != "foo-a b"
assertFalse(cmp.equals(a, 0, b, 0));
assertFalse(cmp.equals(b, 0, a, 0));
// " " == ""
assertTrue(cmp.equals(a, 1, b, 1));
assertTrue(cmp.equals(b, 1, a, 1));
// " a b c" != "ab c"
assertFalse(cmp.equals(a, 2, b, 2));
assertFalse(cmp.equals(b, 2, a, 2));
// "a " == "a"
assertTrue(cmp.equals(a, 3, b, 3));
assertTrue(cmp.equals(b, 3, a, 3));
// " foo" != "foo"
assertFalse(cmp.equals(a, 4, b, 4));
assertFalse(cmp.equals(b, 4, a, 4));
// "a b c" == "a b c "
assertTrue(cmp.equals(a, 5, b, 5));
assertTrue(cmp.equals(b, 5, a, 5));
}
private static void assertHash(RevObject id, byte[] bin) {
MessageDigest md = Constants.newMessageDigest();
md.update(Constants.encodedTypeString(id.getType()));
md.update((byte) ' ');
md.update(Constants.encodeASCII(bin.length));
md.update((byte) 0);
md.update(bin);
assertEquals(id, ObjectId.fromRaw(md.digest()));
}
void writeTo(final OutputStream os) throws IOException {
final MessageDigest foot = Constants.newMessageDigest();
final DigestOutputStream dos = new DigestOutputStream(os, foot);
boolean extended = false;
for (int i = 0; i < entryCnt; i++) extended |= sortedEntries[i].isExtended();
// Write the header.
//
final byte[] tmp = new byte[128];
System.arraycopy(SIG_DIRC, 0, tmp, 0, SIG_DIRC.length);
NB.encodeInt32(tmp, 4, extended ? 3 : 2);
NB.encodeInt32(tmp, 8, entryCnt);
dos.write(tmp, 0, 12);
// Write the individual file entries.
final int smudge_s;
final int smudge_ns;
if (myLock != null) {
// For new files we need to smudge the index entry
// if they have been modified "now". Ideally we'd
// want the timestamp when we're done writing the index,
// so we use the current timestamp as a approximation.
myLock.createCommitSnapshot();
snapshot = myLock.getCommitSnapshot();
smudge_s = (int) (snapshot.lastModified() / 1000);
smudge_ns = ((int) (snapshot.lastModified() % 1000)) * 1000000;
} else {
// Used in unit tests only
smudge_ns = 0;
smudge_s = 0;
}
// Check if tree is non-null here since calling updateSmudgedEntries
// will automatically build it via creating a DirCacheIterator
final boolean writeTree = tree != null;
if (repository != null && entryCnt > 0) updateSmudgedEntries();
for (int i = 0; i < entryCnt; i++) {
final DirCacheEntry e = sortedEntries[i];
if (e.mightBeRacilyClean(smudge_s, smudge_ns)) e.smudgeRacilyClean();
e.write(dos);
}
if (writeTree) {
final TemporaryBuffer bb = new TemporaryBuffer.LocalFile();
tree.write(tmp, bb);
bb.close();
NB.encodeInt32(tmp, 0, EXT_TREE);
NB.encodeInt32(tmp, 4, (int) bb.length());
dos.write(tmp, 0, 8);
bb.writeTo(dos, null);
}
writeIndexChecksum = foot.digest();
os.write(writeIndexChecksum);
os.close();
}
/**
* Save the configuration as a Git text style configuration file.
*
* <p><b>Warning:</b> Although this method uses the traditional Git file locking approach to
* protect against concurrent writes of the configuration file, it does not ensure that the file
* has not been modified since the last read, which means updates performed by other objects
* accessing the same backing file may be lost.
*
* @throws IOException the file could not be written.
*/
public void save() throws IOException {
final byte[] out;
final String text = toText();
if (utf8Bom) {
final ByteArrayOutputStream bos = new ByteArrayOutputStream();
bos.write(0xEF);
bos.write(0xBB);
bos.write(0xBF);
bos.write(text.getBytes(RawParseUtils.UTF8_CHARSET.name()));
out = bos.toByteArray();
} else {
out = Constants.encode(text);
}
final LockFile lf = new LockFile(getFile(), fs);
if (!lf.lock()) throw new LockFailedException(getFile());
try {
lf.setNeedSnapshot(true);
lf.write(out);
if (!lf.commit())
throw new IOException(MessageFormat.format(JGitText.get().cannotCommitWriteTo, getFile()));
} finally {
lf.unlock();
}
snapshot = lf.getCommitSnapshot();
hash = hash(out);
// notify the listeners
fireConfigChangedEvent();
}
/**
* Update the dumb client server info files.
*
* @throws Exception
*/
public void updateServerInfo() throws Exception {
if (db instanceof FileRepository) {
final FileRepository fr = (FileRepository) db;
RefWriter rw =
new RefWriter(fr.getAllRefs().values()) {
@Override
protected void writeFile(final String name, final byte[] bin) throws IOException {
File path = new File(fr.getDirectory(), name);
TestRepository.this.writeFile(path, bin);
}
};
rw.writePackedRefs();
rw.writeInfoRefs();
final StringBuilder w = new StringBuilder();
for (PackFile p : fr.getObjectDatabase().getPacks()) {
w.append("P ");
w.append(p.getPackFile().getName());
w.append('\n');
}
writeFile(
new File(new File(fr.getObjectDatabase().getDirectory(), "info"), "packs"),
Constants.encodeASCII(w.toString()));
}
}
/**
* Send a message to the client, if it supports receiving them.
