/* * Copyright (C) 2007, Dave Watson * Copyright (C) 2008, Robin Rosenberg * Copyright (C) 2008, Roger C. Soares * Copyright (C) 2006, Shawn O. Pearce * Copyright (C) 2010, Chrisian Halstrick 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.dircache; import java.io.File; import java.io.FileOutputStream; import java.io.IOException; import java.io.OutputStream; import java.text.MessageFormat; import java.util.ArrayList; import java.util.HashMap; import java.util.List; import java.util.Map; import org.eclipse.jgit.errors.CheckoutConflictException; import org.eclipse.jgit.errors.CorruptObjectException; import org.eclipse.jgit.errors.IncorrectObjectTypeException; import org.eclipse.jgit.errors.IndexWriteException; import org.eclipse.jgit.errors.MissingObjectException; import org.eclipse.jgit.internal.JGitText; import org.eclipse.jgit.lib.Constants; import org.eclipse.jgit.lib.CoreConfig.AutoCRLF; import org.eclipse.jgit.lib.FileMode; import org.eclipse.jgit.lib.ObjectId; import org.eclipse.jgit.lib.ObjectLoader; import org.eclipse.jgit.lib.ObjectReader; import org.eclipse.jgit.lib.Repository; import org.eclipse.jgit.treewalk.AbstractTreeIterator; import org.eclipse.jgit.treewalk.CanonicalTreeParser; import org.eclipse.jgit.treewalk.EmptyTreeIterator; import org.eclipse.jgit.treewalk.FileTreeIterator; import org.eclipse.jgit.treewalk.NameConflictTreeWalk; import org.eclipse.jgit.treewalk.TreeWalk; import org.eclipse.jgit.treewalk.WorkingTreeIterator; import org.eclipse.jgit.treewalk.WorkingTreeOptions; import org.eclipse.jgit.treewalk.filter.PathFilter; import org.eclipse.jgit.util.FS; import org.eclipse.jgit.util.FileUtils; import org.eclipse.jgit.util.RawParseUtils; import org.eclipse.jgit.util.SystemReader; import org.eclipse.jgit.util.io.AutoCRLFOutputStream; /** * This class handles checking out one or two trees merging with the index. */ public class DirCacheCheckout { private Repository repo; private HashMap updated = new HashMap(); private ArrayList conflicts = new ArrayList(); private ArrayList removed = new ArrayList(); private ObjectId mergeCommitTree; private DirCache dc; private DirCacheBuilder builder; private NameConflictTreeWalk walk; private ObjectId headCommitTree; private WorkingTreeIterator workingTree; private boolean failOnConflict = true; private ArrayList toBeDeleted = new ArrayList(); /** * @return a list of updated paths and objectIds */ public Map getUpdated() { return updated; } /** * @return a list of conflicts created by this checkout */ public List getConflicts() { return conflicts; } /** * @return a list of paths (relative to the start of the working tree) of * files which couldn't be deleted during last call to * {@link #checkout()} . {@link #checkout()} detected that these * files should be deleted but the deletion in the filesystem failed * (e.g. because a file was locked). To have a consistent state of * the working tree these files have to be deleted by the callers of * {@link DirCacheCheckout}. */ public List getToBeDeleted() { return toBeDeleted; } /** * @return a list of all files removed by this checkout */ public List getRemoved() { return removed; } /** * Constructs a DirCacheCeckout for merging and checking out two trees (HEAD * and mergeCommitTree) and the index. * * @param repo * the repository in which we do the checkout * @param headCommitTree * the id of the tree of the head commit * @param dc * the (already locked) Dircache for this repo * @param mergeCommitTree * the id of the tree we want to fast-forward to * @param workingTree * an iterator over the repositories Working Tree * @throws IOException */ public DirCacheCheckout(Repository repo, ObjectId headCommitTree, DirCache dc, ObjectId mergeCommitTree, WorkingTreeIterator workingTree) throws IOException { this.repo = repo; this.dc = dc; this.headCommitTree = headCommitTree; this.mergeCommitTree = mergeCommitTree; this.workingTree = workingTree; } /** * Constructs a DirCacheCeckout for merging and checking out two trees (HEAD * and mergeCommitTree) and the index. As iterator over the working tree * this constructor creates a standard {@link FileTreeIterator} * * @param repo * the repository in which we do the checkout * @param headCommitTree * the id of the tree of the head commit * @param dc * the (already locked) Dircache for this repo * @param mergeCommitTree * the id of the tree we want to fast-forward to * @throws IOException */ public DirCacheCheckout(Repository repo, ObjectId headCommitTree, DirCache dc, ObjectId mergeCommitTree) throws IOException { this(repo, headCommitTree, dc, mergeCommitTree, new