diff options
Diffstat (limited to 'subversion/libsvn_fs_fs/tree.c')
-rw-r--r-- | subversion/libsvn_fs_fs/tree.c | 4420 |
1 files changed, 4420 insertions, 0 deletions
diff --git a/subversion/libsvn_fs_fs/tree.c b/subversion/libsvn_fs_fs/tree.c new file mode 100644 index 0000000..c14955d --- /dev/null +++ b/subversion/libsvn_fs_fs/tree.c @@ -0,0 +1,4420 @@ +/* tree.c : tree-like filesystem, built on DAG filesystem + * + * ==================================================================== + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + * ==================================================================== + */ + + +/* The job of this layer is to take a filesystem with lots of node + sharing going on --- the real DAG filesystem as it appears in the + database --- and make it look and act like an ordinary tree + filesystem, with no sharing. + + We do just-in-time cloning: you can walk from some unfinished + transaction's root down into directories and files shared with + committed revisions; as soon as you try to change something, the + appropriate nodes get cloned (and parent directory entries updated) + invisibly, behind your back. Any other references you have to + nodes that have been cloned by other changes, even made by other + processes, are automatically updated to point to the right clones. */ + + +#include <stdlib.h> +#include <string.h> +#include <assert.h> +#include <apr_pools.h> +#include <apr_hash.h> + +#include "svn_hash.h" +#include "svn_private_config.h" +#include "svn_pools.h" +#include "svn_error.h" +#include "svn_path.h" +#include "svn_mergeinfo.h" +#include "svn_fs.h" +#include "svn_props.h" + +#include "fs.h" +#include "key-gen.h" +#include "dag.h" +#include "lock.h" +#include "tree.h" +#include "fs_fs.h" +#include "id.h" +#include "temp_serializer.h" + +#include "private/svn_mergeinfo_private.h" +#include "private/svn_subr_private.h" +#include "private/svn_fs_util.h" +#include "private/svn_fspath.h" +#include "../libsvn_fs/fs-loader.h" + + +/* ### I believe this constant will become internal to reps-strings.c. + ### see the comment in window_consumer() for more information. */ + +/* ### the comment also seems to need tweaking: the log file stuff + ### is no longer an issue... */ +/* Data written to the filesystem through the svn_fs_apply_textdelta() + interface is cached in memory until the end of the data stream, or + until a size trigger is hit. Define that trigger here (in bytes). + Setting the value to 0 will result in no filesystem buffering at + all. The value only really matters when dealing with file contents + bigger than the value itself. Above that point, large values here + allow the filesystem to buffer more data in memory before flushing + to the database, which increases memory usage but greatly decreases + the amount of disk access (and log-file generation) in database. + Smaller values will limit your overall memory consumption, but can + drastically hurt throughput by necessitating more write operations + to the database (which also generates more log-files). */ +#define WRITE_BUFFER_SIZE 512000 + + + +/* The root structures. + + Why do they contain different data? Well, transactions are mutable + enough that it isn't safe to cache the DAG node for the root + directory or the hash of copyfrom data: somebody else might modify + them concurrently on disk! (Why is the DAG node cache safer than + the root DAG node? When cloning transaction DAG nodes in and out + of the cache, all of the possibly-mutable data from the + node_revision_t inside the dag_node_t is dropped.) Additionally, + revisions are immutable enough that their DAG node cache can be + kept in the FS object and shared among multiple revision root + objects. +*/ +typedef struct fs_rev_root_data_t +{ + /* A dag node for the revision's root directory. */ + dag_node_t *root_dir; + + /* Cache structure for mapping const char * PATH to const char + *COPYFROM_STRING, so that paths_changed can remember all the + copyfrom information in the changes file. + COPYFROM_STRING has the format "REV PATH", or is the empty string if + the path was added without history. */ + apr_hash_t *copyfrom_cache; + +} fs_rev_root_data_t; + +typedef struct fs_txn_root_data_t +{ + const char *txn_id; + + /* Cache of txn DAG nodes (without their nested noderevs, because + * it's mutable). Same keys/values as ffd->rev_node_cache. */ + svn_cache__t *txn_node_cache; +} fs_txn_root_data_t; + +/* Declared here to resolve the circular dependencies. */ +static svn_error_t * get_dag(dag_node_t **dag_node_p, + svn_fs_root_t *root, + const char *path, + svn_boolean_t needs_lock_cache, + apr_pool_t *pool); + +static svn_fs_root_t *make_revision_root(svn_fs_t *fs, svn_revnum_t rev, + dag_node_t *root_dir, + apr_pool_t *pool); + +static svn_error_t *make_txn_root(svn_fs_root_t **root_p, + svn_fs_t *fs, const char *txn, + svn_revnum_t base_rev, apr_uint32_t flags, + apr_pool_t *pool); + + +/*** Node Caching ***/ + +/* 1st level cache */ + +/* An entry in the first-level cache. REVISION and PATH form the key that + will ultimately be matched. + */ +typedef struct cache_entry_t +{ + /* hash value derived from PATH, REVISION. + Used to short-circuit failed lookups. */ + long int hash_value; + + /* revision to which the NODE belongs */ + svn_revnum_t revision; + + /* path of the NODE */ + char *path; + + /* cached value of strlen(PATH). */ + apr_size_t path_len; + + /* the node allocated in the cache's pool. NULL for empty entries. */ + dag_node_t *node; +} cache_entry_t; + +/* Number of entries in the cache. Keep this low to keep pressure on the + CPU caches low as well. A binary value is most efficient. If we walk + a directory tree, we want enough entries to store nodes for all files + without overwriting the nodes for the parent folder. That way, there + will be no unnecessary misses (except for a few random ones caused by + hash collision). + + The actual number of instances may be higher but entries that got + overwritten are no longer visible. + */ +enum { BUCKET_COUNT = 256 }; + +/* Each pool that has received a DAG node, will hold at least on lock on + our cache to ensure that the node remains valid despite being allocated + in the cache's pool. This is the structure to represent the lock. + */ +typedef struct cache_lock_t +{ + /* pool holding the lock */ + apr_pool_t *pool; + + /* cache being locked */ + fs_fs_dag_cache_t *cache; + + /* next lock. NULL at EOL */ + struct cache_lock_t *next; + + /* previous lock. NULL at list head. Only then this==cache->first_lock */ + struct cache_lock_t *prev; +} cache_lock_t; + +/* The actual cache structure. All nodes will be allocated in POOL. + When the number of INSERTIONS (i.e. objects created form that pool) + exceeds a certain threshold, the pool will be cleared and the cache + with it. + + To ensure that nodes returned from this structure remain valid, the + cache will get locked for the lifetime of the _receiving_ pools (i.e. + those in which we would allocate the node if there was no cache.). + The cache will only be cleared FIRST_LOCK is 0. + */ +struct fs_fs_dag_cache_t +{ + /* fixed number of (possibly empty) cache entries */ + cache_entry_t buckets[BUCKET_COUNT]; + + /* pool used for all node allocation */ + apr_pool_t *pool; + + /* number of entries created from POOL since the last cleanup */ + apr_size_t insertions; + + /* Property lookups etc. have a very high locality (75% re-hit). + Thus, remember the last hit location for optimistic lookup. */ + apr_size_t last_hit; + + /* List of receiving pools that are still alive. */ + cache_lock_t *first_lock; +}; + +/* Cleanup function to be called when a receiving pool gets cleared. + Unlocks the cache once. + */ +static apr_status_t +unlock_cache(void *baton_void) +{ + cache_lock_t *lock = baton_void; + + /* remove lock from chain. Update the head */ + if (lock->next) + lock->next->prev = lock->prev; + if (lock->prev) + lock->prev->next = lock->next; + else + lock->cache->first_lock = lock->next; + + return APR_SUCCESS; +} + +/* Cleanup function to be called when the cache itself gets destroyed. + In that case, we must unregister all unlock requests. + */ +static apr_status_t +unregister_locks(void *baton_void) +{ + fs_fs_dag_cache_t *cache = baton_void; + cache_lock_t *lock; + + for (lock = cache->first_lock; lock; lock = lock->next) + apr_pool_cleanup_kill(lock->pool, + lock, + unlock_cache); + + return APR_SUCCESS; +} + +fs_fs_dag_cache_t* +svn_fs_fs__create_dag_cache(apr_pool_t *pool) +{ + fs_fs_dag_cache_t *result = apr_pcalloc(pool, sizeof(*result)); + result->pool = svn_pool_create(pool); + + apr_pool_cleanup_register(pool, + result, + unregister_locks, + apr_pool_cleanup_null); + + return result; +} + +/* Prevent the entries in CACHE from being destroyed, for as long as the + POOL lives. + */ +static void +lock_cache(fs_fs_dag_cache_t* cache, apr_pool_t *pool) +{ + /* we only need to lock / unlock once per pool. Since we will often ask + for multiple nodes with the same pool, we can reduce the overhead. + However, if e.g. pools are being used in an alternating pattern, + we may lock the cache more than once for the same pool (and register + just as many cleanup actions). + */ + cache_lock_t *lock = cache->first_lock; + + /* try to find an existing lock for POOL. + But limit the time spent on chasing pointers. */ + int limiter = 8; + while (lock && --limiter) + if (lock->pool == pool) + return; + + /* create a new lock and put it at the beginning of the lock chain */ + lock = apr_palloc(pool, sizeof(*lock)); + lock->cache = cache; + lock->pool = pool; + lock->next = cache->first_lock; + lock->prev = NULL; + + if (cache->first_lock) + cache->first_lock->prev = lock; + cache->first_lock = lock; + + /* instruct POOL to remove the look upon cleanup */ + apr_pool_cleanup_register(pool, + lock, + unlock_cache, + apr_pool_cleanup_null); +} + +/* Clears the CACHE at regular intervals (destroying all cached nodes) + */ +static void +auto_clear_dag_cache(fs_fs_dag_cache_t* cache) +{ + if (cache->first_lock == NULL && cache->insertions > BUCKET_COUNT) + { + svn_pool_clear(cache->pool); + + memset(cache->buckets, 0, sizeof(cache->buckets)); + cache->insertions = 0; + } +} + +/* For the given REVISION and PATH, return the respective entry in CACHE. + If the entry is empty, its NODE member will be NULL and the caller + may then set it to the corresponding DAG node allocated in CACHE->POOL. + */ +static cache_entry_t * +cache_lookup( fs_fs_dag_cache_t *cache + , svn_revnum_t revision + , const char *path) +{ + apr_size_t i, bucket_index; + apr_size_t path_len = strlen(path); + long int hash_value = revision; + + /* optimistic lookup: hit the same bucket again? */ + cache_entry_t *result = &cache->buckets[cache->last_hit]; + if ( (result->revision == revision) + && (result->path_len == path_len) + && !memcmp(result->path, path, path_len)) + { + return result; + } + + /* need to do a full lookup. Calculate the hash value + (HASH_VALUE has been initialized to REVISION). */ + for (i = 0; i + 4 <= path_len; i += 4) + hash_value = hash_value * 0xd1f3da69 + *(const apr_uint32_t*)(path + i); + + for (; i < path_len; ++i) + hash_value = hash_value * 33 + path[i]; + + bucket_index = hash_value + (hash_value >> 16); + bucket_index = (bucket_index + (bucket_index >> 8)) % BUCKET_COUNT; + + /* access the corresponding bucket and remember its location */ + result = &cache->buckets[bucket_index]; + cache->last_hit = bucket_index; + + /* if it is *NOT* a match, clear the bucket, expect the caller to fill + in the node and count it as an insertion */ + if ( (result->hash_value != hash_value) + || (result->revision != revision) + || (result->path_len != path_len) + || memcmp(result->path, path, path_len)) + { + result->hash_value = hash_value; + result->revision = revision; + if (result->path_len < path_len) + result->path = apr_palloc(cache->pool, path_len + 1); + result->path_len = path_len; + memcpy(result->path, path, path_len + 1); + + result->node = NULL; + + cache->insertions++; + } + + return result; +} + +/* 2nd level cache */ + +/* Find and return the DAG node cache for ROOT and the key that + should be used for PATH. */ +static void +locate_cache(svn_cache__t **cache, + const char **key, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + if (root->is_txn_root) + { + fs_txn_root_data_t *frd = root->fsap_data; + if (cache) *cache = frd->txn_node_cache; + if (key && path) *key = path; + } + else + { + fs_fs_data_t *ffd = root->fs->fsap_data; + if (cache) *cache = ffd->rev_node_cache; + if (key && path) *key + = svn_fs_fs__combine_number_and_string(root->rev, path, pool); + } +} + +/* Return NODE for PATH from ROOT's node cache, or NULL if the node + isn't cached; read it from the FS. *NODE remains valid until either + POOL or the FS gets cleared or destroyed (whichever comes first). + + Since locking can be expensive and POOL may be long-living, for + nodes that will not need to survive the next call to this function, + set NEEDS_LOCK_CACHE to FALSE. */ +static svn_error_t * +dag_node_cache_get(dag_node_t **node_p, + svn_fs_root_t *root, + const char *path, + svn_boolean_t needs_lock_cache, + apr_pool_t *pool) +{ + svn_boolean_t found; + dag_node_t *node = NULL; + svn_cache__t *cache; + const char *key; + + SVN_ERR_ASSERT(*path == '/'); + + if (!root->is_txn_root) + { + /* immutable DAG node. use the global caches for it */ + + fs_fs_data_t *ffd = root->fs->fsap_data; + cache_entry_t *bucket; + + auto_clear_dag_cache(ffd->dag_node_cache); + bucket = cache_lookup(ffd->dag_node_cache, root->rev, path); + if (bucket->node == NULL) + { + locate_cache(&cache, &key, root, path, pool); + SVN_ERR(svn_cache__get((void **)&node, &found, cache, key, + ffd->dag_node_cache->pool)); + if (found && node) + { + /* Patch up the FS, since this might have come from an old FS + * object. */ + svn_fs_fs__dag_set_fs(node, root->fs); + bucket->node = node; + } + } + else + { + node = bucket->node; + } + + /* if we found a node, make sure it remains valid at least as long + as it would when allocated in POOL. */ + if (node && needs_lock_cache) + lock_cache(ffd->dag_node_cache, pool); + } + else + { + /* DAG is mutable / may become invalid. Use the TXN-local cache */ + + locate_cache(&cache, &key, root, path, pool); + + SVN_ERR(svn_cache__get((void **) &node, &found, cache, key, pool)); + if (found && node) + { + /* Patch up the FS, since this might have come from an old FS + * object. */ + svn_fs_fs__dag_set_fs(node, root->fs); + } + } + + *node_p = node; + + return SVN_NO_ERROR; +} + + +/* Add the NODE for PATH to ROOT's node cache. */ +static svn_error_t * +dag_node_cache_set(svn_fs_root_t *root, + const char *path, + dag_node_t *node, + apr_pool_t *pool) +{ + svn_cache__t *cache; + const char *key; + + SVN_ERR_ASSERT(*path == '/'); + + /* Do *not* attempt to dup and put the node into L1. + * dup() is twice as expensive as an L2 lookup (which will set also L1). + */ + locate_cache(&cache, &key, root, path, pool); + + return svn_cache__set(cache, key, node, pool); +} + + +/* Baton for find_descendents_in_cache. */ +struct fdic_baton { + const char *path; + apr_array_header_t *list; + apr_pool_t *pool; +}; + +/* If the given item is a descendent of BATON->PATH, push + * it onto BATON->LIST (copying into BATON->POOL). Implements + * the svn_iter_apr_hash_cb_t prototype. */ +static svn_error_t * +find_descendents_in_cache(void *baton, + const void *key, + apr_ssize_t klen, + void *val, + apr_pool_t *pool) +{ + struct fdic_baton *b = baton; + const char *item_path = key; + + if (svn_fspath__skip_ancestor(b->path, item_path)) + APR_ARRAY_PUSH(b->list, const char *) = apr_pstrdup(b->pool, item_path); + + return SVN_NO_ERROR; +} + +/* Invalidate cache entries for PATH and any of its children. This + should *only* be called on a transaction root! */ +static svn_error_t * +dag_node_cache_invalidate(svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + struct fdic_baton b; + svn_cache__t *cache; + apr_pool_t *iterpool; + int i; + + b.path = path; + b.pool = svn_pool_create(pool); + b.list = apr_array_make(b.pool, 1, sizeof(const char *)); + + SVN_ERR_ASSERT(root->is_txn_root); + locate_cache(&cache, NULL, root, NULL, b.pool); + + + SVN_ERR(svn_cache__iter(NULL, cache, find_descendents_in_cache, + &b, b.pool)); + + iterpool = svn_pool_create(b.pool); + + for (i = 0; i < b.list->nelts; i++) + { + const char *descendent = APR_ARRAY_IDX(b.list, i, const char *); + svn_pool_clear(iterpool); + SVN_ERR(svn_cache__set(cache, descendent, NULL, iterpool)); + } + + svn_pool_destroy(iterpool); + svn_pool_destroy(b.pool); + return SVN_NO_ERROR; +} + + + +/* Creating transaction and revision root nodes. */ + +svn_error_t * +svn_fs_fs__txn_root(svn_fs_root_t **root_p, + svn_fs_txn_t *txn, + apr_pool_t *pool) +{ + apr_uint32_t flags = 0; + apr_hash_t *txnprops; + + /* Look for the temporary txn props representing 'flags'. */ + SVN_ERR(svn_fs_fs__txn_proplist(&txnprops, txn, pool)); + if (txnprops) + { + if (svn_hash_gets(txnprops, SVN_FS__PROP_TXN_CHECK_OOD)) + flags |= SVN_FS_TXN_CHECK_OOD; + + if (svn_hash_gets(txnprops, SVN_FS__PROP_TXN_CHECK_LOCKS)) + flags |= SVN_FS_TXN_CHECK_LOCKS; + } + + return make_txn_root(root_p, txn->fs, txn->id, txn->base_rev, flags, pool); +} + + +svn_error_t * +svn_fs_fs__revision_root(svn_fs_root_t **root_p, + svn_fs_t *fs, + svn_revnum_t rev, + apr_pool_t *pool) +{ + dag_node_t *root_dir; + + SVN_ERR(svn_fs__check_fs(fs, TRUE)); + + SVN_ERR(svn_fs_fs__dag_revision_root(&root_dir, fs, rev, pool)); + + *root_p = make_revision_root(fs, rev, root_dir, pool); + + return SVN_NO_ERROR; +} + + + +/* Getting dag nodes for roots. */ + + +/* Set *NODE_P to a freshly opened dag node referring to the root + directory of ROOT, allocating from POOL. */ +static svn_error_t * +root_node(dag_node_t **node_p, + svn_fs_root_t *root, + apr_pool_t *pool) +{ + if (root->is_txn_root) + { + /* It's a transaction root. Open a fresh copy. */ + return svn_fs_fs__dag_txn_root(node_p, root->fs, root->txn, pool); + } + else + { + /* It's a revision root, so we already have its root directory + opened. */ + fs_rev_root_data_t *frd = root->fsap_data; + *node_p = svn_fs_fs__dag_dup(frd->root_dir, pool); + return SVN_NO_ERROR; + } +} + + +/* Set *NODE_P to a mutable root directory for ROOT, cloning if + necessary, allocating in POOL. ROOT must be a transaction root. + Use ERROR_PATH in error messages. */ +static svn_error_t * +mutable_root_node(dag_node_t **node_p, + svn_fs_root_t *root, + const char *error_path, + apr_pool_t *pool) +{ + if (root->is_txn_root) + return svn_fs_fs__dag_clone_root(node_p, root->fs, root->txn, pool); + else + /* If it's not a transaction root, we can't change its contents. */ + return SVN_FS__ERR_NOT_MUTABLE(root->fs, root->rev, error_path); +} + + + +/* Traversing directory paths. */ + +typedef enum copy_id_inherit_t +{ + copy_id_inherit_unknown = 0, + copy_id_inherit_self, + copy_id_inherit_parent, + copy_id_inherit_new + +} copy_id_inherit_t; + +/* A linked list representing the path from a node up to a root + directory. We use this for cloning, and for operations that need + to deal with both a node and its parent directory. For example, a + `delete' operation needs to know that the node actually exists, but + also needs to change the parent directory. */ +typedef struct parent_path_t +{ + + /* A node along the path. This could be the final node, one of its + parents, or the root. Every parent path ends with an element for + the root directory. */ + dag_node_t *node; + + /* The name NODE has in its parent directory. This is zero for the + root directory, which (obviously) has no name in its parent. */ + char *entry; + + /* The parent of NODE, or zero if NODE is the root directory. */ + struct parent_path_t *parent; + + /* The copy ID inheritance style. */ + copy_id_inherit_t copy_inherit; + + /* If copy ID inheritance style is copy_id_inherit_new, this is the + path which should be implicitly copied; otherwise, this is NULL. */ + const char *copy_src_path; + +} parent_path_t; + +/* Return a text string describing the absolute path of parent_path + PARENT_PATH. It will be allocated in POOL. */ +static const char * +parent_path_path(parent_path_t *parent_path, + apr_pool_t *pool) +{ + const char *path_so_far = "/"; + if (parent_path->parent) + path_so_far = parent_path_path(parent_path->parent, pool); + return parent_path->entry + ? svn_fspath__join(path_so_far, parent_path->entry, pool) + : path_so_far; +} + + +/* Return the FS path for the parent path chain object CHILD relative + to its ANCESTOR in the same chain, allocated in POOL. */ +static const char * +parent_path_relpath(parent_path_t *child, + parent_path_t *ancestor, + apr_pool_t *pool) +{ + const char *path_so_far = ""; + parent_path_t *this_node = child; + while (this_node != ancestor) + { + assert(this_node != NULL); + path_so_far = svn_relpath_join(this_node->entry, path_so_far, pool); + this_node = this_node->parent; + } + return path_so_far; +} + + + +/* Choose a copy ID inheritance method *INHERIT_P to be used in the + event that immutable node CHILD in FS needs to be made mutable. If + the inheritance method is copy_id_inherit_new, also return a + *COPY_SRC_PATH on which to base the new copy ID (else return NULL + for that path). CHILD must have a parent (it cannot be the root + node). TXN_ID is the transaction in which these items might be + mutable. Allocations are taken from POOL. */ +static svn_error_t * +get_copy_inheritance(copy_id_inherit_t *inherit_p, + const char **copy_src_path, + svn_fs_t *fs, + parent_path_t *child, + const char *txn_id, + apr_pool_t *pool) +{ + const svn_fs_id_t *child_id, *parent_id, *copyroot_id; + const char *child_copy_id, *parent_copy_id; + const char *id_path = NULL; + svn_fs_root_t *copyroot_root; + dag_node_t *copyroot_node; + svn_revnum_t copyroot_rev; + const char *copyroot_path; + + SVN_ERR_ASSERT(child && child->parent && txn_id); + + /* Initialize some convenience variables. */ + child_id = svn_fs_fs__dag_get_id(child->node); + parent_id = svn_fs_fs__dag_get_id(child->parent->node); + child_copy_id = svn_fs_fs__id_copy_id(child_id); + parent_copy_id = svn_fs_fs__id_copy_id(parent_id); + + /* If this child is already mutable, we have nothing to do. */ + if (svn_fs_fs__id_txn_id(child_id)) + { + *inherit_p = copy_id_inherit_self; + *copy_src_path = NULL; + return SVN_NO_ERROR; + } + + /* From this point on, we'll assume that the child will just take + its copy ID from its parent. */ + *inherit_p = copy_id_inherit_parent; + *copy_src_path = NULL; + + /* Special case: if the child's copy ID is '0', use the parent's + copy ID. */ + if (strcmp(child_copy_id, "0") == 0) + return SVN_NO_ERROR; + + /* Compare the copy IDs of the child and its parent. If they are + the same, then the child is already on the same branch as the + parent, and should use the same mutability copy ID that the + parent will use. */ + if (svn_fs_fs__key_compare(child_copy_id, parent_copy_id) == 0) + return SVN_NO_ERROR; + + /* If the child is on the same branch that the parent is on, the + child should just use the same copy ID that the parent would use. + Else, the child needs to generate a new copy ID to use should it + need to be made mutable. We will claim that child is on the same + branch as its parent if the child itself is not a branch point, + or if it is a branch point that we are accessing via its original + copy destination path. */ + SVN_ERR(svn_fs_fs__dag_get_copyroot(©root_rev, ©root_path, + child->node)); + SVN_ERR(svn_fs_fs__revision_root(©root_root, fs, copyroot_rev, pool)); + SVN_ERR(get_dag(©root_node, copyroot_root, copyroot_path, FALSE, pool)); + copyroot_id = svn_fs_fs__dag_get_id(copyroot_node); + + if (svn_fs_fs__id_compare(copyroot_id, child_id) == -1) + return SVN_NO_ERROR; + + /* Determine if we are looking at the child via its original path or + as a subtree item of a copied tree. */ + id_path = svn_fs_fs__dag_get_created_path(child->node); + if (strcmp(id_path, parent_path_path(child, pool)) == 0) + { + *inherit_p = copy_id_inherit_self; + return SVN_NO_ERROR; + } + + /* We are pretty sure that the child node is an unedited nested + branched node. When it needs to be made mutable, it should claim + a new copy ID. */ + *inherit_p = copy_id_inherit_new; + *copy_src_path = id_path; + return SVN_NO_ERROR; +} + +/* Allocate a new parent_path_t node from POOL, referring to NODE, + ENTRY, PARENT, and COPY_ID. */ +static parent_path_t * +make_parent_path(dag_node_t *node, + char *entry, + parent_path_t *parent, + apr_pool_t *pool) +{ + parent_path_t *parent_path = apr_pcalloc(pool, sizeof(*parent_path)); + parent_path->node = node; + parent_path->entry = entry; + parent_path->parent = parent; + parent_path->copy_inherit = copy_id_inherit_unknown; + parent_path->copy_src_path = NULL; + return parent_path; +} + + +/* Flags for open_path. */ +typedef enum open_path_flags_t { + + /* The last component of the PATH need not exist. (All parent + directories must exist, as usual.) If the last component doesn't + exist, simply leave the `node' member of the bottom parent_path + component zero. */ + open_path_last_optional = 1, + + /* When this flag is set, don't bother to lookup the DAG node in + our caches because we already tried this. Ignoring this flag + has no functional impact. */ + open_path_uncached = 2, + + /* The caller does not care about the parent node chain but only + the final DAG node. */ + open_path_node_only = 4 +} open_path_flags_t; + + +/* Open the node identified by PATH in ROOT, allocating in POOL. Set + *PARENT_PATH_P to a path from the node up to ROOT. The resulting + **PARENT_PATH_P value is guaranteed to contain at least one + *element, for the root directory. PATH must be in canonical form. + + If resulting *PARENT_PATH_P will eventually be made mutable and + modified, or if copy ID inheritance information is otherwise + needed, TXN_ID should be the ID of the mutability transaction. If + TXN_ID is NULL, no copy ID inheritance information will be + calculated for the *PARENT_PATH_P chain. + + If FLAGS & open_path_last_optional is zero, return the error + SVN_ERR_FS_NOT_FOUND if the node PATH refers to does not exist. If + non-zero, require all the parent directories to exist as normal, + but if the final path component doesn't exist, simply return a path + whose bottom `node' member is zero. This option is useful for + callers that create new nodes --- we find the parent directory for + them, and tell them whether the entry exists already. + + The remaining bits in FLAGS are hints that allow this function + to take shortcuts based on knowledge that the caller provides, + such as the caller is not actually being interested in PARENT_PATH_P, + but only in (*PARENT_PATH_P)->NODE. + + NOTE: Public interfaces which only *read* from the filesystem + should not call this function directly, but should instead use + get_dag(). +*/ +static svn_error_t * +open_path(parent_path_t **parent_path_p, + svn_fs_root_t *root, + const char *path, + int flags, + const char *txn_id, + apr_pool_t *pool) +{ + svn_fs_t *fs = root->fs; + dag_node_t *here = NULL; /* The directory we're currently looking at. */ + parent_path_t *parent_path; /* The path from HERE up to the root. */ + const char *rest; /* The portion of PATH we haven't traversed yet. */ + + /* ensure a canonical path representation */ + const char *path_so_far = "/"; + apr_pool_t *iterpool = svn_pool_create(pool); + + /* callers often traverse the tree in some path-based order. That means + a sibling of PATH has been presently accessed. Try to start the lookup + directly at the parent node, if the caller did not requested the full + parent chain. */ + const char *directory; + assert(svn_fs__is_canonical_abspath(path)); + if (flags & open_path_node_only) + { + directory = svn_dirent_dirname(path, pool); + if (directory[1] != 0) /* root nodes are covered anyway */ + SVN_ERR(dag_node_cache_get(&here, root, directory, TRUE, pool)); + } + + /* did the shortcut work? */ + if (here) + { + path_so_far = directory; + rest = path + strlen(directory) + 1; + } + else + { + /* Make a parent_path item for the root node, using its own current + copy id. */ + SVN_ERR(root_node(&here, root, pool)); + rest = path + 1; /* skip the leading '/', it saves in iteration */ + } + + parent_path = make_parent_path(here, 0, 0, pool); + parent_path->copy_inherit = copy_id_inherit_self; + + /* Whenever we are at the top of this loop: + - HERE is our current directory, + - ID is the node revision ID of HERE, + - REST is the path we're going to find in HERE, and + - PARENT_PATH includes HERE and all its parents. */ + for (;;) + { + const char *next; + char *entry; + dag_node_t *child; + + svn_pool_clear(iterpool); + + /* Parse out the next entry from the path. */ + entry = svn_fs__next_entry_name(&next, rest, pool); + + /* Calculate the path traversed thus far. */ + path_so_far = svn_fspath__join(path_so_far, entry, pool); + + if (*entry == '\0') + { + /* Given the behavior of svn_fs__next_entry_name(), this + happens when the path either starts or ends with a slash. + In either case, we stay put: the current directory stays + the same, and we add nothing to the parent path. */ + child = here; + } + else + { + copy_id_inherit_t inherit; + const char *copy_path = NULL; + svn_error_t *err = SVN_NO_ERROR; + dag_node_t *cached_node = NULL; + + /* If we found a directory entry, follow it. First, we + check our node cache, and, failing that, we hit the DAG + layer. Don't bother to contact the cache for the last + element if we already know the lookup to fail for the + complete path. */ + if (next || !(flags & open_path_uncached)) + SVN_ERR(dag_node_cache_get(&cached_node, root, path_so_far, + TRUE, pool)); + if (cached_node) + child = cached_node; + else + err = svn_fs_fs__dag_open(&child, here, entry, pool, iterpool); + + /* "file not found" requires special handling. */ + if (err && err->apr_err == SVN_ERR_FS_NOT_FOUND) + { + /* If this was the last path component, and the caller + said it was optional, then don't return an error; + just put a NULL node pointer in the path. */ + + svn_error_clear(err); + + if ((flags & open_path_last_optional) + && (! next || *next == '\0')) + { + parent_path = make_parent_path(NULL, entry, parent_path, + pool); + break; + } + else + { + /* Build a better error message than svn_fs_fs__dag_open + can provide, giving the root and full path name. */ + return SVN_FS__NOT_FOUND(root, path); + } + } + + /* Other errors we return normally. */ + SVN_ERR(err); + + if (flags & open_path_node_only) + { + /* Shortcut: the caller only wan'ts the final DAG node. */ + parent_path->node = child; + } + else + { + /* Now, make a parent_path item for CHILD. */ + parent_path = make_parent_path(child, entry, parent_path, pool); + if (txn_id) + { + SVN_ERR(get_copy_inheritance(&inherit, ©_path, fs, + parent_path, txn_id, iterpool)); + parent_path->copy_inherit = inherit; + parent_path->copy_src_path = apr_pstrdup(pool, copy_path); + } + } + + /* Cache the node we found (if it wasn't already cached). */ + if (! cached_node) + SVN_ERR(dag_node_cache_set(root, path_so_far, child, iterpool)); + } + + /* Are we finished traversing the path? */ + if (! next) + break; + + /* The path isn't finished yet; we'd better be in a directory. */ + if (svn_fs_fs__dag_node_kind(child) != svn_node_dir) + SVN_ERR_W(SVN_FS__ERR_NOT_DIRECTORY(fs, path_so_far), + apr_psprintf(iterpool, _("Failure opening '%s'"), path)); + + rest = next; + here = child; + } + + svn_pool_destroy(iterpool); + *parent_path_p = parent_path; + return SVN_NO_ERROR; +} + + +/* Make the node referred to by PARENT_PATH mutable, if it isn't + already, allocating from POOL. ROOT must be the root from which + PARENT_PATH descends. Clone any parent directories as needed. + Adjust the dag nodes in PARENT_PATH to refer to the clones. Use + ERROR_PATH in error messages. */ +static svn_error_t * +make_path_mutable(svn_fs_root_t *root, + parent_path_t *parent_path, + const char *error_path, + apr_pool_t *pool) +{ + dag_node_t *clone; + const char *txn_id = root->txn; + + /* Is the node mutable already? */ + if (svn_fs_fs__dag_check_mutable(parent_path->node)) + return SVN_NO_ERROR; + + /* Are we trying to clone the root, or somebody's child node? */ + if (parent_path->parent) + { + const svn_fs_id_t *parent_id, *child_id, *copyroot_id; + const char *copy_id = NULL; + copy_id_inherit_t inherit = parent_path->copy_inherit; + const char *clone_path, *copyroot_path; + svn_revnum_t copyroot_rev; + svn_boolean_t is_parent_copyroot = FALSE; + svn_fs_root_t *copyroot_root; + dag_node_t *copyroot_node; + + /* We're trying to clone somebody's child. Make sure our parent + is mutable. */ + SVN_ERR(make_path_mutable(root, parent_path->parent, + error_path, pool)); + + switch (inherit) + { + case copy_id_inherit_parent: + parent_id = svn_fs_fs__dag_get_id(parent_path->parent->node); + copy_id = svn_fs_fs__id_copy_id(parent_id); + break; + + case copy_id_inherit_new: + SVN_ERR(svn_fs_fs__reserve_copy_id(©_id, root->fs, txn_id, + pool)); + break; + + case copy_id_inherit_self: + copy_id = NULL; + break; + + case copy_id_inherit_unknown: + default: + SVN_ERR_MALFUNCTION(); /* uh-oh -- somebody didn't calculate copy-ID + inheritance data. */ + } + + /* Determine what copyroot our new child node should use. */ + SVN_ERR(svn_fs_fs__dag_get_copyroot(©root_rev, ©root_path, + parent_path->node)); + SVN_ERR(svn_fs_fs__revision_root(©root_root, root->fs, + copyroot_rev, pool)); + SVN_ERR(get_dag(©root_node, copyroot_root, copyroot_path, + FALSE, pool)); + + child_id = svn_fs_fs__dag_get_id(parent_path->node); + copyroot_id = svn_fs_fs__dag_get_id(copyroot_node); + if (strcmp(svn_fs_fs__id_node_id(child_id), + svn_fs_fs__id_node_id(copyroot_id)) != 0) + is_parent_copyroot = TRUE; + + /* Now make this node mutable. */ + clone_path = parent_path_path(parent_path->parent, pool); + SVN_ERR(svn_fs_fs__dag_clone_child(&clone, + parent_path->parent->node, + clone_path, + parent_path->entry, + copy_id, txn_id, + is_parent_copyroot, + pool)); + + /* Update the path cache. */ + SVN_ERR(dag_node_cache_set(root, parent_path_path(parent_path, pool), + clone, pool)); + } + else + { + /* We're trying to clone the root directory. */ + SVN_ERR(mutable_root_node(&clone, root, error_path, pool)); + } + + /* Update the PARENT_PATH link to refer to the clone. */ + parent_path->node = clone; + + return SVN_NO_ERROR; +} + + +/* Open the node identified by PATH in ROOT. Set DAG_NODE_P to the + node we find, allocated in POOL. Return the error + SVN_ERR_FS_NOT_FOUND if this node doesn't exist. + + Since locking can be expensive and POOL may be long-living, for + nodes that will not need to survive the next call to this function, + set NEEDS_LOCK_CACHE to FALSE. */ +static svn_error_t * +get_dag(dag_node_t **dag_node_p, + svn_fs_root_t *root, + const char *path, + svn_boolean_t needs_lock_cache, + apr_pool_t *pool) +{ + parent_path_t *parent_path; + dag_node_t *node = NULL; + + /* First we look for the DAG in our cache + (if the path may be canonical). */ + if (*path == '/') + SVN_ERR(dag_node_cache_get(&node, root, path, needs_lock_cache, pool)); + + if (! node) + { + /* Canonicalize the input PATH. */ + if (! svn_fs__is_canonical_abspath(path)) + { + path = svn_fs__canonicalize_abspath(path, pool); + + /* Try again with the corrected path. */ + SVN_ERR(dag_node_cache_get(&node, root, path, needs_lock_cache, + pool)); + } + + if (! node) + { + /* Call open_path with no flags, as we want this to return an + * error if the node for which we are searching doesn't exist. */ + SVN_ERR(open_path(&parent_path, root, path, + open_path_uncached | open_path_node_only, + NULL, pool)); + node = parent_path->node; + + /* No need to cache our find -- open_path() will do that for us. */ + } + } + + *dag_node_p = node; + return SVN_NO_ERROR; +} + + + +/* Populating the `changes' table. */ + +/* Add a change to the changes table in FS, keyed on transaction id + TXN_ID, and indicated that a change of kind CHANGE_KIND occurred on + PATH (whose node revision id is--or was, in the case of a + deletion--NODEREV_ID), and optionally that TEXT_MODs or PROP_MODs + occurred. If the change resulted from a copy, COPYFROM_REV and + COPYFROM_PATH specify under which revision and path the node was + copied from. If this was not part of a copy, COPYFROM_REV should + be SVN_INVALID_REVNUM. Do all this as part of POOL. */ +static svn_error_t * +add_change(svn_fs_t *fs, + const char *txn_id, + const char *path, + const svn_fs_id_t *noderev_id, + svn_fs_path_change_kind_t change_kind, + svn_boolean_t text_mod, + svn_boolean_t prop_mod, + svn_node_kind_t node_kind, + svn_revnum_t copyfrom_rev, + const char *copyfrom_path, + apr_pool_t *pool) +{ + return svn_fs_fs__add_change(fs, txn_id, + svn_fs__canonicalize_abspath(path, pool), + noderev_id, change_kind, text_mod, prop_mod, + node_kind, copyfrom_rev, copyfrom_path, + pool); +} + + + +/* Generic node operations. */ + +/* Get the id of a node referenced by path PATH in ROOT. Return the + id in *ID_P allocated in POOL. */ +svn_error_t * +svn_fs_fs__node_id(const svn_fs_id_t **id_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + if ((! root->is_txn_root) + && (path[0] == '\0' || ((path[0] == '/') && (path[1] == '\0')))) + { + /* Optimize the case where we don't need any db access at all. + The root directory ("" or "/") node is stored in the + svn_fs_root_t object, and never changes when it's a revision + root, so we can just reach in and grab it directly. */ + fs_rev_root_data_t *frd = root->fsap_data; + *id_p = svn_fs_fs__id_copy(svn_fs_fs__dag_get_id(frd->root_dir), pool); + } + else + { + dag_node_t *node; + + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + *id_p = svn_fs_fs__id_copy(svn_fs_fs__dag_get_id(node), pool); + } + return SVN_NO_ERROR; +} + + +svn_error_t * +svn_fs_fs__node_created_rev(svn_revnum_t *revision, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *node; + + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + return svn_fs_fs__dag_get_revision(revision, node, pool); +} + + +/* Set *CREATED_PATH to the path at which PATH under ROOT was created. + Return a string allocated in POOL. */ +static svn_error_t * +fs_node_created_path(const char **created_path, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *node; + + SVN_ERR(get_dag(&node, root, path, TRUE, pool)); + *created_path = svn_fs_fs__dag_get_created_path(node); + + return SVN_NO_ERROR; +} + + +/* Set *KIND_P to the type of node located at PATH under ROOT. + Perform temporary allocations in POOL. */ +static svn_error_t * +node_kind(svn_node_kind_t *kind_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + const svn_fs_id_t *node_id; + dag_node_t *node; + + /* Get the node id. */ + SVN_ERR(svn_fs_fs__node_id(&node_id, root, path, pool)); + + /* Use the node id to get the real kind. */ + SVN_ERR(svn_fs_fs__dag_get_node(&node, root->fs, node_id, pool)); + *kind_p = svn_fs_fs__dag_node_kind(node); + + return SVN_NO_ERROR; +} + + +/* Set *KIND_P to the type of node present at PATH under ROOT. If + PATH does not exist under ROOT, set *KIND_P to svn_node_none. Use + POOL for temporary allocation. */ +svn_error_t * +svn_fs_fs__check_path(svn_node_kind_t *kind_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + svn_error_t *err = node_kind(kind_p, root, path, pool); + if (err && + ((err->apr_err == SVN_ERR_FS_NOT_FOUND) + || (err->apr_err == SVN_ERR_FS_NOT_DIRECTORY))) + { + svn_error_clear(err); + err = SVN_NO_ERROR; + *kind_p = svn_node_none; + } + + return svn_error_trace(err); +} + +/* Set *VALUE_P to the value of the property named PROPNAME of PATH in + ROOT. If the node has no property by that name, set *VALUE_P to + zero. Allocate the result in POOL. */ +static svn_error_t * +fs_node_prop(svn_string_t **value_p, + svn_fs_root_t *root, + const char *path, + const char *propname, + apr_pool_t *pool) +{ + dag_node_t *node; + apr_hash_t *proplist; + + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + SVN_ERR(svn_fs_fs__dag_get_proplist(&proplist, node, pool)); + *value_p = NULL; + if (proplist) + *value_p = svn_hash_gets(proplist, propname); + + return SVN_NO_ERROR; +} + + +/* Set *TABLE_P to the entire property list of PATH under ROOT, as an + APR hash table allocated in POOL. The resulting property table + maps property names to pointers to svn_string_t objects containing + the property value. */ +static svn_error_t * +fs_node_proplist(apr_hash_t **table_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + apr_hash_t *table; + dag_node_t *node; + + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + SVN_ERR(svn_fs_fs__dag_get_proplist(&table, node, pool)); + *table_p = table ? table : apr_hash_make(pool); + + return SVN_NO_ERROR; +} + + +static svn_error_t * +increment_mergeinfo_up_tree(parent_path_t *pp, + apr_int64_t increment, + apr_pool_t *pool) +{ + for (; pp; pp = pp->parent) + SVN_ERR(svn_fs_fs__dag_increment_mergeinfo_count(pp->node, + increment, + pool)); + + return