*
* <p>If the client doesn't support receiving messages, the message will be discarded, with no
* other indication to the caller or to the client.
*
* @param what string describing the problem identified by the hook. The string must not end with
* an LF, and must not contain an LF.
*/
public void sendMessage(final String what) {
try {
if (msgOut != null) msgOut.write(Constants.encode(what + "\n"));
} catch (IOException e) {
// Ignore write failures.
}
}
/**
* Creates the version 1 pack bitmap index files.
*
* @param dst the output stream to which the index will be written.
*/
public PackBitmapIndexWriterV1(final OutputStream dst) {
out =
new DigestOutputStream(
dst instanceof BufferedOutputStream ? dst : new SafeBufferedOutputStream(dst),
Constants.newMessageDigest());
dataOutput = new SimpleDataOutput(out);
}
public void setProjectDescription(final Project.NameKey name, final String description) {
// Update git's description file, in case gitweb is being used
//
try {
final Repository e;
final LockFile f;
e = openRepository(name);
try {
f = new LockFile(new File(e.getDirectory(), "description"), FS.DETECTED);
if (f.lock()) {
String d = description;
if (d != null) {
d = d.trim();
if (d.length() > 0) {
d += "\n";
}
} else {
d = "";
}
f.write(Constants.encode(d));
f.commit();
}
} finally {
e.close();
}
} catch (RepositoryNotFoundException e) {
log.error("Cannot update description for " + name, e);
} catch (IOException e) {
log.error("Cannot update description for " + name, e);
}
}
/**
* Construct a new substring pattern.
*
* @param patternText text to locate. This should be a literal string, as no meta-characters are
* supported by this implementation. The string may not be the empty string.
*/
public RawSubStringPattern(final String patternText) {
if (patternText.length() == 0)
throw new IllegalArgumentException(JGitText.get().cannotMatchOnEmptyString);
needleString = patternText;
final byte[] b = Constants.encode(patternText);
needle = new byte[b.length];
for (int i = 0; i < b.length; i++) needle[i] = lc(b[i]);
}
private DirCacheEntry makeFile(final String path) throws Exception {
final byte[] pathBytes = Constants.encode(path);
final DirCacheEntry ent = new DirCacheEntry(path);
ent.setFileMode(REGULAR_FILE);
ent.setObjectId(
new ObjectWriter(db)
.computeBlobSha1(pathBytes.length, new ByteArrayInputStream(pathBytes)));
return ent;
}
public FakeTreeIterator(String pathName, FileMode fileMode) {
super(prefix(pathName), new WorkingTreeOptions(AutoCRLF.FALSE));
mode = fileMode.getBits();
final int s = pathName.lastIndexOf('/');
final byte[] name = Constants.encode(pathName.substring(s + 1));
ensurePathCapacity(pathOffset + name.length, pathOffset);
System.arraycopy(name, 0, path, pathOffset, name.length);
pathLen = pathOffset + name.length;
}
public void testEqualsWithoutWhitespace() {
final RawText a = new RawText(cmp, Constants.encodeASCII("foo-a\nfoo-b\nfoo\n"));
final RawText b = new RawText(cmp, Constants.encodeASCII("foo-b\nfoo-c\nf\n"));
assertEquals(3, a.size());
assertEquals(3, b.size());
// foo-a != foo-b
assertFalse(cmp.equals(a, 0, b, 0));
assertFalse(cmp.equals(b, 0, a, 0));
// foo-b == foo-b
assertTrue(cmp.equals(a, 1, b, 0));
assertTrue(cmp.equals(b, 0, a, 1));
// foo != f
assertFalse(cmp.equals(a, 2, b, 2));
assertFalse(cmp.equals(b, 2, a, 2));
}
@Override
public String toString() {
final StringBuilder s = new StringBuilder();
s.append(Constants.typeString(getType()));
s.append(' ');
s.append(name());
s.append(' ');
appendCoreFlags(s);
return s.toString();
}
/**
* Send an error message to the client.
*
* <p>If any error messages are sent before the references are advertised to the client, the
* errors will be sent instead of the advertisement and the receive operation will be aborted. All
* clients should receive and display such early stage errors.
*
* <p>If the reference advertisements have already been sent, messages are sent in a side channel.
* If the client doesn't support receiving messages, the message will be discarded, with no other
* indication to the caller or to the client.
*
* <p>{@link PreReceiveHook}s should always try to use {@link ReceiveCommand#setResult(Result,
* String)} with a result status of {@link Result#REJECTED_OTHER_REASON} to indicate any reasons
* for rejecting an update. Messages attached to a command are much more likely to be returned to
* the client.
*
* @param what string describing the problem identified by the hook. The string must not end with
* an LF, and must not contain an LF.
*/
public void sendError(final String what) {
if (refs == null) {
if (advertiseError == null) advertiseError = new StringBuilder();
advertiseError.append(what).append('\n');
} else {
try {
if (msgOut != null) msgOut.write(Constants.encode("error: " + what + "\n"));
} catch (IOException e) {
// Ignore write failures.