FileTreeIterator(repo)); } /** * Constructs a DirCacheCeckout for checking out one tree, merging with the * index. * * @param repo * the repository in which we do the checkout * @param dc * the (already locked) Dircache for this repo * @param mergeCommitTree * the id of the tree we want to fast-forward to * @param workingTree * an iterator over the repositories Working Tree * @throws IOException */ public DirCacheCheckout(Repository repo, DirCache dc, ObjectId mergeCommitTree, WorkingTreeIterator workingTree) throws IOException { this(repo, null, dc, mergeCommitTree, workingTree); } /** * Constructs a DirCacheCeckout for checking out one tree, merging with the * index. As iterator over the working tree this constructor creates a * standard {@link FileTreeIterator} * * @param repo * the repository in which we do the checkout * @param dc * the (already locked) Dircache for this repo * @param mergeCommitTree * the id of the tree of the * @throws IOException */ public DirCacheCheckout(Repository repo, DirCache dc, ObjectId mergeCommitTree) throws IOException { this(repo, null, dc, mergeCommitTree, new FileTreeIterator(repo)); } /** * Scan head, index and merge tree. Used during normal checkout or merge * operations. * * @throws CorruptObjectException * @throws IOException */ public void preScanTwoTrees() throws CorruptObjectException, IOException { removed.clear(); updated.clear(); conflicts.clear(); walk = new NameConflictTreeWalk(repo); builder = dc.builder(); addTree(walk, headCommitTree); addTree(walk, mergeCommitTree); walk.addTree(new DirCacheBuildIterator(builder)); walk.addTree(workingTree); while (walk.next()) { processEntry(walk.getTree(0, CanonicalTreeParser.class), walk.getTree(1, CanonicalTreeParser.class), walk.getTree(2, DirCacheBuildIterator.class), walk.getTree(3, WorkingTreeIterator.class)); if (walk.isSubtree()) walk.enterSubtree(); } } private void addTree(TreeWalk tw, ObjectId id) throws MissingObjectException, IncorrectObjectTypeException, IOException { if (id == null) tw.addTree(new EmptyTreeIterator()); else tw.addTree(id); } /** * Scan index and merge tree (no HEAD). Used e.g. for initial checkout when * there is no head yet. * * @throws MissingObjectException * @throws IncorrectObjectTypeException * @throws CorruptObjectException * @throws IOException */ public void prescanOneTree() throws MissingObjectException, IncorrectObjectTypeException, CorruptObjectException, IOException { removed.clear(); updated.clear(); conflicts.clear(); builder = dc.builder(); walk = new NameConflictTreeWalk(repo); addTree(walk, mergeCommitTree); walk.addTree(new DirCacheBuildIterator(builder)); walk.addTree(workingTree); while (walk.next()) { processEntry(walk.getTree(0, CanonicalTreeParser.class), walk.getTree(1, DirCacheBuildIterator.class), walk.getTree(2, WorkingTreeIterator.class)); if (walk.isSubtree()) walk.enterSubtree(); } conflicts.removeAll(removed); } /** * Processing an entry in the context of {@link #prescanOneTree()} when only * one tree is given * * @param m the tree to merge * @param i the index * @param f the working tree * @throws IOException */ void processEntry(CanonicalTreeParser m, DirCacheBuildIterator i, WorkingTreeIterator f) throws IOException { if (m != null) { checkValidPath(m); // There is an entry in the merge commit. Means: we want to update // what's currently in the index and working-tree to that one if (i == null) { // The index entry is missing if (f != null && !FileMode.TREE.equals(f.getEntryFileMode()) && !f.isEntryIgnored()) { // don't overwrite an untracked and not ignored file conflicts.add(walk.getPathString()); } else update(m.getEntryPathString(), m.getEntryObjectId(), m.getEntryFileMode()); } else if (f == null || !m.idEqual(i)) { // The working tree file is missing or the merge content differs // from index content update(m.getEntryPathString(), m.getEntryObjectId(), m.getEntryFileMode()); } else if (i.getDirCacheEntry() != null) { // The index contains a file (and not a folder) if (f.isModified(i.getDirCacheEntry(), true) || i.getDirCacheEntry().getStage() != 0) // The working tree file is dirty or the index contains a // conflict update(m.getEntryPathString(), m.getEntryObjectId(), m.getEntryFileMode()); else { // update the timestamp of the index with the one from the // file if not set, as we are sure to be in sync here. DirCacheEntry entry = i.getDirCacheEntry(); if (entry.getLastModified() == 0) entry.setLastModified(f.getEntryLastModified()); keep(entry); } } else // The index contains a folder keep(i.getDirCacheEntry()); } else { // There is no entry in the merge commit. Means: we want to delete // what's currently in the index and working tree if (f != null) { // There is a file/folder for that path in the working tree if (walk.isDirectoryFileConflict()) { conflicts.add(walk.getPathString()); } else { // No file/folder conflict exists. All entries are files or // all entries are folders if (i != null) { // ... and the working tree contained a file or folder // -> add it to the removed set and remove it from // conflicts set remove(i.getEntryPathString()); conflicts.remove(i.getEntryPathString()); } else { // untracked file, neither contained in tree to merge // nor in index } } } else { // There is no file/folder for that path in the working tree, // nor in the merge head. // The only entry we have is the index entry. Like the case // where there is a file with the same name, remove it, } } } /** * Execute this checkout * * @return false if this method could not delete all the files * which should be deleted (e.g. because of of the files was * locked). In this case {@link #getToBeDeleted()} lists the files * which should be tried to be deleted outside of this method. * Although false is returned the checkout was * successful and the working tree was updated for all other files. * true is returned when no such problem occurred * * @throws IOException */ public boolean checkout() throws IOException { try { return doCheckout(); } finally { dc.unlock(); } } private boolean doCheckout() throws CorruptObjectException, IOException, MissingObjectException, IncorrectObjectTypeException, CheckoutConflictException, IndexWriteException { toBeDeleted.clear(); ObjectReader objectReader = repo.getObjectDatabase().newReader(); try { if (headCommitTree != null) preScanTwoTrees(); else prescanOneTree(); if (!conflicts.isEmpty()) { if (failOnConflict) throw new CheckoutConflictException(conflicts.toArray(new String[conflicts.size()])); else cleanUpConflicts(); } // update our index builder.finish(); File file = null; String last = null; // when deleting files process them in the opposite order as they have // been reported. This ensures the files are deleted before we delete // their parent folders for (int i = removed.size() - 1; i >= 0; i--) { String r = removed.get(i); file = new File(repo.getWorkTree(), r); if (!file.delete() && file.exists()) { // The list of stuff to delete comes from the index // which will only contain a directory if it is // a submodule, in which case we shall not attempt // to delete it. A submodule is not empty, so it // is safe to check this after a failed delete. if (!file.isDirectory()) toBeDeleted.add(r); } else { if (last != null && !isSamePrefix(r, last)) removeEmptyParents(new File(repo.getWorkTree(), last)); last = r; } } if (file != null) removeEmptyParents(file); for (String path : updated.keySet()) { // ... create/overwrite this file ... file = new File(repo.getWorkTree(), path); if (!file.getParentFile().mkdirs()) { // ignore } DirCacheEntry entry = dc.getEntry(path); // submodules are handled with separate operations if (FileMode.GITLINK.equals(entry.getRawMode())) continue; checkoutEntry(repo, file, entry, objectReader); } // commit the index builder - a new index is persisted if (!builder.commit()) throw new IndexWriteException(); } finally { objectReader.release(); } return toBeDeleted.size() == 0; } private static boolean isSamePrefix(String a, String b) { int as = a.lastIndexOf('/'); int bs = b.lastIndexOf('/'); return a.substring(0, as + 1).equals(b.substring(0, bs + 1)); } private void removeEmptyParents(File f) { File parentFile = f.getParentFile(); while (parentFile != null && !parentFile.equals(repo.getWorkTree())) { if (!parentFile.delete()) break; parentFile = parentFile.getParentFile(); } } /** * Compares whether two pairs of ObjectId and FileMode are equal. * * @param id1 * @param mode1 * @param id2 * @param mode2 * @return true if FileModes and ObjectIds are equal. * false otherwise */ private boolean equalIdAndMode(ObjectId id1, FileMode mode1, ObjectId id2, FileMode mode2) { if (!mode1.equals(mode2)) return false; return id1 != null ? id1.equals(id2) : id2 == null; } /** * Here the main work is done. This method is called for each existing path * in head, index and merge. This method decides what to do with the * corresponding index entry: keep it, update it, remove it or mark a * conflict. * * @param h * the entry for the head * @param m * the entry for the merge * @param i * the entry for the index * @param f * the file in the working tree * @throws IOException */ void processEntry(CanonicalTreeParser h, CanonicalTreeParser m, DirCacheBuildIterator i, WorkingTreeIterator f) throws IOException { DirCacheEntry dce = i != null ? i.getDirCacheEntry() : null; String name = walk.getPathString(); if (m != null) checkValidPath(m); if (i == null && m == null && h == null) { // File/Directory conflict case #20 if (walk.isDirectoryFileConflict()) // TODO: check whether it is always correct to report a conflict here conflict(name, null, null, null); // file only exists in working tree -> ignore it return; } ObjectId iId = (i == null ? null : i.getEntryObjectId()); ObjectId mId = (m == null ? null : m.getEntryObjectId()); ObjectId hId = (h == null ? null : h.getEntryObjectId()); FileMode iMode = (i == null ? null : i.getEntryFileMode()); FileMode mMode = (m == null ? null : m.getEntryFileMode()); FileMode hMode = (h == null ? null : h.getEntryFileMode()); /** *
		 *  File/Directory conflicts:
		 *  the following table from ReadTreeTest tells what to do in case of directory/file
		 *  conflicts. I give comments here
		 *
		 *      H        I       M     Clean     H==M     H==I    I==M         Result
		 *      ------------------------------------------------------------------
		 * 1    D        D       F       Y         N       Y       N           Update
		 * 2    D        D       F       N         N       Y       N           Conflict
		 * 3    D        F       D                 Y       N       N           Keep
		 * 4    D        F       D                 N       N       N           Conflict
		 * 5    D        F       F       Y         N       N       Y           Keep
		 * 5b   D        F       F       Y         N       N       N           Conflict
		 * 6    D        F       F       N         N       N       Y           Keep
		 * 6b   D        F       F       N         N       N       N           Conflict
		 * 7    F        D       F       Y         Y       N       N           Update
		 * 8    F        D       F       N         Y       N       N           Conflict
		 * 9    F        D       F       Y         N       N       N           Update
		 * 10   F        D       D                 N       N       Y           Keep
		 * 11   F        D       D                 N       N       N           Conflict
		 * 12   F        F       D       Y         N       Y       N           Update
		 * 13   F        F       D       N         N       Y       N           Conflict
		 * 14   F        F       D                 N       N       N           Conflict
		 * 15   0        F       D                 N       N       N           Conflict
		 * 16   0        D       F       Y         N       N       N           Update
		 * 17   0        D       F                 N       N       N           Conflict
		 * 18   F        0       D                                             Update
		 * 19   D        0       F                                             Update
		 * 20   0        0       F       N (worktree=dir)                      Conflict
		 * 
*/ // The information whether head,index,merge iterators are currently // pointing to file/folder/non-existing is encoded into this variable. // // To decode write down ffMask in hexadecimal form. The last digit // represents the state for the merge iterator, the second last the // state for the index iterator and the third last represents the state // for the head iterator. The hexadecimal constant "F" stands for // "file", // an "D" stands for "directory" (tree), and a "0" stands for // non-existing // // Examples: // ffMask == 0xFFD -> Head=File, Index=File, Merge=Tree // ffMask == 0xDD0 -> Head=Tree, Index=Tree, Merge=Non-Existing int ffMask = 0; if (h != null) ffMask = FileMode.TREE.equals(hMode) ? 0xD00 : 0xF00; if (i != null) ffMask |= FileMode.TREE.equals(iMode) ? 0x0D0 : 0x0F0; if (m != null) ffMask |= FileMode.TREE.equals(mMode) ? 0x00D : 0x00F; // Check whether we have a possible file/folder conflict. Therefore we // need a least one file and one folder. if (((ffMask & 0x222) != 0x000) && (((ffMask & 0x00F) == 0x00D) || ((ffMask & 0x0F0) == 0x0D0) || ((ffMask & 0xF00) == 0xD00))) { // There are 3*3*3=27 possible combinations of file/folder // conflicts. Some of them are not-relevant because // they represent no conflict, e.g. 0xFFF, 0xDDD, ... The following // switch processes all relevant cases. switch (ffMask) { case 0xDDF: // 1 2 if (isModified(name)) { conflict(name, dce, h, m); // 1 } else { update(name, mId, mMode); // 2 } break; case 0xDFD: // 3 4 keep(dce); break; case 0xF0D: // 18 remove(name); break; case 0xDFF: // 5 5b 6 6b if (equalIdAndMode(iId, iMode, mId, mMode)) keep(dce); // 5 6 else conflict(name, dce, h, m); // 5b 6b case 0xFDD: // 10 11 // TODO: make use of tree extension as soon as available in jgit // we would like to do something like // if (!equalIdAndMode(iId, iMode, mId, mMode) // conflict(name, i.getDirCacheEntry(), h, m); // But since we don't know the id of a tree in the index we do // nothing here and wait that conflicts between index and merge // are found later break; case 0xD0F: // 19 update(name, mId, mMode); break; case 0xDF0: // conflict without a rule case 0x0FD: // 15 conflict(name, dce, h, m); break; case 0xFDF: // 7 8 9 if (equalIdAndMode(hId, hMode, mId, mMode)) { if (isModified(name)) conflict(name, dce, h, m); // 8 else update(name, mId, mMode); // 7 } else if (!isModified(name)) update(name, mId, mMode); // 9 else // To be confirmed - this case is not in the table. conflict(name, dce, h, m); break; case 0xFD0: // keep without a rule keep(dce); break; case 0xFFD: // 12 13 14 if (equalIdAndMode(hId, hMode, iId, iMode)) if (f == null || f.isModified(dce, true)) conflict(name, dce, h, m); else remove(name); else conflict(name, dce, h, m); break; case 0x0DF: // 16 17 if (!isModified(name)) update(name, mId, mMode); else conflict(name, dce, h, m); break; default: keep(dce); } return; } // if we have no file at all then there is nothing to do if ((ffMask & 0x222) == 0) return; if ((ffMask == 0x00F) && f != null && FileMode.TREE.equals(f.getEntryFileMode())) { // File/Directory conflict case #20 conflict(name, null, h, m); } if (i == null) { // Nothing in Index // At least one of Head, Index, Merge is not empty // make sure not to overwrite untracked files if (f != null) { // A submodule is not a file. We should ignore it if (!FileMode.GITLINK.equals(mMode)) { // a dirty worktree: the index is empty but we have a // workingtree-file if (mId == null || !equalIdAndMode(mId, mMode, f.getEntryObjectId(), f.getEntryFileMode())) { conflict(name, null, h, m); return; } } } /** *
			 * 	          I (index)     H        M     H==M  Result
			 * 	        -------------------------------------------
			 * 	        0 nothing    nothing  nothing        (does not happen)
			 * 	        1 nothing    nothing  exists         use M
			 * 	        2 nothing    exists   nothing        remove path from index
			 * 	        3 nothing    exists   exists   yes   keep index
			 * 	          nothing    exists   exists   no    fail
			 * 
*/ if (h == null) // Nothing in Head // Nothing in Index // At least one of Head, Index, Merge is not empty // -> only Merge contains something for this path. Use it! // Potentially update the file update(name, mId, mMode); // 1 else if (m == null) // Nothing in Merge // Something in Head // Nothing in Index // -> only Head contains something for this path and it should // be deleted. Potentially removes the file! remove(name); // 2 else { // 3 // Something in Merge // Something in Head // Nothing in Index // -> Head and Merge contain something (maybe not the same) and // in the index there is nothing (e.g. 'git rm ...' was // called before). Ignore the cached deletion and use what we // find in Merge. Potentially updates the file. if (equalIdAndMode(hId, hMode, mId, mMode)) keep(dce); else conflict(name, dce, h, m); } } else { // Something in Index if (h == null) { // Nothing in Head // Something in Index /** *
				 * 	          clean I==H  I==M       H        M        Result
				 * 	         -----------------------------------------------------
				 * 	        4 yes   N/A   N/A     nothing  nothing  keep index
				 * 	        5 no    N/A   N/A     nothing  nothing  keep index
				 *
				 * 	        6 yes   N/A   yes     nothing  exists   keep index
				 * 	        7 no    N/A   yes     nothing  exists   keep index
				 * 	        8 yes   N/A   no      nothing  exists   fail
				 * 	        9 no    N/A   no      nothing  exists   fail
				 * 