SVN_NO_ERROR; +} + +/* Change, add, or delete a node's property value. The affected node + is PATH under ROOT, the property value to modify is NAME, and VALUE + points to either a string value to set the new contents to, or NULL + if the property should be deleted. Perform temporary allocations + in POOL. */ +static svn_error_t * +fs_change_node_prop(svn_fs_root_t *root, + const char *path, + const char *name, + const svn_string_t *value, + apr_pool_t *pool) +{ + parent_path_t *parent_path; + apr_hash_t *proplist; + const char *txn_id; + + if (! root->is_txn_root) + return SVN_FS__NOT_TXN(root); + txn_id = root->txn; + + path = svn_fs__canonicalize_abspath(path, pool); + SVN_ERR(open_path(&parent_path, root, path, 0, txn_id, pool)); + + /* Check (non-recursively) to see if path is locked; if so, check + that we can use it. */ + if (root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(path, root->fs, FALSE, FALSE, + pool)); + + SVN_ERR(make_path_mutable(root, parent_path, path, pool)); + SVN_ERR(svn_fs_fs__dag_get_proplist(&proplist, parent_path->node, pool)); + + /* If there's no proplist, but we're just deleting a property, exit now. */ + if ((! proplist) && (! value)) + return SVN_NO_ERROR; + + /* Now, if there's no proplist, we know we need to make one. */ + if (! proplist) + proplist = apr_hash_make(pool); + + if (svn_fs_fs__fs_supports_mergeinfo(root->fs) + && strcmp (name, SVN_PROP_MERGEINFO) == 0) + { + apr_int64_t increment = 0; + svn_boolean_t had_mergeinfo; + SVN_ERR(svn_fs_fs__dag_has_mergeinfo(&had_mergeinfo, parent_path->node)); + + if (value && !had_mergeinfo) + increment = 1; + else if (!value && had_mergeinfo) + increment = -1; + + if (increment != 0) + { + SVN_ERR(increment_mergeinfo_up_tree(parent_path, increment, pool)); + SVN_ERR(svn_fs_fs__dag_set_has_mergeinfo(parent_path->node, + (value != NULL), pool)); + } + } + + /* Set the property. */ + svn_hash_sets(proplist, name, value); + + /* Overwrite the node's proplist. */ + SVN_ERR(svn_fs_fs__dag_set_proplist(parent_path->node, proplist, + pool)); + + /* Make a record of this modification in the changes table. */ + return add_change(root->fs, txn_id, path, + svn_fs_fs__dag_get_id(parent_path->node), + svn_fs_path_change_modify, FALSE, TRUE, + svn_fs_fs__dag_node_kind(parent_path->node), + SVN_INVALID_REVNUM, NULL, pool); +} + + +/* Determine if the properties of two path/root combinations are + different. Set *CHANGED_P to TRUE if the properties at PATH1 under + ROOT1 differ from those at PATH2 under ROOT2, or FALSE otherwise. + Both roots must be in the same filesystem. */ +static svn_error_t * +fs_props_changed(svn_boolean_t *changed_p, + svn_fs_root_t *root1, + const char *path1, + svn_fs_root_t *root2, + const char *path2, + apr_pool_t *pool) +{ + dag_node_t *node1, *node2; + + /* Check that roots are in the same fs. */ + if (root1->fs != root2->fs) + return svn_error_create + (SVN_ERR_FS_GENERAL, NULL, + _("Cannot compare property value between two different filesystems")); + + SVN_ERR(get_dag(&node1, root1, path1, TRUE, pool)); + SVN_ERR(get_dag(&node2, root2, path2, TRUE, pool)); + return svn_fs_fs__dag_things_different(changed_p, NULL, + node1, node2); +} + + + +/* Merges and commits. */ + +/* Set *NODE to the root node of ROOT. */ +static svn_error_t * +get_root(dag_node_t **node, svn_fs_root_t *root, apr_pool_t *pool) +{ + return get_dag(node, root, "/", TRUE, pool); +} + + +/* Set the contents of CONFLICT_PATH to PATH, and return an + SVN_ERR_FS_CONFLICT error that indicates that there was a conflict + at PATH. Perform all allocations in POOL (except the allocation of + CONFLICT_PATH, which should be handled outside this function). */ +static svn_error_t * +conflict_err(svn_stringbuf_t *conflict_path, + const char *path) +{ + svn_stringbuf_set(conflict_path, path); + return svn_error_createf(SVN_ERR_FS_CONFLICT, NULL, + _("Conflict at '%s'"), path); +} + + +/* Merge changes between ANCESTOR and SOURCE into TARGET. ANCESTOR + * and TARGET must be distinct node revisions. TARGET_PATH should + * correspond to TARGET's full path in its filesystem, and is used for + * reporting conflict location. + * + * SOURCE, TARGET, and ANCESTOR are generally directories; this + * function recursively merges the directories' contents. If any are + * files, this function simply returns an error whenever SOURCE, + * TARGET, and ANCESTOR are all distinct node revisions. + * + * If there are differences between ANCESTOR and SOURCE that conflict + * with changes between ANCESTOR and TARGET, this function returns an + * SVN_ERR_FS_CONFLICT error, and updates CONFLICT_P to the name of the + * conflicting node in TARGET, with TARGET_PATH prepended as a path. + * + * If there are no conflicting differences, CONFLICT_P is updated to + * the empty string. + * + * CONFLICT_P must point to a valid svn_stringbuf_t. + * + * Do any necessary temporary allocation in POOL. + */ +static svn_error_t * +merge(svn_stringbuf_t *conflict_p, + const char *target_path, + dag_node_t *target, + dag_node_t *source, + dag_node_t *ancestor, + const char *txn_id, + apr_int64_t *mergeinfo_increment_out, + apr_pool_t *pool) +{ + const svn_fs_id_t *source_id, *target_id, *ancestor_id; + apr_hash_t *s_entries, *t_entries, *a_entries; + apr_hash_index_t *hi; + svn_fs_t *fs; + apr_pool_t *iterpool; + apr_int64_t mergeinfo_increment = 0; + svn_boolean_t fs_supports_mergeinfo; + + /* Make sure everyone comes from the same filesystem. */ + fs = svn_fs_fs__dag_get_fs(ancestor); + if ((fs != svn_fs_fs__dag_get_fs(source)) + || (fs != svn_fs_fs__dag_get_fs(target))) + { + return svn_error_create + (SVN_ERR_FS_CORRUPT, NULL, + _("Bad merge; ancestor, source, and target not all in same fs")); + } + + /* We have the same fs, now check it. */ + SVN_ERR(svn_fs__check_fs(fs, TRUE)); + + source_id = svn_fs_fs__dag_get_id(source); + target_id = svn_fs_fs__dag_get_id(target); + ancestor_id = svn_fs_fs__dag_get_id(ancestor); + + /* It's improper to call this function with ancestor == target. */ + if (svn_fs_fs__id_eq(ancestor_id, target_id)) + { + svn_string_t *id_str = svn_fs_fs__id_unparse(target_id, pool); + return svn_error_createf + (SVN_ERR_FS_GENERAL, NULL, + _("Bad merge; target '%s' has id '%s', same as ancestor"), + target_path, id_str->data); + } + + svn_stringbuf_setempty(conflict_p); + + /* Base cases: + * Either no change made in source, or same change as made in target. + * Both mean nothing to merge here. + */ + if (svn_fs_fs__id_eq(ancestor_id, source_id) + || (svn_fs_fs__id_eq(source_id, target_id))) + return SVN_NO_ERROR; + + /* Else proceed, knowing all three are distinct node revisions. + * + * How to merge from this point: + * + * if (not all 3 are directories) + * { + * early exit with conflict; + * } + * + * // Property changes may only be made to up-to-date + * // directories, because once the client commits the prop + * // change, it bumps the directory's revision, and therefore + * // must be able to depend on there being no other changes to + * // that directory in the repository. + * if (target's property list differs from ancestor's) + * conflict; + * + * For each entry NAME in the directory ANCESTOR: + * + * Let ANCESTOR-ENTRY, SOURCE-ENTRY, and TARGET-ENTRY be the IDs of + * the name within ANCESTOR, SOURCE, and TARGET respectively. + * (Possibly null if NAME does not exist in SOURCE or TARGET.) + * + * If ANCESTOR-ENTRY == SOURCE-ENTRY, then: + * No changes were made to this entry while the transaction was in + * progress, so do nothing to the target. + * + * Else if ANCESTOR-ENTRY == TARGET-ENTRY, then: + * A change was made to this entry while the transaction was in + * process, but the transaction did not touch this entry. Replace + * TARGET-ENTRY with SOURCE-ENTRY. + * + * Else: + * Changes were made to this entry both within the transaction and + * to the repository while the transaction was in progress. They + * must be merged or declared to be in conflict. + * + * If SOURCE-ENTRY and TARGET-ENTRY are both null, that's a + * double delete; flag a conflict. + * + * If any of the three entries is of type file, declare a conflict. + * + * If either SOURCE-ENTRY or TARGET-ENTRY is not a direct + * modification of ANCESTOR-ENTRY (determine by comparing the + * node-id fields), declare a conflict. A replacement is + * incompatible with a modification or other replacement--even + * an identical replacement. + * + * Direct modifications were made to the directory ANCESTOR-ENTRY + * in both SOURCE and TARGET. Recursively merge these + * modifications. + * + * For each leftover entry NAME in the directory SOURCE: + * + * If NAME exists in TARGET, declare a conflict. Even if SOURCE and + * TARGET are adding exactly the same thing, two additions are not + * auto-mergeable with each other. + * + * Add NAME to TARGET with the entry from SOURCE. + * + * Now that we are done merging the changes from SOURCE into the + * directory TARGET, update TARGET's predecessor to be SOURCE. + */ + + if ((svn_fs_fs__dag_node_kind(source) != svn_node_dir) + || (svn_fs_fs__dag_node_kind(target) != svn_node_dir) + || (svn_fs_fs__dag_node_kind(ancestor) != svn_node_dir)) + { + return conflict_err(conflict_p, target_path); + } + + + /* Possible early merge failure: if target and ancestor have + different property lists, then the merge should fail. + Propchanges can *only* be committed on an up-to-date directory. + ### TODO: see issue #418 about the inelegance of this. + + Another possible, similar, early merge failure: if source and + ancestor have different property lists (meaning someone else + changed directory properties while our commit transaction was + happening), the merge should fail. See issue #2751. + */ + { + node_revision_t *tgt_nr, *anc_nr, *src_nr; + + /* Get node revisions for our id's. */ + SVN_ERR(svn_fs_fs__get_node_revision(&tgt_nr, fs, target_id, pool)); + SVN_ERR(svn_fs_fs__get_node_revision(&anc_nr, fs, ancestor_id, pool)); + SVN_ERR(svn_fs_fs__get_node_revision(&src_nr, fs, source_id, pool)); + + /* Now compare the prop-keys of the skels. Note that just because + the keys are different -doesn't- mean the proplists have + different contents. But merge() isn't concerned with contents; + it doesn't do a brute-force comparison on textual contents, so + it won't do that here either. Checking to see if the propkey + atoms are `equal' is enough. */ + if (! svn_fs_fs__noderev_same_rep_key(tgt_nr->prop_rep, anc_nr->prop_rep)) + return conflict_err(conflict_p, target_path); + if (! svn_fs_fs__noderev_same_rep_key(src_nr->prop_rep, anc_nr->prop_rep)) + return conflict_err(conflict_p, target_path); + } + + /* ### todo: it would be more efficient to simply check for a NULL + entries hash where necessary below than to allocate an empty hash + here, but another day, another day... */ + SVN_ERR(svn_fs_fs__dag_dir_entries(&s_entries, source, pool)); + SVN_ERR(svn_fs_fs__dag_dir_entries(&t_entries, target, pool)); + SVN_ERR(svn_fs_fs__dag_dir_entries(&a_entries, ancestor, pool)); + + fs_supports_mergeinfo = svn_fs_fs__fs_supports_mergeinfo(fs); + + /* for each entry E in a_entries... */ + iterpool = svn_pool_create(pool); + for (hi = apr_hash_first(pool, a_entries); + hi; + hi = apr_hash_next(hi)) + { + svn_fs_dirent_t *s_entry, *t_entry, *a_entry; + const char *name; + apr_ssize_t klen; + + svn_pool_clear(iterpool); + + name = svn__apr_hash_index_key(hi); + klen = svn__apr_hash_index_klen(hi); + a_entry = svn__apr_hash_index_val(hi); + + s_entry = apr_hash_get(s_entries, name, klen); + t_entry = apr_hash_get(t_entries, name, klen); + + /* No changes were made to this entry while the transaction was + in progress, so do nothing to the target. */ + if (s_entry && svn_fs_fs__id_eq(a_entry->id, s_entry->id)) + goto end; + + /* A change was made to this entry while the transaction was in + process, but the transaction did not touch this entry. */ + else if (t_entry && svn_fs_fs__id_eq(a_entry->id, t_entry->id)) + { + dag_node_t *t_ent_node; + SVN_ERR(svn_fs_fs__dag_get_node(&t_ent_node, fs, + t_entry->id, iterpool)); + if (fs_supports_mergeinfo) + { + apr_int64_t mergeinfo_start; + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_start, + t_ent_node)); + mergeinfo_increment -= mergeinfo_start; + } + + if (s_entry) + { + dag_node_t *s_ent_node; + SVN_ERR(svn_fs_fs__dag_get_node(&s_ent_node, fs, + s_entry->id, iterpool)); + + if (fs_supports_mergeinfo) + { + apr_int64_t mergeinfo_end; + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_end, + s_ent_node)); + mergeinfo_increment += mergeinfo_end; + } + + SVN_ERR(svn_fs_fs__dag_set_entry(target, name, + s_entry->id, + s_entry->kind, + txn_id, + iterpool)); + } + else + { + SVN_ERR(svn_fs_fs__dag_delete(target, name, txn_id, iterpool)); + } + } + + /* Changes were made to this entry both within the transaction + and to the repository while the transaction was in progress. + They must be merged or declared to be in conflict. */ + else + { + dag_node_t *s_ent_node, *t_ent_node, *a_ent_node; + const char *new_tpath; + apr_int64_t sub_mergeinfo_increment; + + /* If SOURCE-ENTRY and TARGET-ENTRY are both null, that's a + double delete; if one of them is null, that's a delete versus + a modification. In any of these cases, flag a conflict. */ + if (s_entry == NULL || t_entry == NULL) + return conflict_err(conflict_p, + svn_fspath__join(target_path, + a_entry->name, + iterpool)); + + /* If any of the three entries is of type file, flag a conflict. */ + if (s_entry->kind == svn_node_file + || t_entry->kind == svn_node_file + || a_entry->kind == svn_node_file) + return conflict_err(conflict_p, + svn_fspath__join(target_path, + a_entry->name, + iterpool)); + + /* If