}
}
}
static {
String[] list =
new String[] {
"AUX", "COM1", "COM2", "COM3",
"COM4", //$NON-NLS-1$ //$NON-NLS-2$ //$NON-NLS-3$ //$NON-NLS-4$ //$NON-NLS-5$
"COM5", "COM6", "COM7", "COM8", "COM9", "CON", "LPT1",
"LPT2", //$NON-NLS-1$ //$NON-NLS-2$ //$NON-NLS-3$ //$NON-NLS-4$ //$NON-NLS-5$
// //$NON-NLS-6$ //$NON-NLS-7$ //$NON-NLS-8$
"LPT3", "LPT4", "LPT5", "LPT6", "LPT7", "LPT8", "LPT9",
"NUL", //$NON-NLS-1$ //$NON-NLS-2$ //$NON-NLS-3$ //$NON-NLS-4$ //$NON-NLS-5$ //$NON-NLS-6$
// //$NON-NLS-7$ //$NON-NLS-8$
"PRN"
}; //$NON-NLS-1$
forbidden = new byte[list.length][];
for (int i = 0; i < list.length; ++i) forbidden[i] = Constants.encodeASCII(list[i]);
}
public void start(final Environment env) throws IOException {
Context old = sshScope.set(context);
String message;
try {
message = messageFactory.get().getMessage();
} finally {
sshScope.set(old);
}
err.write(Constants.encode(message.toString()));
err.flush();
in.close();
out.close();
err.close();
exit.onExit(127);
}
void parseCanonical(final RevWalk walk, final byte[] rawTag) throws CorruptObjectException {
final MutableInteger pos = new MutableInteger();
final int oType;
pos.value = 53; // "object $sha1\ntype "
oType = Constants.decodeTypeString(this, rawTag, (byte) '\n', pos);
walk.idBuffer.fromString(rawTag, 7);
object = walk.lookupAny(walk.idBuffer, oType);
int p = pos.value += 4; // "tag "
final int nameEnd = RawParseUtils.nextLF(rawTag, p) - 1;
tagName = RawParseUtils.decode(Constants.CHARSET, rawTag, p, nameEnd);
if (walk.isRetainBody()) buffer = rawTag;
flags |= PARSED;
}
/**
* Recursively get all entries within a subtree.
*
* @param path the subtree path to get all entries within.
* @return all entries recursively contained within the subtree.
*/
public DirCacheEntry[] getEntriesWithin(String path) {
if (path.length() == 0) {
final DirCacheEntry[] r = new DirCacheEntry[sortedEntries.length];
System.arraycopy(sortedEntries, 0, r, 0, sortedEntries.length);
return r;
}
if (!path.endsWith("/")) // $NON-NLS-1$
path += "/"; // $NON-NLS-1$
final byte[] p = Constants.encode(path);
final int pLen = p.length;
int eIdx = findEntry(p, pLen);
if (eIdx < 0) eIdx = -(eIdx + 1);
final int lastIdx = nextEntry(p, pLen, eIdx);
final DirCacheEntry[] r = new DirCacheEntry[lastIdx - eIdx];
System.arraycopy(sortedEntries, eIdx, r, 0, r.length);
return r;
}
private String rand(final HttpServletRequest req, final String suffix)
throws UnsupportedEncodingException {
// Produce a random suffix that is difficult (or nearly impossible)
// for an attacker to guess in advance. This reduces the risk that
// an attacker could upload a *.class file and have us send a ZIP
// that can be invoked through an applet tag in the victim's browser.
//
final MessageDigest md = Constants.newMessageDigest();
final byte[] buf = new byte[8];
NB.encodeInt32(buf, 0, req.getRemotePort());
md.update(req.getRemoteAddr().getBytes("UTF-8"));
md.update(buf, 0, 4);
NB.encodeInt64(buf, 0, System.currentTimeMillis());
md.update(buf, 0, 8);
rng.nextBytes(buf);
md.update(buf, 0, 8);
return suffix + "-" + ObjectId.fromRaw(md.digest()).name();
}
private PackedRefList readPackedRefs() throws IOException {
int maxStaleRetries = 5;
int retries = 0;
while (true) {
final FileSnapshot snapshot = FileSnapshot.save(packedRefsFile);
final BufferedReader br;
final MessageDigest digest = Constants.newMessageDigest();
try {
br =
new BufferedReader(
new InputStreamReader(
new DigestInputStream(new FileInputStream(packedRefsFile), digest), CHARSET));
} catch (FileNotFoundException noPackedRefs) {
if (packedRefsFile.exists()) {
throw noPackedRefs;
}
// Ignore it and leave the new list empty.