*/ if (m == null || equalIdAndMode(mId, mMode, iId, iMode)) { // Merge contains nothing or the same as Index // Nothing in Head // Something in Index if (m==null && walk.isDirectoryFileConflict()) { // Nothing in Merge and current path is part of // File/Folder conflict // Nothing in Head // Something in Index if (dce != null && (f == null || f.isModified(dce, true))) // No file or file is dirty // Nothing in Merge and current path is part of // File/Folder conflict // Nothing in Head // Something in Index // -> File folder conflict and Merge wants this // path to be removed. Since the file is dirty // report a conflict conflict(name, dce, h, m); else // A file is present and file is not dirty // Nothing in Merge and current path is part of // File/Folder conflict // Nothing in Head // Something in Index // -> File folder conflict and Merge wants this path // to be removed. Since the file is not dirty remove // file and index entry remove(name); } else // Something in Merge or current path is not part of // File/Folder conflict // Merge contains nothing or the same as Index // Nothing in Head // Something in Index // -> Merge contains nothing new. Keep the index. keep(dce); } else // Merge contains something and it is not the same as Index // Nothing in Head // Something in Index // -> Index contains something new (different from Head) // and Merge is different from Index. Report a conflict conflict(name, dce, h, m); } else if (m == null) { // Nothing in Merge // Something in Head // Something in Index /** *
				 * 	           clean I==H  I==M       H        M        Result
				 * 	         -----------------------------------------------------
				 * 	        10 yes   yes   N/A     exists   nothing  remove path from index
				 * 	        11 no    yes   N/A     exists   nothing  fail
				 * 	        12 yes   no    N/A     exists   nothing  fail
				 * 	        13 no    no    N/A     exists   nothing  fail
				 * 
*/ if (iMode == FileMode.GITLINK) { // A submodule in Index // Nothing in Merge // Something in Head // Submodules that disappear from the checkout must // be removed from the index, but not deleted from disk. remove(name); } else { // Something different from a submodule in Index // Nothing in Merge // Something in Head if (equalIdAndMode(hId, hMode, iId, iMode)) { // Index contains the same as Head // Something different from a submodule in Index // Nothing in Merge // Something in Head if (f == null || f.isModified(dce, true)) // file is dirty // Index contains the same as Head // Something different from a submodule in Index // Nothing in Merge // Something in Head // -> file is dirty but is should be removed. That's // a conflict conflict(name, dce, h, m); else // file doesn't exist or is clean // Index contains the same as Head // Something different from a submodule in Index // Nothing in Merge // Something in Head // -> Remove from index and delete the file remove(name); } else // Index contains something different from Head // Something different from a submodule in Index // Nothing in Merge // Something in Head // -> Something new is in index (and maybe even on the // filesystem). But Merge wants the path to be removed. // Report a conflict conflict(name, dce, h, m); } } else { // Something in Merge // Something in Head // Something in Index if (!equalIdAndMode(hId, hMode, mId, mMode) && !equalIdAndMode(hId, hMode, iId, iMode) && !equalIdAndMode(mId, mMode, iId, iMode)) // All three contents in Head, Merge, Index differ from each // other // -> All contents differ. Report a conflict. conflict(name, dce, h, m); else // At least two of the contents of Head, Index, Merge // are the same // Something in Merge // Something in Head // Something in Index if (equalIdAndMode(hId, hMode, iId, iMode) && !equalIdAndMode(mId, mMode, iId, iMode)) { // Head contains the same as Index. Merge differs // Something in Merge // For submodules just update the index with the new SHA-1 if (dce != null && FileMode.GITLINK.equals(dce.getFileMode())) { // Index and Head contain the same submodule. Merge // differs // Something in Merge // -> Nothing new in index. Move to merge. // Potentially updates the file // TODO check that we don't overwrite some unsaved // file content update(name, mId, mMode); } else if (dce != null && (f == null || f.isModified(dce, true))) { // File doesn't exist or is dirty // Head and Index don't contain a submodule // Head contains the same as Index. Merge differs // Something in Merge // -> Merge wants the index and file to be updated // but the file is dirty. Report a conflict conflict(name, dce, h, m); } else { // File exists and is clean // Head and Index don't contain a submodule // Head contains the same as Index. Merge differs // Something in Merge // -> Standard case when switching between branches: // Nothing new in index but something different in // Merge. Update index and file update(name, mId, mMode); } } else { // Head differs from index or merge is same as index // At least two of the contents of Head, Index, Merge // are the same // Something in Merge // Something in Head // Something in Index // Can be formulated as: Either all three states are // equal or Merge is equal to Head or Index and