either SOURCE-ENTRY or TARGET-ENTRY is not a direct + modification of ANCESTOR-ENTRY, declare a conflict. */ + if (strcmp(svn_fs_fs__id_node_id(s_entry->id), + svn_fs_fs__id_node_id(a_entry->id)) != 0 + || strcmp(svn_fs_fs__id_copy_id(s_entry->id), + svn_fs_fs__id_copy_id(a_entry->id)) != 0 + || strcmp(svn_fs_fs__id_node_id(t_entry->id), + svn_fs_fs__id_node_id(a_entry->id)) != 0 + || strcmp(svn_fs_fs__id_copy_id(t_entry->id), + svn_fs_fs__id_copy_id(a_entry->id)) != 0) + return conflict_err(conflict_p, + svn_fspath__join(target_path, + a_entry->name, + iterpool)); + + /* Direct modifications were made to the directory + ANCESTOR-ENTRY in both SOURCE and TARGET. Recursively + merge these modifications. */ + SVN_ERR(svn_fs_fs__dag_get_node(&s_ent_node, fs, + s_entry->id, iterpool)); + SVN_ERR(svn_fs_fs__dag_get_node(&t_ent_node, fs, + t_entry->id, iterpool)); + SVN_ERR(svn_fs_fs__dag_get_node(&a_ent_node, fs, + a_entry->id, iterpool)); + new_tpath = svn_fspath__join(target_path, t_entry->name, iterpool); + SVN_ERR(merge(conflict_p, new_tpath, + t_ent_node, s_ent_node, a_ent_node, + txn_id, + &sub_mergeinfo_increment, + iterpool)); + if (fs_supports_mergeinfo) + mergeinfo_increment += sub_mergeinfo_increment; + } + + /* We've taken care of any possible implications E could have. + Remove it from source_entries, so it's easy later to loop + over all the source entries that didn't exist in + ancestor_entries. */ + end: + apr_hash_set(s_entries, name, klen, NULL); + } + + /* For each entry E in source but not in ancestor */ + for (hi = apr_hash_first(pool, s_entries); + hi; + hi = apr_hash_next(hi)) + { + svn_fs_dirent_t *s_entry, *t_entry; + const char *name = svn__apr_hash_index_key(hi); + apr_ssize_t klen = svn__apr_hash_index_klen(hi); + dag_node_t *s_ent_node; + + svn_pool_clear(iterpool); + + s_entry = svn__apr_hash_index_val(hi); + t_entry = apr_hash_get(t_entries, name, klen); + + /* If NAME exists in TARGET, declare a conflict. */ + if (t_entry) + return conflict_err(conflict_p, + svn_fspath__join(target_path, + t_entry->name, + iterpool)); + + SVN_ERR(svn_fs_fs__dag_get_node(&s_ent_node, fs, + s_entry->id, iterpool)); + if (fs_supports_mergeinfo) + { + apr_int64_t mergeinfo_s; + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_s, + s_ent_node)); + mergeinfo_increment += mergeinfo_s; + } + + SVN_ERR(svn_fs_fs__dag_set_entry + (target, s_entry->name, s_entry->id, s_entry->kind, + txn_id, iterpool)); + } + svn_pool_destroy(iterpool); + + SVN_ERR(svn_fs_fs__dag_update_ancestry(target, source, pool)); + + if (fs_supports_mergeinfo) + SVN_ERR(svn_fs_fs__dag_increment_mergeinfo_count(target, + mergeinfo_increment, + pool)); + + if (mergeinfo_increment_out) + *mergeinfo_increment_out = mergeinfo_increment; + + return SVN_NO_ERROR; +} + +/* Merge changes between an ancestor and SOURCE_NODE into + TXN. The ancestor is either ANCESTOR_NODE, or if + that is null, TXN's base node. + + If the merge is successful, TXN's base will become + SOURCE_NODE, and its root node will have a new ID, a + successor of SOURCE_NODE. + + If a conflict results, update *CONFLICT to the path in the txn that + conflicted; see the CONFLICT_P parameter of merge() for details. */ +static svn_error_t * +merge_changes(dag_node_t *ancestor_node, + dag_node_t *source_node, + svn_fs_txn_t *txn, + svn_stringbuf_t *conflict, + apr_pool_t *pool) +{ + dag_node_t *txn_root_node; + svn_fs_t *fs = txn->fs; + const char *txn_id = txn->id; + + SVN_ERR(svn_fs_fs__dag_txn_root(&txn_root_node, fs, txn_id, pool)); + + if (ancestor_node == NULL) + { + SVN_ERR(svn_fs_fs__dag_txn_base_root(&ancestor_node, fs, + txn_id, pool)); + } + + if (svn_fs_fs__id_eq(svn_fs_fs__dag_get_id(ancestor_node), + svn_fs_fs__dag_get_id(txn_root_node))) + { + /* If no changes have been made in TXN since its current base, + then it can't conflict with any changes since that base. + The caller isn't supposed to call us in that case. */ + SVN_ERR_MALFUNCTION(); + } + else + SVN_ERR(merge(conflict, "/", txn_root_node, + source_node, ancestor_node, txn_id, NULL, pool)); + + return SVN_NO_ERROR; +} + + +svn_error_t * +svn_fs_fs__commit_txn(const char **conflict_p, + svn_revnum_t *new_rev, + svn_fs_txn_t *txn, + apr_pool_t *pool) +{ + /* How do commits work in Subversion? + * + * When you're ready to commit, here's what you have: + * + * 1. A transaction, with a mutable tree hanging off it. + * 2. A base revision, against which TXN_TREE was made. + * 3. A latest revision, which may be newer than the base rev. + * + * The problem is that if latest != base, then one can't simply + * attach the txn root as the root of the new revision, because that + * would lose all the changes between base and latest. It is also + * not acceptable to insist that base == latest; in a busy + * repository, commits happen too fast to insist that everyone keep + * their entire tree up-to-date at all times. Non-overlapping + * changes should not interfere with each other. + * + * The solution is to merge the changes between base and latest into + * the txn tree [see the function merge()]. The txn tree is the + * only one of the three trees that is mutable, so it has to be the + * one to adjust. + * + * You might have to adjust it more than once, if a new latest + * revision gets committed while you were merging in the previous + * one. For example: + * + * 1. Jane starts txn T, based at revision 6. + * 2. Someone commits (or already committed) revision 7. + * 3. Jane's starts merging the changes between 6 and 7 into T. + * 4. Meanwhile, someone commits revision 8. + * 5. Jane finishes the 6-->7 merge. T could now be committed + * against a latest revision of 7, if only that were still the + * latest. Unfortunately, 8 is now the latest, so... + * 6. Jane starts merging the changes between 7 and 8 into T. + * 7. Meanwhile, no one commits any new revisions. Whew. + * 8. Jane commits T, creating revision 9, whose tree is exactly + * T's tree, except immutable now. + * + * Lather, rinse, repeat. + */ + + svn_error_t *err = SVN_NO_ERROR; + svn_stringbuf_t *conflict = svn_stringbuf_create_empty(pool); + svn_fs_t *fs = txn->fs; + + /* Limit memory usage when the repository has a high commit rate and + needs to run the following while loop multiple times. The memory + growth without an iteration pool is very noticeable when the + transaction modifies a node that has 20,000 sibling nodes. */ + apr_pool_t *iterpool = svn_pool_create(pool); + + /* Initialize output params. */ + *new_rev = SVN_INVALID_REVNUM; + if (conflict_p) + *conflict_p = NULL; + + while (1729) + { + svn_revnum_t youngish_rev; + svn_fs_root_t *youngish_root; + dag_node_t *youngish_root_node; + + svn_pool_clear(iterpool); + + /* Get the *current* youngest revision. We call it "youngish" + because new revisions might get committed after we've + obtained it. */ + + SVN_ERR(svn_fs_fs__youngest_rev(&youngish_rev, fs, iterpool)); + SVN_ERR(svn_fs_fs__revision_root(&youngish_root, fs, youngish_rev, + iterpool)); + + /* Get the dag node for the youngest revision. Later we'll use + it as the SOURCE argument to a merge, and if the merge + succeeds, this youngest root node will become the new base + root for the svn txn that was the target of the merge (but + note that the youngest rev may have changed by then -- that's + why we're careful to get this root in its own bdb txn + here). */ + SVN_ERR(get_root(&youngish_root_node, youngish_root, iterpool)); + + /* Try to merge. If the merge succeeds, the base root node of + TARGET's txn will become the same as youngish_root_node, so + any future merges will only be between that node and whatever + the root node of the youngest rev is by then. */ + err = merge_changes(NULL, youngish_root_node, txn, conflict, iterpool); + if (err) + { + if ((err->apr_err == SVN_ERR_FS_CONFLICT) && conflict_p) + *conflict_p = conflict->data; + goto cleanup; + } + txn->base_rev = youngish_rev; + + /* Try to commit. */ + err = svn_fs_fs__commit(new_rev, fs, txn, iterpool); + if (err && (err->apr_err == SVN_ERR_FS_TXN_OUT_OF_DATE)) + { + /* Did someone else finish committing a new revision while we + were in mid-merge or mid-commit? If so, we'll need to + loop again to merge the new changes in, then try to + commit again. Or if that's not what happened, then just + return the error. */ + svn_revnum_t youngest_rev; + SVN_ERR(svn_fs_fs__youngest_rev(&youngest_rev, fs, iterpool)); + if (youngest_rev == youngish_rev) + goto cleanup; + else + svn_error_clear(err); + } + else if (err) + { + goto cleanup; + } + else + { + err = SVN_NO_ERROR; + goto cleanup; + } + } + + cleanup: + + svn_fs_fs__reset_txn_caches(fs); + + svn_pool_destroy(iterpool); + return svn_error_trace(err); +} + + +/* Merge changes between two nodes into a third node. Given nodes + SOURCE_PATH under SOURCE_ROOT, TARGET_PATH under TARGET_ROOT and + ANCESTOR_PATH under ANCESTOR_ROOT, modify target to contain all the + changes between the ancestor and source. If there are conflicts, + return SVN_ERR_FS_CONFLICT and set *CONFLICT_P to a textual + description of the offending changes. Perform any temporary + allocations in POOL. */ +static svn_error_t * +fs_merge(const char **conflict_p, + svn_fs_root_t *source_root, + const char *source_path, + svn_fs_root_t *target_root, + const char *target_path, + svn_fs_root_t *ancestor_root, + const char *ancestor_path, + apr_pool_t *pool) +{ + dag_node_t *source, *ancestor; + svn_fs_txn_t *txn; + svn_error_t *err; + svn_stringbuf_t *conflict = svn_stringbuf_create_empty(pool); + + if (! target_root->is_txn_root) + return SVN_FS__NOT_TXN(target_root); + + /* Paranoia. */ + if ((source_root->fs != ancestor_root->fs) + || (target_root->fs != ancestor_root->fs)) + { + return svn_error_create + (SVN_ERR_FS_CORRUPT, NULL, + _("Bad merge; ancestor, source, and target not all in same fs")); + } + + /* ### kff todo: is there any compelling reason to get the nodes in + one db transaction? Right now we don't; txn_body_get_root() gets + one node at a time. This will probably need to change: + + Jim Blandy <jimb@zwingli.cygnus.com> writes: + > svn_fs_merge needs to be a single transaction, to protect it against + > people deleting parents of nodes it's working on, etc. + */ + + /* Get the ancestor node. */ + SVN_ERR(get_root(&ancestor, ancestor_root, pool)); + + /* Get the source node. */ + SVN_ERR(get_root(&source, source_root, pool)); + + /* Open a txn for the txn root into which we're merging. */ + SVN_ERR(svn_fs_fs__open_txn(&txn, ancestor_root->fs, target_root->txn, + pool)); + + /* Merge changes between ANCESTOR and SOURCE into TXN. */ + err = merge_changes(ancestor, source, txn, conflict, pool); + if (err) + { + if ((err->apr_err == SVN_ERR_FS_CONFLICT) && conflict_p) + *conflict_p = conflict->data; + return svn_error_trace(err); + } + + return SVN_NO_ERROR; +} + +svn_error_t * +svn_fs_fs__deltify(svn_fs_t *fs, + svn_revnum_t revision, + apr_pool_t *pool) +{ + /* Deltify is a no-op for fs_fs. */ + + return SVN_NO_ERROR; +} + + + +/* Directories. */ + +/* Set *TABLE_P to a newly allocated APR hash table containing the + entries of the directory at PATH in ROOT. The keys of the table + are entry names, as byte strings, excluding the final null + character; the table's values are pointers to svn_fs_dirent_t + structures. Allocate the table and its contents in POOL. */ +static svn_error_t * +fs_dir_entries(apr_hash_t **table_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *node; + + /* Get the entries for this path in the caller's pool. */ + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + return svn_fs_fs__dag_dir_entries(table_p, node, pool); +} + +/* Raise an error if PATH contains a newline because FSFS cannot handle + * such paths. See issue #4340. */ +static svn_error_t * +check_newline(const char *path, apr_pool_t *pool) +{ + char *c = strchr(path, '\n'); + + if (c) + return svn_error_createf(SVN_ERR_FS_PATH_SYNTAX, NULL, + _("Invalid control character '0x%02x' in path '%s'"), + (unsigned char)*c, svn_path_illegal_path_escape(path, pool)); + + return SVN_NO_ERROR; +} + +/* Create a new directory named PATH in ROOT. The new directory has + no entries, and no properties. ROOT must be the root of a + transaction, not a revision. Do any necessary temporary allocation + in POOL. */ +static svn_error_t * +fs_make_dir(svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + parent_path_t *parent_path; + dag_node_t *sub_dir; + const char *txn_id = root->txn; + + SVN_ERR(check_newline(path, pool)); + + path = svn_fs__canonicalize_abspath(path, pool); + SVN_ERR(open_path(&parent_path, root, path, open_path_last_optional, + txn_id, pool)); + + /* Check (recursively) to see if some lock is 'reserving' a path at + that location, or even some child-path; if so, check that we can + use it. */ + if (root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(path, root->fs, TRUE, FALSE, + pool)); + + /* If there's already a sub-directory by that name, complain. This + also catches the case of trying to make a subdirectory named `/'. */ + if (parent_path->node) + return SVN_FS__ALREADY_EXISTS(root, path); + + /* Create the subdirectory. */ + SVN_ERR(make_path_mutable(root, parent_path->parent, path, pool)); + SVN_ERR(svn_fs_fs__dag_make_dir(&sub_dir, + parent_path->parent->node, + parent_path_path(parent_path->parent, + pool), + parent_path->entry, + txn_id, + pool)); + + /* Add this directory to the path cache. */ + SVN_ERR(dag_node_cache_set(root, parent_path_path(parent_path, pool), + sub_dir, pool)); + + /* Make a record of this modification in the changes table. */ + return