return PackedRefList.NO_PACKED_REFS;
}
try {
return new PackedRefList(parsePackedRefs(br), snapshot, ObjectId.fromRaw(digest.digest()));
} catch (IOException e) {
if (FileUtils.isStaleFileHandle(e) && retries < maxStaleRetries) {
if (LOG.isDebugEnabled()) {
LOG.debug(
MessageFormat.format(
JGitText.get().packedRefsHandleIsStale, Integer.valueOf(retries)),
e);
}
retries++;
continue;
}
throw e;
} finally {
br.close();
}
}
}
@Test
public void testRejectInvalidWindowsPaths() throws Exception {
SystemReader.setInstance(
new MockSystemReader() {
{
setUnix();
}
});
String path = "src/con.txt";
DirCache dc = db.lockDirCache();
DirCacheBuilder b = dc.builder();
DirCacheEntry e = new DirCacheEntry(path);
e.setFileMode(FileMode.REGULAR_FILE);
try (ObjectInserter.Formatter formatter = new ObjectInserter.Formatter()) {
e.setObjectId(formatter.idFor(Constants.OBJ_BLOB, Constants.encode(path)));
}
b.add(e);
b.commit();
db.readDirCache();
SystemReader.setInstance(
new MockSystemReader() {
{
setWindows();
}
});
try {
db.readDirCache();
fail("should have rejected " + path);
} catch (CorruptObjectException err) {
assertEquals(MessageFormat.format(JGitText.get().invalidPath, path), err.getMessage());
assertNotNull(err.getCause());
assertEquals("invalid name 'CON'", err.getCause().getMessage());
}
}
/**
* Locate the position a path's entry is at in the index.
*
* <p>If there is at least one entry in the index for this path the position of the lowest stage
* is returned. Subsequent stages can be identified by testing consecutive entries until the path
* differs.
*
* <p>If no path matches the entry -(position+1) is returned, where position is the location it
* would have gone within the index.
*
* @param path the path to search for.
* @return if >= 0 then the return value is the position of the entry in the index; pass to {@link
* #getEntry(int)} to obtain the entry information. If < 0 the entry does not exist in the
* index.
*/
public int findEntry(final String path) {
final byte[] p = Constants.encode(path);
return findEntry(p, p.length);
}
private void readFrom(final InputStream inStream) throws IOException, CorruptObjectException {
final BufferedInputStream in = new BufferedInputStream(inStream);
final MessageDigest md = Constants.newMessageDigest();
// Read the index header and verify we understand it.
//
final byte[] hdr = new byte[20];
IO.readFully(in, hdr, 0, 12);
md.update(hdr, 0, 12);
if (!is_DIRC(hdr)) throw new CorruptObjectException(JGitText.get().notADIRCFile);
final int ver = NB.decodeInt32(hdr, 4);
boolean extended = false;
if (ver == 3) extended = true;
else if (ver != 2)
throw new CorruptObjectException(
MessageFormat.format(JGitText.get().unknownDIRCVersion, Integer.valueOf(ver)));
entryCnt = NB.decodeInt32(hdr, 8);
if (entryCnt < 0) throw new CorruptObjectException(JGitText.get().DIRCHasTooManyEntries);
snapshot = FileSnapshot.save(liveFile);
int smudge_s = (int) (snapshot.lastModified() / 1000);
int smudge_ns = ((int) (snapshot.lastModified() % 1000)) * 1000000;
// Load the individual file entries.
//
final int infoLength = DirCacheEntry.getMaximumInfoLength(extended);
final byte[] infos = new byte[infoLength * entryCnt];
sortedEntries = new DirCacheEntry[entryCnt];
final MutableInteger infoAt = new MutableInteger();
for (int i = 0; i < entryCnt; i++)
sortedEntries[i] = new DirCacheEntry(infos, infoAt, in, md, smudge_s, smudge_ns);
// After the file entries are index extensions, and then a footer.
//
for (; ; ) {
in.mark(21);
IO.readFully(in, hdr, 0, 20);
if (in.read() < 0) {
// No extensions present; the file ended where we expected.
//
break;
}
in.reset();
md.update(hdr, 0, 8);
IO.skipFully(in, 8);
long sz = NB.decodeUInt32(hdr, 4);
switch (NB.decodeInt32(hdr, 0)) {
case EXT_TREE:
{
if (Integer.MAX_VALUE < sz) {
throw new CorruptObjectException(
MessageFormat.format(
JGitText.get().DIRCExtensionIsTooLargeAt,
formatExtensionName(hdr),
Long.valueOf(sz)));
}
final byte[] raw = new byte[(int) sz];
IO.readFully(in, raw, 0, raw.length);
md.update(raw, 0, raw.length);
tree = new DirCacheTree(raw, new MutableInteger(), null);
break;
}
default:
if (hdr[0] >= 'A' && hdr[0] <= 'Z') {
// The extension is optional and is here only as
// a performance optimization. Since we do not
// understand it, we can safely skip past it, after
// we include its data in our checksum.
//
skipOptionalExtension(in, md, hdr, sz);
} else {
// The extension is not an optimization and is
// _required_ to understand this index format.
// Since we did not trap it above we must abort.
//
throw new CorruptObjectException(
MessageFormat.format(
JGitText.get().DIRCExtensionNotSupportedByThisVersion,
formatExtensionName(hdr)));
}
}
}
readIndexChecksum = md.digest();
if (!Arrays.equals(readIndexChecksum, hdr)) {
throw new CorruptObjectException(JGitText.get().DIRCChecksumMismatch);
}
}
/**
* Encode a string pattern for faster matching on byte arrays.
*
* <p>Force the characters to our funny UTF-8 only convention that we use on raw buffers. This
* avoids needing to perform character set decodes on the individual commit buffers.
*
* @param patternText original pattern string supplied by the user or the application.
* @return same pattern, but re-encoded to match our funny raw UTF-8 character sequence {@link
* RawCharSequence}.