differs // to the other one. // -> In all three cases we don't touch index and file. keep(dce); } } } } /** * A conflict is detected - add the three different stages to the index * @param path the path of the conflicting entry * @param e the previous index entry * @param h the first tree you want to merge (the HEAD) * @param m the second tree you want to merge */ private void conflict(String path, DirCacheEntry e, AbstractTreeIterator h, AbstractTreeIterator m) { conflicts.add(path); DirCacheEntry entry; if (e != null) { entry = new DirCacheEntry(e.getPathString(), DirCacheEntry.STAGE_1); entry.copyMetaData(e, true); builder.add(entry); } if (h != null && !FileMode.TREE.equals(h.getEntryFileMode())) { entry = new DirCacheEntry(h.getEntryPathString(), DirCacheEntry.STAGE_2); entry.setFileMode(h.getEntryFileMode()); entry.setObjectId(h.getEntryObjectId()); builder.add(entry); } if (m != null && !FileMode.TREE.equals(m.getEntryFileMode())) { entry = new DirCacheEntry(m.getEntryPathString(), DirCacheEntry.STAGE_3); entry.setFileMode(m.getEntryFileMode()); entry.setObjectId(m.getEntryObjectId()); builder.add(entry); } } private void keep(DirCacheEntry e) { if (e != null && !FileMode.TREE.equals(e.getFileMode())) builder.add(e); } private void remove(String path) { removed.add(path); } private void update(String path, ObjectId mId, FileMode mode) { if (!FileMode.TREE.equals(mode)) { updated.put(path, mId); DirCacheEntry entry = new DirCacheEntry(path, DirCacheEntry.STAGE_0); entry.setObjectId(mId); entry.setFileMode(mode); builder.add(entry); } } /** * If true, will scan first to see if it's possible to check * out, otherwise throw {@link CheckoutConflictException}. If * false, it will silently deal with the problem. * * @param failOnConflict */ public void setFailOnConflict(boolean failOnConflict) { this.failOnConflict = failOnConflict; } /** * This method implements how to handle conflicts when * {@link #failOnConflict} is false * * @throws CheckoutConflictException */ private void cleanUpConflicts() throws CheckoutConflictException { // TODO: couldn't we delete unsaved worktree content here? for (String c : conflicts) { File conflict = new File(repo.getWorkTree(), c); if (!conflict.delete()) throw new CheckoutConflictException(MessageFormat.format( JGitText.get().cannotDeleteFile, c)); removeEmptyParents(conflict); } for (String r : removed) { File file = new File(repo.getWorkTree(), r); if (!file.delete()) throw new CheckoutConflictException( MessageFormat.format(JGitText.get().cannotDeleteFile, file.getAbsolutePath())); removeEmptyParents(file); } } private boolean isModified(String path) throws CorruptObjectException, IOException { NameConflictTreeWalk tw = new NameConflictTreeWalk(repo); tw.addTree(new DirCacheIterator(dc)); tw.addTree(new FileTreeIterator(repo)); tw.setRecursive(true); tw.setFilter(PathFilter.create(path)); DirCacheIterator dcIt; WorkingTreeIterator wtIt; while(tw.next()) { dcIt = tw.getTree(0, DirCacheIterator.class); wtIt = tw.getTree(1, WorkingTreeIterator.class); if (dcIt == null || wtIt == null) return true; if (wtIt.isModified(dcIt.getDirCacheEntry(), true)) { return true; } } return false; } /** * Updates the file in the working tree with content and mode from an entry * in the index. The new content is first written to a new temporary file in * the same directory as the real file. Then that new file is renamed to the * final filename. Use this method only for checkout of a single entry. * Otherwise use * {@code checkoutEntry(Repository, File f, DirCacheEntry, ObjectReader)} * instead which allows to reuse one {@code ObjectReader} for multiple * entries. * *

* TODO: this method works directly on File IO, we may need another * abstraction (like WorkingTreeIterator). This way we could tell e.g. * Eclipse that Files in the workspace got changed *

* * @param repository * @param f * the file to be modified. The parent directory for this file * has to exist already * @param entry * the entry containing new mode and content * @throws IOException */ public static void checkoutEntry(final Repository repository, File f, DirCacheEntry entry) throws IOException { ObjectReader or = repository.newObjectReader(); try { checkoutEntry(repository, f, entry, repository.newObjectReader()); } finally { or.release(); } } /** * Updates the file in the working tree with content and mode from an entry * in the index. The new content is first written to a new temporary file in * the same directory as the real file. Then that new file is renamed to the * final filename. * *