add_change(root->fs, txn_id, path, svn_fs_fs__dag_get_id(sub_dir), + svn_fs_path_change_add, FALSE, FALSE, svn_node_dir, + SVN_INVALID_REVNUM, NULL, pool); +} + + +/* Delete the node at PATH under ROOT. ROOT must be a transaction + root. Perform temporary allocations in POOL. */ +static svn_error_t * +fs_delete_node(svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + parent_path_t *parent_path; + const char *txn_id = root->txn; + apr_int64_t mergeinfo_count = 0; + svn_node_kind_t kind; + + if (! root->is_txn_root) + return SVN_FS__NOT_TXN(root); + + path = svn_fs__canonicalize_abspath(path, pool); + SVN_ERR(open_path(&parent_path, root, path, 0, txn_id, pool)); + kind = svn_fs_fs__dag_node_kind(parent_path->node); + + /* We can't remove the root of the filesystem. */ + if (! parent_path->parent) + return svn_error_create(SVN_ERR_FS_ROOT_DIR, NULL, + _("The root directory cannot be deleted")); + + /* Check to see if path (or any child thereof) is locked; if so, + check that we can use the existing lock(s). */ + if (root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(path, root->fs, TRUE, FALSE, + pool)); + + /* Make the parent directory mutable, and do the deletion. */ + SVN_ERR(make_path_mutable(root, parent_path->parent, path, pool)); + if (svn_fs_fs__fs_supports_mergeinfo(root->fs)) + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_count, + parent_path->node)); + SVN_ERR(svn_fs_fs__dag_delete(parent_path->parent->node, + parent_path->entry, + txn_id, pool)); + + /* Remove this node and any children from the path cache. */ + SVN_ERR(dag_node_cache_invalidate(root, parent_path_path(parent_path, pool), + pool)); + + /* Update mergeinfo counts for parents */ + if (mergeinfo_count > 0) + SVN_ERR(increment_mergeinfo_up_tree(parent_path->parent, + -mergeinfo_count, + pool)); + + /* Make a record of this modification in the changes table. */ + return add_change(root->fs, txn_id, path, + svn_fs_fs__dag_get_id(parent_path->node), + svn_fs_path_change_delete, FALSE, FALSE, kind, + SVN_INVALID_REVNUM, NULL, pool); +} + + +/* Set *SAME_P to TRUE if FS1 and FS2 have the same UUID, else set to FALSE. + Use POOL for temporary allocation only. + Note: this code is duplicated between libsvn_fs_fs and libsvn_fs_base. */ +static svn_error_t * +fs_same_p(svn_boolean_t *same_p, + svn_fs_t *fs1, + svn_fs_t *fs2, + apr_pool_t *pool) +{ + *same_p = ! strcmp(fs1->uuid, fs2->uuid); + return SVN_NO_ERROR; +} + +/* Copy the node at FROM_PATH under FROM_ROOT to TO_PATH under + TO_ROOT. If PRESERVE_HISTORY is set, then the copy is recorded in + the copies table. Perform temporary allocations in POOL. */ +static svn_error_t * +copy_helper(svn_fs_root_t *from_root, + const char *from_path, + svn_fs_root_t *to_root, + const char *to_path, + svn_boolean_t preserve_history, + apr_pool_t *pool) +{ + dag_node_t *from_node; + parent_path_t *to_parent_path; + const char *txn_id = to_root->txn; + svn_boolean_t same_p; + + /* Use an error check, not an assert, because even the caller cannot + guarantee that a filesystem's UUID has not changed "on the fly". */ + SVN_ERR(fs_same_p(&same_p, from_root->fs, to_root->fs, pool)); + if (! same_p) + return svn_error_createf + (SVN_ERR_UNSUPPORTED_FEATURE, NULL, + _("Cannot copy between two different filesystems ('%s' and '%s')"), + from_root->fs->path, to_root->fs->path); + + if (from_root->is_txn_root) + return svn_error_create + (SVN_ERR_UNSUPPORTED_FEATURE, NULL, + _("Copy from mutable tree not currently supported")); + + /* Get the NODE for FROM_PATH in FROM_ROOT.*/ + SVN_ERR(get_dag(&from_node, from_root, from_path, TRUE, pool)); + + /* Build up the parent path from TO_PATH in TO_ROOT. If the last + component does not exist, it's not that big a deal. We'll just + make one there. */ + SVN_ERR(open_path(&to_parent_path, to_root, to_path, + open_path_last_optional, txn_id, pool)); + + /* Check to see if path (or any child thereof) is locked; if so, + check that we can use the existing lock(s). */ + if (to_root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(to_path, to_root->fs, + TRUE, FALSE, pool)); + + /* If the destination node already exists as the same node as the + source (in other words, this operation would result in nothing + happening at all), just do nothing an return successfully, + proud that you saved yourself from a tiresome task. */ + if (to_parent_path->node && + svn_fs_fs__id_eq(svn_fs_fs__dag_get_id(from_node), + svn_fs_fs__dag_get_id(to_parent_path->node))) + return SVN_NO_ERROR; + + if (! from_root->is_txn_root) + { + svn_fs_path_change_kind_t kind; + dag_node_t *new_node; + const char *from_canonpath; + apr_int64_t mergeinfo_start; + apr_int64_t mergeinfo_end; + + /* If TO_PATH already existed prior to the copy, note that this + operation is a replacement, not an addition. */ + if (to_parent_path->node) + { + kind = svn_fs_path_change_replace; + if (svn_fs_fs__fs_supports_mergeinfo(to_root->fs)) + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_start, + to_parent_path->node)); + } + else + { + kind = svn_fs_path_change_add; + mergeinfo_start = 0; + } + + if (svn_fs_fs__fs_supports_mergeinfo(to_root->fs)) + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_end, + from_node)); + + /* Make sure the target node's parents are mutable. */ + SVN_ERR(make_path_mutable(to_root, to_parent_path->parent, + to_path, pool)); + + /* Canonicalize the copyfrom path. */ + from_canonpath = svn_fs__canonicalize_abspath(from_path, pool); + + SVN_ERR(svn_fs_fs__dag_copy(to_parent_path->parent->node, + to_parent_path->entry, + from_node, + preserve_history, + from_root->rev, + from_canonpath, + txn_id, pool)); + + if (kind == svn_fs_path_change_replace) + SVN_ERR(dag_node_cache_invalidate(to_root, + parent_path_path(to_parent_path, + pool), pool)); + + if (svn_fs_fs__fs_supports_mergeinfo(to_root->fs) + && mergeinfo_start != mergeinfo_end) + SVN_ERR(increment_mergeinfo_up_tree(to_parent_path->parent, + mergeinfo_end - mergeinfo_start, + pool)); + + /* Make a record of this modification in the changes table. */ + SVN_ERR(get_dag(&new_node, to_root, to_path, TRUE, pool)); + SVN_ERR(add_change(to_root->fs, txn_id, to_path, + svn_fs_fs__dag_get_id(new_node), kind, FALSE, FALSE, + svn_fs_fs__dag_node_kind(from_node), + from_root->rev, from_canonpath, pool)); + } + else + { + /* See IZ Issue #436 */ + /* Copying from transaction roots not currently available. + + ### cmpilato todo someday: make this not so. :-) Note that + when copying from mutable trees, you have to make sure that + you aren't creating a cyclic graph filesystem, and a simple + referencing operation won't cut it. Currently, we should not + be able to reach this clause, and the interface reports that + this only works from immutable trees anyway, but JimB has + stated that this requirement need not be necessary in the + future. */ + + SVN_ERR_MALFUNCTION(); + } + + return SVN_NO_ERROR; +} + + +/* Create a copy of FROM_PATH in FROM_ROOT named TO_PATH in TO_ROOT. + If FROM_PATH is a directory, copy it recursively. Temporary + allocations are from POOL.*/ +static svn_error_t * +fs_copy(svn_fs_root_t *from_root, + const char *from_path, + svn_fs_root_t *to_root, + const char *to_path, + apr_pool_t *pool) +{ + SVN_ERR(check_newline(to_path, pool)); + + return svn_error_trace(copy_helper(from_root, + svn_fs__canonicalize_abspath(from_path, + pool), + to_root, + svn_fs__canonicalize_abspath(to_path, + pool), + TRUE, pool)); +} + + +/* Create a copy of FROM_PATH in FROM_ROOT named TO_PATH in TO_ROOT. + If FROM_PATH is a directory, copy it recursively. No history is + preserved. Temporary allocations are from POOL. */ +static svn_error_t * +fs_revision_link(svn_fs_root_t *from_root, + svn_fs_root_t *to_root, + const char *path, + apr_pool_t *pool) +{ + if (! to_root->is_txn_root) + return SVN_FS__NOT_TXN(to_root); + + path = svn_fs__canonicalize_abspath(path, pool); + return svn_error_trace(copy_helper(from_root, path, to_root, path, + FALSE, pool)); +} + + +/* Discover the copy ancestry of PATH under ROOT. Return a relevant + ancestor/revision combination in *PATH_P and *REV_P. Temporary + allocations are in POOL. */ +static svn_error_t * +fs_copied_from(svn_revnum_t *rev_p, + const char **path_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *node; + const char *copyfrom_path, *copyfrom_str = NULL; + svn_revnum_t copyfrom_rev; + char *str, *buf; + + /* Check to see if there is a cached version of this copyfrom + entry. */ + if (! root->is_txn_root) { + fs_rev_root_data_t *frd = root->fsap_data; + copyfrom_str = svn_hash_gets(frd->copyfrom_cache, path); + } + + if (copyfrom_str) + { + if (*copyfrom_str == 0) + { + /* We have a cached entry that says there is no copyfrom + here. */ + copyfrom_rev = SVN_INVALID_REVNUM; + copyfrom_path = NULL; + } + else + { + /* Parse the copyfrom string for our cached entry. */ + buf = apr_pstrdup(pool, copyfrom_str); + str = svn_cstring_tokenize(" ", &buf); + copyfrom_rev = SVN_STR_TO_REV(str); + copyfrom_path = buf; + } + } + else + { + /* There is no cached entry, look it up the old-fashioned + way. */ + SVN_ERR(get_dag(&node, root, path, TRUE, pool)); + SVN_ERR(svn_fs_fs__dag_get_copyfrom_rev(©from_rev, node)); + SVN_ERR(svn_fs_fs__dag_get_copyfrom_path(©from_path, node)); + } + + *rev_p = copyfrom_rev; + *path_p = copyfrom_path; + + return SVN_NO_ERROR; +} + + + +/* Files. */ + +/* Create the empty file PATH under ROOT. Temporary allocations are + in POOL. */ +static svn_error_t * +fs_make_file(svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + parent_path_t *parent_path; + dag_node_t *child; + const char *txn_id = root->txn; + + SVN_ERR(check_newline(path, pool)); + + path = svn_fs__canonicalize_abspath(path, pool); + SVN_ERR(open_path(&parent_path, root, path, open_path_last_optional, + txn_id, pool)); + + /* If there's already a file by that name, complain. + This also catches the case of trying to make a file named `/'. */ + if (parent_path->node) + return SVN_FS__ALREADY_EXISTS(root, path); + + /* Check (non-recursively) to see if path is locked; if so, check + that we can use it. */ + if (root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(path, root->fs, FALSE, FALSE, + pool)); + + /* Create the file. */ + SVN_ERR(make_path_mutable(root, parent_path->parent, path, pool)); + SVN_ERR(svn_fs_fs__dag_make_file(&child, + parent_path->parent->node, + parent_path_path(parent_path->parent, + pool), + parent_path->entry, + txn_id, + pool)); + + /* Add this file to the path cache. */ + SVN_ERR(dag_node_cache_set(root, parent_path_path(parent_path, pool), child, + pool)); + + /* Make a record of this modification in the changes table. */ + return add_change(root->fs, txn_id, path, svn_fs_fs__dag_get_id(child), + svn_fs_path_change_add, TRUE, FALSE, svn_node_file, + SVN_INVALID_REVNUM, NULL, pool); +} + + +/* Set *LENGTH_P to the size of the file PATH under ROOT. Temporary + allocations are in POOL. */ +static svn_error_t * +fs_file_length(svn_filesize_t *length_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *file; + + /* First create a dag_node_t from the root/path pair. */ + SVN_ERR(get_dag(&file, root, path, FALSE, pool)); + + /* Now fetch its length */ + return svn_fs_fs__dag_file_length(length_p, file, pool); +} + + +/* Set *CHECKSUM to the checksum of type KIND for PATH under ROOT, or + NULL if that information isn't available. Temporary allocations + are from POOL. */ +static svn_error_t * +fs_file_checksum(svn_checksum_t **checksum, + svn_checksum_kind_t kind, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *file; + + SVN_ERR(get_dag(&file, root, path, FALSE, pool)); + return svn_fs_fs__dag_file_checksum(checksum, file, kind, pool); +} + + +/* --- Machinery for svn_fs_file_contents() --- */ + +/* Set *CONTENTS to a readable stream that will return the contents of + PATH under ROOT. The stream is allocated in POOL. */ +static svn_error_t * +fs_file_contents(svn_stream_t **contents, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + dag_node_t *node; + svn_stream_t *file_stream; + + /* First create a dag_node_t from the root/path pair. */ + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + + /* Then create a readable stream from the dag_node_t. */ + SVN_ERR(svn_fs_fs__dag_get_contents(&file_stream, node, pool)); + + *contents = file_stream; + return SVN_NO_ERROR; +} + +/* --- End machinery for svn_fs_file_contents() --- */ + + +/* --- Machinery for svn_fs_try_process_file_contents() --- */ + +static svn_error_t * +fs_try_process_file_contents(svn_boolean_t *success, + svn_fs_root_t *root, + const char *path, + svn_fs_process_contents_func_t processor, + void* baton, + apr_pool_t *pool) +{ + dag_node_t *node; + SVN_ERR(get_dag(&node, root, path, FALSE, pool)); + + return svn_fs_fs__dag_try_process_file_contents(success, node, + processor, baton, pool); +} + +/* --- End machinery for svn_fs_try_process_file_contents() --- */ + + +/* --- Machinery for svn_fs_apply_textdelta() --- */ + + +/* Local baton type for all the helper functions below. */ +typedef struct txdelta_baton_t +{ + /* This is the custom-built window consumer given to us by the delta + library; it uniquely knows how to read data from our designated + "source" stream, interpret the window, and write data to our + designated "target" stream (in this case, our repos file.) */ + svn_txdelta_window_handler_t interpreter; + void *interpreter_baton; + + /* The original file info */ + svn_fs_root_t *root; + const char *path; + + /* Derived from the file info */ + dag_node_t *node; + + svn_stream_t *source_stream; + svn_stream_t *target_stream; + svn_stream_t *string_stream; + svn_stringbuf_t *target_string; + + /* MD5 digest for the base text against which a delta is to be + applied, and for the