*/
protected static final String forceToRaw(final String patternText) {
final byte[] b = Constants.encode(patternText);
final StringBuilder needle = new StringBuilder(b.length);
for (int i = 0; i < b.length; i++) needle.append((char) (b[i] & 0xff));
return needle.toString();
}
/**
* Walks a working directory tree as part of a {@link TreeWalk}.
*
* <p>Most applications will want to use the standard implementation of this iterator, {@link
* FileTreeIterator}, as that does all IO through the standard <code>java.io</code> package. Plugins
* for a Java based IDE may however wish to create their own implementations of this class to allow
* traversal of the IDE's project space, as well as benefit from any caching the IDE may have.
*
* @see FileTreeIterator
*/
public abstract class WorkingTreeIterator extends AbstractTreeIterator {
/** An empty entry array, suitable for {@link #init(Entry[])}. */
protected static final Entry[] EOF = {};
/** Size we perform file IO in if we have to read and hash a file. */
static final int BUFFER_SIZE = 2048;
/** Maximum size of files which may be read fully into memory for performance reasons. */
private static final long MAXIMUM_FILE_SIZE_TO_READ_FULLY = 65536;
/** Inherited state of this iterator, describing working tree, etc. */
private final IteratorState state;
/** The {@link #idBuffer()} for the current entry. */
private byte[] contentId;
/** Index within {@link #entries} that {@link #contentId} came from. */
private int contentIdFromPtr;
/** List of entries obtained from the subclass. */
private Entry[] entries;
/** Total number of entries in {@link #entries} that are valid. */
private int entryCnt;
/** Current position within {@link #entries}. */
private int ptr;
/** If there is a .gitignore file present, the parsed rules from it. */
private IgnoreNode ignoreNode;
/**
* Create a new iterator with no parent.
*
* @param options working tree options to be used
*/
protected WorkingTreeIterator(WorkingTreeOptions options) {
super();
state = new IteratorState(options);
}
/**
* Create a new iterator with no parent and a prefix.
*
* <p>The prefix path supplied is inserted in front of all paths generated by this iterator. It is
* intended to be used when an iterator is being created for a subsection of an overall repository
* and needs to be combined with other iterators that are created to run over the entire
* repository namespace.
*
* @param prefix position of this iterator in the repository tree. The value may be null or the
* empty string to indicate the prefix is the root of the repository. A trailing slash ('/')
* is automatically appended if the prefix does not end in '/'.
* @param options working tree options to be used
*/
protected WorkingTreeIterator(final String prefix, WorkingTreeOptions options) {
super(prefix);
state = new IteratorState(options);
}
/**
* Create an iterator for a subtree of an existing iterator.
*
* @param p parent tree iterator.
*/
protected WorkingTreeIterator(final WorkingTreeIterator p) {
super(p);
state = p.state;
}
/**
* Initialize this iterator for the root level of a repository.
*
* <p>This method should only be invoked after calling {@link #init(Entry[])}, and only for the
* root iterator.
*
* @param repo the repository.
*/
protected void initRootIterator(Repository repo) {
Entry entry;
if (ignoreNode instanceof PerDirectoryIgnoreNode)
entry = ((PerDirectoryIgnoreNode) ignoreNode).entry;
else entry = null;
ignoreNode = new RootIgnoreNode(entry, repo);
}
@Override
public boolean hasId() {
if (contentIdFromPtr == ptr) return true;
return (mode & FileMode.TYPE_MASK) == FileMode.TYPE_FILE;
}
@Override
public byte[] idBuffer() {
if (contentIdFromPtr == ptr) return contentId;
switch (mode & FileMode.TYPE_MASK) {
case FileMode.TYPE_FILE:
contentIdFromPtr = ptr;
return contentId = idBufferBlob(entries[ptr]);
case FileMode.TYPE_SYMLINK:
// Java does not support symbolic links, so we should not
// have reached this particular part of the walk code.
//
return zeroid;
case FileMode.TYPE_GITLINK:
// TODO: Support obtaining current HEAD SHA-1 from nested repository
//
return zeroid;
}
return zeroid;
}
private static final byte[] digits = {'0', '1', '2', '3', '4', '5', '6', '7', '8', '9'};
private static final byte[] hblob = Constants.encodedTypeString(Constants.OBJ_BLOB);
private byte[] idBufferBlob(final Entry e) {
try {
final InputStream is = e.openInputStream();
if (is == null) return zeroid;
try {
state.initializeDigestAndReadBuffer();
final long len = e.getLength();
if (!mightNeedCleaning()) return computeHash(is, len);
if (len <= MAXIMUM_FILE_SIZE_TO_READ_FULLY) {
ByteBuffer rawbuf = IO.readWholeStream(is, (int) len);
byte[] raw = rawbuf.array();
int n = rawbuf.limit();
if (!isBinary(raw, n)) {
rawbuf = filterClean(raw, n);
raw = rawbuf.array();
n = rawbuf.limit();
}
return computeHash(new ByteArrayInputStream(raw, 0, n), n);
}
if (isBinary(e)) return computeHash(is, len);
final long canonLen;
final InputStream lenIs = filterClean(e.openInputStream());
try {
canonLen = computeLength(lenIs);
} finally {
safeClose(lenIs);
}
return computeHash(filterClean(is), canonLen);
} finally {
safeClose(is);
}
} catch (IOException err) {
// Can't read the file? Don't report the failure either.
return zeroid;
}
}
private static void safeClose(final InputStream in) {
try {
in.close();
} catch (IOException err2) {
// Suppress any error related to closing an input
// stream. We don't care, we should not have any
// outstanding data to flush or anything like that.