* TODO: this method works directly on File IO, we may need another * abstraction (like WorkingTreeIterator). This way we could tell e.g. * Eclipse that Files in the workspace got changed *

* * @param repo * @param f * the file to be modified. The parent directory for this file * has to exist already * @param entry * the entry containing new mode and content * @param or * object reader to use for checkout * @throws IOException */ public static void checkoutEntry(final Repository repo, File f, DirCacheEntry entry, ObjectReader or) throws IOException { ObjectLoader ol = or.open(entry.getObjectId()); File parentDir = f.getParentFile(); parentDir.mkdirs(); File tmpFile = File.createTempFile("._" + f.getName(), null, parentDir); //$NON-NLS-1$ WorkingTreeOptions opt = repo.getConfig().get(WorkingTreeOptions.KEY); FileOutputStream rawChannel = new FileOutputStream(tmpFile); OutputStream channel; if (opt.getAutoCRLF() == AutoCRLF.TRUE) channel = new AutoCRLFOutputStream(rawChannel); else channel = rawChannel; try { ol.copyTo(channel); } finally { channel.close(); } FS fs = repo.getFS(); if (opt.isFileMode() && fs.supportsExecute()) { if (FileMode.EXECUTABLE_FILE.equals(entry.getRawMode())) { if (!fs.canExecute(tmpFile)) fs.setExecute(tmpFile, true); } else { if (fs.canExecute(tmpFile)) fs.setExecute(tmpFile, false); } } try { FileUtils.rename(tmpFile, f); } catch (IOException e) { throw new IOException(MessageFormat.format( JGitText.get().couldNotWriteFile, tmpFile.getPath(), f.getPath())); } entry.setLastModified(f.lastModified()); if (opt.getAutoCRLF() != AutoCRLF.FALSE) entry.setLength(f.length()); // AutoCRLF wants on-disk-size else entry.setLength((int) ol.getSize()); } private static byte[][] forbidden; 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]); } private static void checkValidPath(CanonicalTreeParser t) throws InvalidPathException { for (CanonicalTreeParser i = t; i != null; i = i.getParent()) checkValidPathSegment(i); } private static void checkValidPathSegment(CanonicalTreeParser t) throws InvalidPathException { boolean isWindows = SystemReader.getInstance().isWindows(); boolean isOSX = SystemReader.getInstance().isMacOS(); boolean ignCase = isOSX || isWindows; int ptr = t.getNameOffset(); byte[] raw = t.getEntryPathBuffer(); int end = ptr + t.getNameLength(); // Validate path component at this level of the tree int start = ptr; while (ptr < end) { if (raw[ptr] == '/') throw new InvalidPathException( JGitText.get().invalidPathContainsSeparator, "/", t.getEntryPathString()); //$NON-NLS-1$ if (isWindows) { if (raw[ptr] == '\\') throw new InvalidPathException( JGitText.get().invalidPathContainsSeparator, "\\", t.getEntryPathString()); //$NON-NLS-1$ if (raw[ptr] == ':') throw new InvalidPathException( JGitText.get().invalidPathContainsSeparator, ":", t.getEntryPathString()); //$NON-NLS-1$ } ptr++; } // '.' and '..' are invalid here if (ptr - start == 1) { if (raw[start] == '.') throw new InvalidPathException(t.getEntryPathString()); } else if (ptr - start == 2) { if (raw[start] == '.') if (raw[start + 1] == '.') throw new InvalidPathException(t.getEntryPathString()); } else if (ptr - start == 4) { // .git (possibly case insensitive) is disallowed if (raw[start] == '.') if (raw[start + 1] == 'g' || (ignCase && raw[start + 1] == 'G')) if (raw[start + 2] == 'i' || (ignCase && raw[start + 2] == 'I')) if (raw[start + 3] == 't' || (ignCase && raw[start + 3] == 'T')) throw new InvalidPathException( t.getEntryPathString()); } if (isWindows) { // Space or period at end of file name is ignored by Windows. // Treat this as a bad path for now. We may want to handle // this as case insensitivity in the future. if (ptr > 0) { if (raw[ptr - 1] == '.') throw new InvalidPathException( JGitText.get().invalidPathPeriodAtEndWindows, t.getEntryPathString()); if (raw[ptr - 1] == ' ') throw new InvalidPathException( JGitText.get().invalidPathSpaceAtEndWindows, t.getEntryPathString()); } int i; // Bad names, eliminate suffix first for (i = start; i < ptr; ++i) if (raw[i] == '.') break; int len = i - start; if (len == 3 || len == 4) { for (int j = 0; j < forbidden.length; ++j) { if (forbidden[j].length == len) { if (toUpper(raw[start]) < forbidden[j][0]) break; int k; for (k = 0; k < len; ++k) { if (toUpper(raw[start + k]) != forbidden[j][k]) break; } if (k == len) throw new InvalidPathException( JGitText.get().invalidPathReservedOnWindows, RawParseUtils.decode(forbidden[j]), t .getEntryPathString()); } } } } } private static byte toUpper(byte b) { if (b >= 'a' && b <= 'z') return (byte) (b - ('a' - 'A')); return b; } }