}
}
private boolean mightNeedCleaning() {
switch (getOptions().getAutoCRLF()) {
case FALSE:
default:
return false;
case TRUE:
case INPUT:
return true;
}
}
private boolean isBinary(byte[] content, int sz) {
return RawText.isBinary(content, sz);
}
private boolean isBinary(Entry entry) throws IOException {
InputStream in = entry.openInputStream();
try {
return RawText.isBinary(in);
} finally {
safeClose(in);
}
}
private ByteBuffer filterClean(byte[] src, int n) throws IOException {
InputStream in = new ByteArrayInputStream(src);
return IO.readWholeStream(filterClean(in), n);
}
private InputStream filterClean(InputStream in) {
return new EolCanonicalizingInputStream(in);
}
/**
* Returns the working tree options used by this iterator.
*
* @return working tree options
*/
public WorkingTreeOptions getOptions() {
return state.options;
}
@Override
public int idOffset() {
return 0;
}
@Override
public void reset() {
if (!first()) {
ptr = 0;
if (!eof()) parseEntry();
}
}
@Override
public boolean first() {
return ptr == 0;
}
@Override
public boolean eof() {
return ptr == entryCnt;
}
@Override
public void next(final int delta) throws CorruptObjectException {
ptr += delta;
if (!eof()) parseEntry();
}
@Override
public void back(final int delta) throws CorruptObjectException {
ptr -= delta;
parseEntry();
}
private void parseEntry() {
final Entry e = entries[ptr];
mode = e.getMode().getBits();
final int nameLen = e.encodedNameLen;
ensurePathCapacity(pathOffset + nameLen, pathOffset);
System.arraycopy(e.encodedName, 0, path, pathOffset, nameLen);
pathLen = pathOffset + nameLen;
}
/**
* Get the byte length of this entry.
*
* @return size of this file, in bytes.
*/
public long getEntryLength() {
return current().getLength();
}
/**
* Get the last modified time of this entry.
*
* @return last modified time of this file, in milliseconds since the epoch (Jan 1, 1970 UTC).
*/
public long getEntryLastModified() {
return current().getLastModified();
}
/**
* Obtain an input stream to read the file content.
*
* <p>Efficient implementations are not required. The caller will usually obtain the stream only
* once per entry, if at all.
*
* <p>The input stream should not use buffering if the implementation can avoid it. The caller
* will buffer as necessary to perform efficient block IO operations.
*
* <p>The caller will close the stream once complete.
*
* @return a stream to read from the file.
* @throws IOException the file could not be opened for reading.
*/
public InputStream openEntryStream() throws IOException {
return current().openInputStream();
}
/**
* Determine if the current entry path is ignored by an ignore rule.
*
* @return true if the entry was ignored by an ignore rule file.
* @throws IOException a relevant ignore rule file exists but cannot be read.
*/
public boolean isEntryIgnored() throws IOException {
return isEntryIgnored(pathLen);
}
/**
* Determine if the entry path is ignored by an ignore rule.
*
* @param pLen the length of the path in the path buffer.
* @return true if the entry is ignored by an ignore rule.
* @throws IOException a relevant ignore rule file exists but cannot be read.
*/
protected boolean isEntryIgnored(final int pLen) throws IOException {
IgnoreNode rules = getIgnoreNode();
if (rules != null) {
// The ignore code wants path to start with a '/' if possible.
// If we have the '/' in our path buffer because we are inside
// a subdirectory include it in the range we convert to string.
//
int pOff = pathOffset;
if (0 < pOff) pOff--;
String p = TreeWalk.pathOf(path, pOff, pLen);
switch (rules.isIgnored(p, FileMode.TREE.equals(mode))) {
case IGNORED:
return true;
case NOT_IGNORED:
return false;
case CHECK_PARENT:
break;
}
}
if (parent instanceof WorkingTreeIterator)
return ((WorkingTreeIterator) parent).isEntryIgnored(pLen);
return false;
}
private IgnoreNode getIgnoreNode() throws IOException {
if (ignoreNode instanceof PerDirectoryIgnoreNode)
ignoreNode = ((PerDirectoryIgnoreNode) ignoreNode).load();
return ignoreNode;
}
private static final Comparator<Entry> ENTRY_CMP =
new Comparator<Entry>() {
public int compare(final Entry o1, final Entry o2) {
final byte[] a = o1.encodedName;
final byte[] b = o2.encodedName;
final int aLen = o1.encodedNameLen;
final int bLen = o2.encodedNameLen;
int cPos;
for (cPos = 0; cPos < aLen && cPos < bLen; cPos++) {
final int cmp = (a[cPos] & 0xff) - (b[cPos] & 0xff);
if (cmp != 0) return cmp;
}
if (cPos < aLen) return (a[cPos] & 0xff) - lastPathChar(o2);
if (cPos < bLen) return lastPathChar(o1) - (b[cPos] & 0xff);
return lastPathChar(o1) - lastPathChar(o2);
}
};
static int lastPathChar(final Entry e) {
return e.getMode() == FileMode.TREE ? '/' : '\0';
}
/**
* Constructor helper.
*
* @param list files in the subtree of the work tree this iterator operates on
*/
protected void init(final Entry[] list) {
// Filter out nulls, . and .. as these are not valid tree entries,
// also cache the encoded forms of the path names for efficient use
// later on during sorting and iteration.
//
entries = list;
int i, o;
final CharsetEncoder nameEncoder = state.nameEncoder;
for (i = 0, o = 0; i < entries.length; i++) {
final Entry e = entries[i];
if (e == null) continue;
final String name = e.getName();
if (".".equals(name) || "..".equals(name)) continue;
if (Constants.DOT_GIT.equals(name)) continue;
if (Constants.DOT_GIT_IGNORE.equals(name)) ignoreNode = new PerDirectoryIgnoreNode(e);
if (i != o) entries[o] = e;
e.encodeName(nameEncoder);
o++;
}
entryCnt = o;
Arrays.sort(entries, 0, entryCnt, ENTRY_CMP);
contentIdFromPtr = -1;
ptr = 0;
if (!eof()) parseEntry();
}
/**
* Obtain the current entry from this iterator.
*
* @return the currently selected entry.
*/
protected Entry current() {
return entries[ptr];
}
/**
* Checks whether this entry differs from a given entry from the {@link DirCache}.
*
* <p>File status information is used and if status is same we consider the file identical to the
* state in the working directory. Native git uses more stat fields than we have accessible in
* Java.
*
* @param entry the entry from the dircache we want to compare against
* @param forceContentCheck True if the actual file content should be checked if modification time
* differs.
* @return true if content is most likely different.
*/
public boolean isModified(DirCacheEntry entry, boolean forceContentCheck) {
if (entry.isAssumeValid()) return false;
if (entry.isUpdateNeeded()) return true;
if (!entry.isSmudged() && (getEntryLength() != entry.getLength())) return true;
// Determine difference in mode-bits of file and index-entry. In the
// bitwise presentation of modeDiff we'll have a '1' when the two modes
// differ at this position.
int modeDiff = getEntryRawMode() ^ entry.getRawMode();
// Ignore the executable file bits if checkFilemode tells me to do so.
// Ignoring is done by setting the bits representing a EXECUTABLE_FILE
// to '0' in modeDiff
if (!state.options.isFileMode()) modeDiff &= ~FileMode.EXECUTABLE_FILE.getBits();
if (modeDiff != 0)
// Report a modification if the modes still (after potentially
// ignoring EXECUTABLE_FILE bits) differ
return true;
// Git under windows only stores seconds so we round the timestamp
// Java gives us if it looks like the timestamp in index is seconds
// only. Otherwise we compare the timestamp at millisecond precision.
long cacheLastModified = entry.getLastModified();
long fileLastModified = getEntryLastModified();
if (cacheLastModified % 1000 == 0)
fileLastModified = fileLastModified - fileLastModified % 1000;
if (fileLastModified != cacheLastModified) {
// The file is dirty by timestamps
if (forceContentCheck) {
// But we are told to look at content even though timestamps
// tell us about modification
return contentCheck(entry);
} else {
// We are told to assume a modification if timestamps differs
return true;
}
} else {
// The file is clean when you look at timestamps.
if (entry.isSmudged()) {
// The file is clean by timestamps but the entry was smudged.
// Lets do a content check
return contentCheck(entry);
} else {
// The file is clean by timestamps and the entry is not
// smudged: Can't get any cleaner!
return false;
}
}
}
/**
* Compares the entries content with the content in the filesystem. Unsmudges the entry when it is
* detected that it is clean.
*
* @param entry the entry to be checked
* @return <code>true</code> if the content matches, <code>false</code> otherwise
*/
private boolean contentCheck(DirCacheEntry entry) {
if (getEntryObjectId().equals(entry.getObjectId())) {
// Content has not changed
// We know the entry can't be racily clean because it's still clean.
// Therefore we unsmudge the entry!
// If by any chance we now unsmudge although we are still in the
// same time-slot as the last modification to the index file the
// next index write operation will smudge again.
// Caution: we are unsmudging just by setting the length of the
// in-memory entry object. It's the callers task to detect that we
// have modified the entry and to persist the modified index.
entry.setLength((int) getEntryLength());
return false;
} else {
// Content differs: that's a real change!
return true;
}
}
private long computeLength(InputStream in) throws IOException {
// Since we only care about the length, use skip. The stream
// may be able to more efficiently wade through its data.
//
long length = 0;
for (; ; ) {
long n = in.skip(1 << 20);
if (n <= 0) break;
length += n;
}
return length;
}
private byte[] computeHash(InputStream in, long length) throws IOException {
final MessageDigest contentDigest = state.contentDigest;
final byte[] contentReadBuffer = state.contentReadBuffer;
contentDigest.reset();
contentDigest.update(hblob);
contentDigest.update((byte) ' ');
long sz = length;
if (sz == 0) {
contentDigest.update((byte) '0');
} else {
final int bufn = contentReadBuffer.length;
int p = bufn;
do {
contentReadBuffer[--p] = digits[(int) (sz % 10)];
sz /= 10;
} while (sz > 0);
contentDigest.update(contentReadBuffer, p, bufn - p);
}
contentDigest.update((byte) 0);
for (; ; ) {
final int r = in.read(contentReadBuffer);
if (r <= 0) break;
contentDigest.update(contentReadBuffer, 0, r);
sz += r;
}
if (sz != length) return zeroid;
return contentDigest.digest();
}
/** A single entry within a working directory tree. */
protected abstract static class Entry {
byte[] encodedName;
int encodedNameLen;
void encodeName(final CharsetEncoder enc) {
final ByteBuffer b;
try {
b = enc.encode(CharBuffer.wrap(getName()));
} catch (CharacterCodingException e) {
// This should so never happen.
throw new RuntimeException(
MessageFormat.format(JGitText.get().unencodeableFile, getName()));
}
encodedNameLen = b.limit();
if (b.hasArray() && b.arrayOffset() == 0) encodedName = b.array();
else b.get(encodedName = new byte[encodedNameLen]);
}
public String toString() {
return getMode().toString() + " " + getName();
}
/**
* Get the type of this entry.
*
* <p><b>Note: Efficient implementation required.</b>
*
* <p>The implementation of this method must be efficient. If a subclass needs to compute the
* value they should cache the reference within an instance member instead.
*
* @return a file mode constant from {@link FileMode}.
*/
public abstract FileMode getMode();
/**
* Get the byte length of this entry.
*
* <p><b>Note: Efficient implementation required.</b>
*
* <p>The implementation of this method must be efficient. If a subclass needs to compute the
* value they should cache the reference within an instance member instead.
*
* @return size of this file, in bytes.
*/
public abstract long getLength();
/**
* Get the last modified time of this entry.
*
* <p><b>Note: Efficient implementation required.</b>
*
* <p>The implementation of this method must be efficient. If a subclass needs to compute the
* value they should cache the reference within an instance member instead.
*
* @return time since the epoch (in ms) of the last change.
*/
public abstract long getLastModified();
/**
* Get the name of this entry within its directory.
*
* <p>Efficient implementations are not required. The caller will obtain the name only once and
* cache it once obtained.
*
* @return name of the entry.
*/
public abstract String getName();
/**
* Obtain an input stream to read the file content.
*
* <p>Efficient implementations are not required. The caller will usually obtain the stream only
* once per entry, if at all.
*
* <p>The input stream should not use buffering if the implementation can avoid it. The caller
* will buffer as necessary to perform efficient block IO operations.
*
* <p>The caller will close the stream once complete.
*
* @return a stream to read from the file.
* @throws IOException the file could not be opened for reading.
*/
public abstract InputStream openInputStream() throws IOException;
}
/** Magic type indicating we know rules exist, but they aren't loaded. */
private static class PerDirectoryIgnoreNode extends IgnoreNode {
final Entry entry;
PerDirectoryIgnoreNode(Entry entry) {
super(Collections.<IgnoreRule>emptyList());
this.entry = entry;
}
IgnoreNode load() throws IOException {
IgnoreNode r = new IgnoreNode();
InputStream in = entry.openInputStream();
try {
r.parse(in);
} finally {
in.close();
}
return r.getRules().isEmpty() ? null : r;
}
}
/** Magic type indicating there may be rules for the top level. */
private static class RootIgnoreNode extends PerDirectoryIgnoreNode {
final Repository repository;
RootIgnoreNode(Entry entry, Repository repository) {
super(entry);
this.repository = repository;
}
@Override
IgnoreNode load() throws IOException {
IgnoreNode r;
if (entry != null) {
r = super.load();
if (r == null) r = new IgnoreNode();
} else {
r = new IgnoreNode();
}
File exclude = new File(repository.getDirectory(), "info/exclude");
if (exclude.exists()) {
FileInputStream in = new FileInputStream(exclude);
try {
r.parse(in);
} finally {
in.close();
}
}
return r.getRules().isEmpty() ? null : r;
}
}
private static final class IteratorState {
/** Options used to process the working tree. */
final WorkingTreeOptions options;
/** File name character encoder. */
final CharsetEncoder nameEncoder;
/** Digest computer for {@link #contentId} computations. */
MessageDigest contentDigest;
/** Buffer used to perform {@link #contentId} computations. */
byte[] contentReadBuffer;
IteratorState(WorkingTreeOptions options) {
this.options = options;
this.nameEncoder = Constants.CHARSET.newEncoder();
}
void initializeDigestAndReadBuffer() {
if (contentDigest == null) {
contentDigest = Constants.newMessageDigest();
contentReadBuffer = new byte[BUFFER_SIZE];
}
}
}
}
void initializeDigestAndReadBuffer() {
if (contentDigest == null) {
contentDigest = Constants.newMessageDigest();
contentReadBuffer = new byte[BUFFER_SIZE];
}
}
@Test
public void testFindOnEmpty() throws Exception {
final DirCache dc = DirCache.newInCore();
final byte[] path = Constants.encode("a");
assertEquals(-1, dc.findEntry(path, path.length));
}
private void insert(String text) throws IOException {
insert(Constants.encode(text));
}
private static void digest(TemporaryBuffer.Heap buf) throws IOException {
MessageDigest md = Constants.newMessageDigest();
md.update(buf.toByteArray());
buf.write(md.digest());
}
private static ObjectId hash(final byte[] rawText) {
return ObjectId.fromRaw(Constants.newMessageDigest().digest(rawText));
}