resultant fulltext, respectively. Either or + both may be null, in which case ignored. */ + svn_checksum_t *base_checksum; + svn_checksum_t *result_checksum; + + /* Pool used by db txns */ + apr_pool_t *pool; + +} txdelta_baton_t; + + +/* ### see comment in window_consumer() regarding this function. */ + +/* Helper function of generic type `svn_write_fn_t'. Implements a + writable stream which appends to an svn_stringbuf_t. */ +static svn_error_t * +write_to_string(void *baton, const char *data, apr_size_t *len) +{ + txdelta_baton_t *tb = (txdelta_baton_t *) baton; + svn_stringbuf_appendbytes(tb->target_string, data, *len); + return SVN_NO_ERROR; +} + + + +/* The main window handler returned by svn_fs_apply_textdelta. */ +static svn_error_t * +window_consumer(svn_txdelta_window_t *window, void *baton) +{ + txdelta_baton_t *tb = (txdelta_baton_t *) baton; + + /* Send the window right through to the custom window interpreter. + In theory, the interpreter will then write more data to + cb->target_string. */ + SVN_ERR(tb->interpreter(window, tb->interpreter_baton)); + + /* ### the write_to_string() callback for the txdelta's output stream + ### should be doing all the flush determination logic, not here. + ### in a drastic case, a window could generate a LOT more than the + ### maximum buffer size. we want to flush to the underlying target + ### stream much sooner (e.g. also in a streamy fashion). also, by + ### moving this logic inside the stream, the stream becomes nice + ### and encapsulated: it holds all the logic about buffering and + ### flushing. + ### + ### further: I believe the buffering should be removed from tree.c + ### the buffering should go into the target_stream itself, which + ### is defined by reps-string.c. Specifically, I think the + ### rep_write_contents() function will handle the buffering and + ### the spill to the underlying DB. by locating it there, then + ### anybody who gets a writable stream for FS content can take + ### advantage of the buffering capability. this will be important + ### when we export an FS API function for writing a fulltext into + ### the FS, rather than forcing that fulltext thru apply_textdelta. + */ + + /* Check to see if we need to purge the portion of the contents that + have been written thus far. */ + if ((! window) || (tb->target_string->len > WRITE_BUFFER_SIZE)) + { + apr_size_t len = tb->target_string->len; + SVN_ERR(svn_stream_write(tb->target_stream, + tb->target_string->data, + &len)); + svn_stringbuf_setempty(tb->target_string); + } + + /* Is the window NULL? If so, we're done. */ + if (! window) + { + /* Close the internal-use stream. ### This used to be inside of + txn_body_fulltext_finalize_edits(), but that invoked a nested + Berkeley DB transaction -- scandalous! */ + SVN_ERR(svn_stream_close(tb->target_stream)); + + SVN_ERR(svn_fs_fs__dag_finalize_edits(tb->node, tb->result_checksum, + tb->pool)); + } + + return SVN_NO_ERROR; +} + +/* Helper function for fs_apply_textdelta. BATON is of type + txdelta_baton_t. */ +static svn_error_t * +apply_textdelta(void *baton, apr_pool_t *pool) +{ + txdelta_baton_t *tb = (txdelta_baton_t *) baton; + parent_path_t *parent_path; + const char *txn_id = tb->root->txn; + + /* Call open_path with no flags, as we want this to return an error + if the node for which we are searching doesn't exist. */ + SVN_ERR(open_path(&parent_path, tb->root, tb->path, 0, txn_id, pool)); + + /* Check (non-recursively) to see if path is locked; if so, check + that we can use it. */ + if (tb->root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(tb->path, tb->root->fs, + FALSE, FALSE, pool)); + + /* Now, make sure this path is mutable. */ + SVN_ERR(make_path_mutable(tb->root, parent_path, tb->path, pool)); + tb->node = parent_path->node; + + if (tb->base_checksum) + { + svn_checksum_t *checksum; + + /* Until we finalize the node, its data_key points to the old + contents, in other words, the base text. */ + SVN_ERR(svn_fs_fs__dag_file_checksum(&checksum, tb->node, + tb->base_checksum->kind, pool)); + if (!svn_checksum_match(tb->base_checksum, checksum)) + return svn_checksum_mismatch_err(tb->base_checksum, checksum, pool, + _("Base checksum mismatch on '%s'"), + tb->path); + } + + /* Make a readable "source" stream out of the current contents of + ROOT/PATH; obviously, this must done in the context of a db_txn. + The stream is returned in tb->source_stream. */ + SVN_ERR(svn_fs_fs__dag_get_contents(&(tb->source_stream), + tb->node, tb->pool)); + + /* Make a writable "target" stream */ + SVN_ERR(svn_fs_fs__dag_get_edit_stream(&(tb->target_stream), tb->node, + tb->pool)); + + /* Make a writable "string" stream which writes data to + tb->target_string. */ + tb->target_string = svn_stringbuf_create_empty(tb->pool); + tb->string_stream = svn_stream_create(tb, tb->pool); + svn_stream_set_write(tb->string_stream, write_to_string); + + /* Now, create a custom window handler that uses our two streams. */ + svn_txdelta_apply(tb->source_stream, + tb->string_stream, + NULL, + tb->path, + tb->pool, + &(tb->interpreter), + &(tb->interpreter_baton)); + + /* Make a record of this modification in the changes table. */ + return add_change(tb->root->fs, txn_id, tb->path, + svn_fs_fs__dag_get_id(tb->node), + svn_fs_path_change_modify, TRUE, FALSE, svn_node_file, + SVN_INVALID_REVNUM, NULL, pool); +} + + +/* Set *CONTENTS_P and *CONTENTS_BATON_P to a window handler and baton + that will accept text delta windows to modify the contents of PATH + under ROOT. Allocations are in POOL. */ +static svn_error_t * +fs_apply_textdelta(svn_txdelta_window_handler_t *contents_p, + void **contents_baton_p, + svn_fs_root_t *root, + const char *path, + svn_checksum_t *base_checksum, + svn_checksum_t *result_checksum, + apr_pool_t *pool) +{ + txdelta_baton_t *tb = apr_pcalloc(pool, sizeof(*tb)); + + tb->root = root; + tb->path = svn_fs__canonicalize_abspath(path, pool); + tb->pool = pool; + tb->base_checksum = svn_checksum_dup(base_checksum, pool); + tb->result_checksum = svn_checksum_dup(result_checksum, pool); + + SVN_ERR(apply_textdelta(tb, pool)); + + *contents_p = window_consumer; + *contents_baton_p = tb; + return SVN_NO_ERROR; +} + +/* --- End machinery for svn_fs_apply_textdelta() --- */ + +/* --- Machinery for svn_fs_apply_text() --- */ + +/* Baton for svn_fs_apply_text(). */ +struct text_baton_t +{ + /* The original file info */ + svn_fs_root_t *root; + const char *path; + + /* Derived from the file info */ + dag_node_t *node; + + /* The returned stream that will accept the file's new contents. */ + svn_stream_t *stream; + + /* The actual fs stream that the returned stream will write to. */ + svn_stream_t *file_stream; + + /* MD5 digest for the final fulltext written to the file. May + be null, in which case ignored. */ + svn_checksum_t *result_checksum; + + /* Pool used by db txns */ + apr_pool_t *pool; +}; + + +/* A wrapper around svn_fs_fs__dag_finalize_edits, but for + * fulltext data, not text deltas. Closes BATON->file_stream. + * + * Note: If you're confused about how this function relates to another + * of similar name, think of it this way: + * + * svn_fs_apply_textdelta() ==> ... ==> txn_body_txdelta_finalize_edits() + * svn_fs_apply_text() ==> ... ==> txn_body_fulltext_finalize_edits() + */ + +/* Write function for the publically returned stream. */ +static svn_error_t * +text_stream_writer(void *baton, + const char *data, + apr_size_t *len) +{ + struct text_baton_t *tb = baton; + + /* Psst, here's some data. Pass it on to the -real- file stream. */ + return svn_stream_write(tb->file_stream, data, len); +} + +/* Close function for the publically returned stream. */ +static svn_error_t * +text_stream_closer(void *baton) +{ + struct text_baton_t *tb = baton; + + /* Close the internal-use stream. ### This used to be inside of + txn_body_fulltext_finalize_edits(), but that invoked a nested + Berkeley DB transaction -- scandalous! */ + SVN_ERR(svn_stream_close(tb->file_stream)); + + /* Need to tell fs that we're done sending text */ + return svn_fs_fs__dag_finalize_edits(tb->node, tb->result_checksum, + tb->pool); +} + + +/* Helper function for fs_apply_text. BATON is of type + text_baton_t. */ +static svn_error_t * +apply_text(void *baton, apr_pool_t *pool) +{ + struct text_baton_t *tb = baton; + parent_path_t *parent_path; + const char *txn_id = tb->root->txn; + + /* Call open_path with no flags, as we want this to return an error + if the node for which we are searching doesn't exist. */ + SVN_ERR(open_path(&parent_path, tb->root, tb->path, 0, txn_id, pool)); + + /* Check (non-recursively) to see if path is locked; if so, check + that we can use it. */ + if (tb->root->txn_flags & SVN_FS_TXN_CHECK_LOCKS) + SVN_ERR(svn_fs_fs__allow_locked_operation(tb->path, tb->root->fs, + FALSE, FALSE, pool)); + + /* Now, make sure this path is mutable. */ + SVN_ERR(make_path_mutable(tb->root, parent_path, tb->path, pool)); + tb->node = parent_path->node; + + /* Make a writable stream for replacing the file's text. */ + SVN_ERR(svn_fs_fs__dag_get_edit_stream(&(tb->file_stream), tb->node, + tb->pool)); + + /* Create a 'returnable' stream which writes to the file_stream. */ + tb->stream = svn_stream_create(tb, tb->pool); + svn_stream_set_write(tb->stream, text_stream_writer); + svn_stream_set_close(tb->stream, text_stream_closer); + + /* Make a record of this modification in the changes table. */ + return add_change(tb->root->fs, txn_id, tb->path, + svn_fs_fs__dag_get_id(tb->node), + svn_fs_path_change_modify, TRUE, FALSE, svn_node_file, + SVN_INVALID_REVNUM, NULL, pool); +} + + +/* Return a writable stream that will set the contents of PATH under + ROOT. RESULT_CHECKSUM is the MD5 checksum of the final result. + Temporary allocations are in POOL. */ +static svn_error_t * +fs_apply_text(svn_stream_t **contents_p, + svn_fs_root_t *root, + const char *path, + svn_checksum_t *result_checksum, + apr_pool_t *pool) +{ + struct text_baton_t *tb = apr_pcalloc(pool, sizeof(*tb)); + + tb->root = root; + tb->path = svn_fs__canonicalize_abspath(path, pool); + tb->pool = pool; + tb->result_checksum = svn_checksum_dup(result_checksum, pool); + + SVN_ERR(apply_text(tb, pool)); + + *contents_p = tb->stream; + return SVN_NO_ERROR; +} + +/* --- End machinery for svn_fs_apply_text() --- */ + + +/* Check if the contents of PATH1 under ROOT1 are different from the + contents of PATH2 under ROOT2. If they are different set + *CHANGED_P to TRUE, otherwise set it to FALSE. */ +static svn_error_t * +fs_contents_changed(svn_boolean_t *changed_p, + svn_fs_root_t *root1, + const char *path1, + svn_fs_root_t *root2, + const char *path2, + apr_pool_t *pool) +{ + dag_node_t *node1, *node2; + + /* Check that roots are in the same fs. */ + if (root1->fs != root2->fs) + return svn_error_create + (SVN_ERR_FS_GENERAL, NULL, + _("Cannot compare file contents between two different filesystems")); + + /* Check that both paths are files. */ + { + svn_node_kind_t kind; + + SVN_ERR(svn_fs_fs__check_path(&kind, root1, path1, pool)); + if (kind != svn_node_file) + return svn_error_createf + (SVN_ERR_FS_GENERAL, NULL, _("'%s' is not a file"), path1); + + SVN_ERR(svn_fs_fs__check_path(&kind, root2, path2, pool)); + if (kind != svn_node_file) + return svn_error_createf + (SVN_ERR_FS_GENERAL, NULL, _("'%s' is not a file"), path2); + } + + SVN_ERR(get_dag(&node1, root1, path1, TRUE, pool)); + SVN_ERR(get_dag(&node2, root2, path2, TRUE, pool)); + return svn_fs_fs__dag_things_different(NULL, changed_p, + node1, node2); +} + + + +/* Public interface to computing file text deltas. */ + +static svn_error_t * +fs_get_file_delta_stream(svn_txdelta_stream_t **stream_p, + svn_fs_root_t *source_root, + const char *source_path, + svn_fs_root_t *target_root, + const char *target_path, + apr_pool_t *pool) +{ + dag_node_t *source_node, *target_node; + + if (source_root && source_path) + SVN_ERR(get_dag(&source_node, source_root, source_path, TRUE, pool)); + else + source_node = NULL; + SVN_ERR(get_dag(&target_node, target_root, target_path, TRUE, pool)); + + /* Create a delta stream that turns the source into the target. */ + return svn_fs_fs__dag_get_file_delta_stream(stream_p, source_node, + target_node, pool); +} + + + +/* Finding Changes */ + +/* Set *CHANGED_PATHS_P to a newly allocated hash containing + descriptions of the paths changed under ROOT. The hash is keyed + with const char * paths and has svn_fs_path_change2_t * values. Use + POOL for all allocations. */ +static svn_error_t * +fs_paths_changed(apr_hash_t **changed_paths_p, + svn_fs_root_t *root, + apr_pool_t *pool) +{ + if (root->is_txn_root) + return svn_fs_fs__txn_changes_fetch(changed_paths_p, root->fs, root->txn, + pool); + else + { + fs_rev_root_data_t *frd = root->fsap_data; + return svn_fs_fs__paths_changed(changed_paths_p, root->fs, root->rev, + frd->copyfrom_cache, pool); + } +} + + + +/* Our coolio opaque history object. */ +typedef struct fs_history_data_t +{ + /* filesystem object */ + svn_fs_t *fs; + + /* path and revision of historical location */ + const char *path; + svn_revnum_t revision; + + /* internal-use hints about where to resume the history search. */ + const char *path_hint; + svn_revnum_t rev_hint; + + /* FALSE until the first call to svn_fs_history_prev(). */ + svn_boolean_t is_interesting; +} fs_history_data_t; + +static svn_fs_history_t * +assemble_history(svn_fs_t *fs, + const char *path, + svn_revnum_t revision, + svn_boolean_t is_interesting, + const char *path_hint, + svn_revnum_t rev_hint, + apr_pool_t *pool); + + +/* Set *HISTORY_P to an opaque node history object which represents + PATH under ROOT. ROOT must be a revision root. Use POOL for all + allocations. */ +static svn_error_t * +fs_node_history(svn_fs_history_t **history_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + svn_node_kind_t kind; + + /* We require a revision root. */ + if (root->is_txn_root) + return svn_error_create(SVN_ERR_FS_NOT_REVISION_ROOT, NULL, NULL); + + /* And we require that the path exist in the root. */ + SVN_ERR(svn_fs_fs__check_path(&kind, root, path, pool)); + if (kind == svn_node_none) + return SVN_FS__NOT_FOUND(root, path); + + /* Okay, all seems well. Build our history object and return it. */ + *history_p = assemble_history(root->fs, + svn_fs__canonicalize_abspath(path, pool), + root->rev, FALSE, NULL, + SVN_INVALID_REVNUM, pool); + return SVN_NO_ERROR; +} + +/* Find the youngest copyroot for path PARENT_PATH or its parents in + filesystem FS, and store the copyroot in *REV_P and *PATH_P. + Perform all allocations in POOL. */ +static svn_error_t * +find_youngest_copyroot(svn_revnum_t *rev_p, + const char **path_p, + svn_fs_t *fs, + parent_path_t *parent_path, + apr_pool_t *pool) +{ + svn_revnum_t rev_mine; + svn_revnum_t rev_parent = SVN_INVALID_REVNUM; + const char *path_mine; + const char *path_parent = NULL; + + /* First find our parent's youngest copyroot. */ + if (parent_path->parent) + SVN_ERR(find_youngest_copyroot(&rev_parent, &path_parent, fs, + parent_path->parent, pool)); + + /* Find our copyroot. */ + SVN_ERR(svn_fs_fs__dag_get_copyroot(&rev_mine, &path_mine, + parent_path->node)); + + /* If a parent and child were copied to in the same revision, prefer + the child copy target, since it is the copy relevant to the + history of the child. */ + if (rev_mine >= rev_parent) + { + *rev_p = rev_mine; + *path_p = path_mine; + } + else + { + *rev_p = rev_parent; + *path_p = path_parent; + } + + return SVN_NO_ERROR; +} + + +static svn_error_t *fs_closest_copy(svn_fs_root_t **root_p, + const char **path_p, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + svn_fs_t *fs = root->fs; + parent_path_t *parent_path, *copy_dst_parent_path; + svn_revnum_t copy_dst_rev, created_rev; + const char *copy_dst_path; + svn_fs_root_t *copy_dst_root; + dag_node_t *copy_dst_node; + svn_node_kind_t kind; + + /* Initialize return values. */ + *root_p = NULL; + *path_p = NULL; + + path = svn_fs__canonicalize_abspath(path, pool); + SVN_ERR(open_path(&parent_path, root, path, 0, NULL, pool)); + + /* Find the youngest copyroot in the path of this node-rev, which + will indicate the target of the innermost copy affecting the + node-rev. */ + SVN_ERR(find_youngest_copyroot(©_dst_rev, ©_dst_path, + fs, parent_path, pool)); + if (copy_dst_rev == 0) /* There are no copies affecting this node-rev. */ + return SVN_NO_ERROR; + + /* It is possible that this node was created from scratch at some + revision between COPY_DST_REV and REV. Make sure that PATH + exists as of COPY_DST_REV and is related to this node-rev. */ + SVN_ERR(svn_fs_fs__revision_root(©_dst_root, fs, copy_dst_rev, pool)); + SVN_ERR(svn_fs_fs__check_path(&kind, copy_dst_root, path, pool)); + if (kind == svn_node_none) + return SVN_NO_ERROR; + SVN_ERR(open_path(©_dst_parent_path, copy_dst_root, path, + open_path_node_only, NULL, pool)); + copy_dst_node = copy_dst_parent_path->node; + if (! svn_fs_fs__id_check_related(svn_fs_fs__dag_get_id(copy_dst_node), + svn_fs_fs__dag_get_id(parent_path->node))) + return SVN_NO_ERROR; + + /* One final check must be done here. If you copy a directory and + create a new entity somewhere beneath that directory in the same + txn, then we can't claim that the copy affected the new entity. + For example, if you do: + + copy dir1 dir2 + create dir2/new-thing + commit + + then dir2/new-thing was not affected by the copy of dir1 to dir2. + We detect this situation by asking if PATH@COPY_DST_REV's + created-rev is COPY_DST_REV, and that node-revision has no + predecessors, then there is no relevant closest copy. + */ + SVN_ERR(svn_fs_fs__dag_get_revision(&created_rev, copy_dst_node, pool)); + if (created_rev == copy_dst_rev) + { + const svn_fs_id_t *pred; + SVN_ERR(svn_fs_fs__dag_get_predecessor_id(&pred, copy_dst_node)); + if (! pred) + return SVN_NO_ERROR; + } + + /* The copy destination checks out. Return it. */ + *root_p = copy_dst_root; + *path_p = copy_dst_path; + return SVN_NO_ERROR; +} + + +/* Set *PREV_PATH and *PREV_REV to the path and revision which + represent the location at which PATH in FS was located immediately + prior to REVISION iff there was a copy operation (to PATH or one of + its parent directories) between that previous location and + PATH@REVISION. + + If there was no such copy operation in that portion of PATH's + history, set *PREV_PATH to NULL and *PREV_REV to SVN_INVALID_REVNUM. */ +static svn_error_t * +prev_location(const char **prev_path, + svn_revnum_t *prev_rev, + svn_fs_t *fs, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + const char *copy_path, *copy_src_path, *remainder_path; + svn_fs_root_t *copy_root; + svn_revnum_t copy_src_rev; + + /* Ask about the most recent copy which affected PATH@REVISION. If + there was no such copy, we're done. */ + SVN_ERR(fs_closest_copy(©_root, ©_path, root, path, pool)); + if (! copy_root) + { + *prev_rev = SVN_INVALID_REVNUM; + *prev_path = NULL; + return SVN_NO_ERROR; + } + + /* Ultimately, it's not the path of the closest copy's source that + we care about -- it's our own path's location in the copy source + revision. So we'll tack the relative path that expresses the + difference between the copy destination and our path in the copy + revision onto the copy source path to determine this information. + + In other words, if our path is "/branches/my-branch/foo/bar", and + we know that the closest relevant copy was a copy of "/trunk" to + "/branches/my-branch", then that relative path under the copy + destination is "/foo/bar". Tacking that onto the copy source + path tells us that our path was located at "/trunk/foo/bar" + before the copy. + */ + SVN_ERR(fs_copied_from(©_src_rev, ©_src_path, + copy_root, copy_path, pool)); + remainder_path = svn_fspath__skip_ancestor(copy_path, path); + *prev_path = svn_fspath__join(copy_src_path, remainder_path, pool); + *prev_rev = copy_src_rev; + return SVN_NO_ERROR; +} + + +static svn_error_t * +fs_node_origin_rev(svn_revnum_t *revision, + svn_fs_root_t *root, + const char *path, + apr_pool_t *pool) +{ + svn_fs_t *fs = root->fs; + const svn_fs_id_t *given_noderev_id, *cached_origin_id; + const char *node_id, *dash; + + path = svn_fs__canonicalize_abspath(path, pool); + + /* Check the cache first. */ + SVN_ERR(svn_fs_fs__node_id(&given_noderev_id, root, path, pool)); + node_id = svn_fs_fs__id_node_id(given_noderev_id); + + /* Is it a brand new uncommitted node? */ + if (node_id[0] == '_') + { + *revision = SVN_INVALID_REVNUM; + return SVN_NO_ERROR; + } + + /* Maybe this is a new-style node ID that just has the revision + sitting right in it. */ + dash = strchr(node_id, '-'); + if (dash && *(dash+1)) + { + *revision = SVN_STR_TO_REV(dash + 1); + return SVN_NO_ERROR; + } + + /* OK, it's an old-style ID? Maybe it's cached. */ + SVN_ERR(svn_fs_fs__get_node_origin(&cached_origin_id, + fs, + node_id, + pool)); + if (cached_origin_id != NULL) + { + *revision = svn_fs_fs__id_rev(cached_origin_id); + return SVN_NO_ERROR; + } + + { + /* Ah well, the answer isn't in the ID itself or in the cache. + Let's actually calculate it, then. */ + svn_fs_root_t *curroot = root; + apr_pool_t *subpool = svn_pool_create(pool); + apr_pool_t *predidpool = svn_pool_create(pool); + svn_stringbuf_t *lastpath = svn_stringbuf_create(path, pool); + svn_revnum_t lastrev = SVN_INVALID_REVNUM; + dag_node_t *node; + const svn_fs_id_t *pred_id; + + /* Walk the closest-copy chain back to the first copy in our history. + + NOTE: We merely *assume* that this is faster than walking the + predecessor chain, because we *assume* that copies of parent + directories happen less often than modifications to a given item. */ + while (1) + { + svn_revnum_t currev; + const char *curpath = lastpath->data; + + svn_pool_clear(subpool); + + /* Get a root pointing to LASTREV. (The first time around, + LASTREV is invalid, but that's cool because CURROOT is + already initialized.) */ + if (SVN_IS_VALID_REVNUM(lastrev)) + SVN_ERR(svn_fs_fs__revision_root(&curroot, fs, lastrev, subpool)); + + /* Find the previous location using the closest-copy shortcut. */ + SVN_ERR(prev_location(&curpath, &currev, fs, curroot, curpath, + subpool)); + if (! curpath) + break; + + /* Update our LASTPATH and LASTREV variables (which survive + SUBPOOL). */ + svn_stringbuf_set(lastpath, curpath); + lastrev = currev; + } + + /* Walk the predecessor links back to origin. */ + SVN_ERR(svn_fs_fs__node_id(&pred_id, curroot, lastpath->data, predidpool)); + do + { + svn_pool_clear(subpool); + SVN_ERR(svn_fs_fs__dag_get_node(&node, fs, pred_id, subpool)); + + /* Why not just fetch the predecessor ID in PREDIDPOOL? + Because svn_fs_fs__dag_get_predecessor_id() doesn't + necessarily honor the passed-in pool, and might return a + value cached in the node (which is allocated in + SUBPOOL... maybe). */ + svn_pool_clear(predidpool); + SVN_ERR(svn_fs_fs__dag_get_predecessor_id(&pred_id, node)); + pred_id = pred_id ? svn_fs_fs__id_copy(pred_id, predidpool) : NULL; + } + while (pred_id); + + /* When we get here, NODE should be the first node-revision in our + chain. */ + SVN_ERR(svn_fs_fs__dag_get_revision(revision, node, pool)); + + /* Wow, I don't want to have to do all that again. Let's cache + the result. */ + if (node_id[0] != '_') + SVN_ERR(svn_fs_fs__set_node_origin(fs, node_id, + svn_fs_fs__dag_get_id(node), pool)); + + svn_pool_destroy(subpool); + svn_pool_destroy(predidpool); + return SVN_NO_ERROR; + } +} + + +struct history_prev_args +{ + svn_fs_history_t **prev_history_p; + svn_fs_history_t *history; + svn_boolean_t cross_copies; + apr_pool_t *pool; +}; + + +static svn_error_t * +history_prev(void *baton, apr_pool_t *pool) +{ + struct history_prev_args *args = baton; + svn_fs_history_t **prev_history = args->prev_history_p; + svn_fs_history_t *history = args->history; + fs_history_data_t *fhd = history->fsap_data; + const char *commit_path, *src_path, *path = fhd->path; + svn_revnum_t commit_rev, src_rev, dst_rev; + svn_revnum_t revision = fhd->revision; + apr_pool_t *retpool = args->pool; + svn_fs_t *fs = fhd->fs; + parent_path_t *parent_path; + dag_node_t *node; + svn_fs_root_t *root; + svn_boolean_t reported = fhd->is_interesting; + svn_revnum_t copyroot_rev; + const char *copyroot_path; + + /* Initialize our return value. */ + *prev_history = NULL; + + /* If our last history report left us hints about where to pickup + the chase, then our last report was on the destination of a + copy. If we are crossing copies, start from those locations, + otherwise, we're all done here. */ + if (fhd->path_hint && SVN_IS_VALID_REVNUM(fhd->rev_hint)) + { + reported = FALSE; + if (! args->cross_copies) + return SVN_NO_ERROR; + path = fhd->path_hint; + revision = fhd->rev_hint; + } + + /* Construct a ROOT for the current revision. */ + SVN_ERR(svn_fs_fs__revision_root(&root, fs, revision, pool)); + + /* Open PATH/REVISION, and get its node and a bunch of other + goodies. */ + SVN_ERR(open_path(&parent_path, root, path, 0, NULL, pool)); + node = parent_path->node; + commit_path = svn_fs_fs__dag_get_created_path(node); + SVN_ERR(svn_fs_fs__dag_get_revision(&commit_rev, node, pool)); + + /* The Subversion filesystem is written in such a way that a given + line of history may have at most one interesting history point + per filesystem revision. Either that node was edited (and + possibly copied), or it was copied but not edited. And a copy + source cannot be from the same revision as its destination. So, + if our history revision matches its node's commit revision, we + know that ... */ + if (revision == commit_rev) + { + if (! reported) + { + /* ... we either have not yet reported on this revision (and + need now to do so) ... */ + *prev_history = assemble_history(fs, + apr_pstrdup(retpool, commit_path), + commit_rev, TRUE, NULL, + SVN_INVALID_REVNUM, retpool); + return SVN_NO_ERROR; + } + else + { + /* ... or we *have* reported on this revision, and must now + progress toward this node's predecessor (unless there is + no predecessor, in which case we're all done!). */ + const svn_fs_id_t *pred_id; + + SVN_ERR(svn_fs_fs__dag_get_predecessor_id(&pred_id, node)); + if (! pred_id) + return SVN_NO_ERROR; + + /* Replace NODE and friends with the information from its + predecessor. */ + SVN_ERR(svn_fs_fs__dag_get_node(&node, fs, pred_id, pool)); + commit_path = svn_fs_fs__dag_get_created_path(node); + SVN_ERR(svn_fs_fs__dag_get_revision(&commit_rev, node, pool)); + } + } + + /* Find the youngest copyroot in the path of this node, including + itself. */ + SVN_ERR(find_youngest_copyroot(©root_rev, ©root_path, fs, + parent_path, pool)); + + /* Initialize some state variables. */ + src_path = NULL; + src_rev = SVN_INVALID_REVNUM; + dst_rev = SVN_INVALID_REVNUM; + + if (copyroot_rev > commit_rev) + { + const char *remainder_path; + const char *copy_dst, *copy_src; + svn_fs_root_t *copyroot_root; + + SVN_ERR(svn_fs_fs__revision_root(©root_root, fs, copyroot_rev, + pool)); + SVN_ERR(get_dag(&node, copyroot_root, copyroot_path, FALSE, pool)); + copy_dst = svn_fs_fs__dag_get_created_path(node); + + /* If our current path was the very destination of the copy, + then our new current path will be the copy source. If our + current path was instead the *child* of the destination of + the copy, then figure out its previous location by taking its + path relative to the copy destination and appending that to + the copy source. Finally, if our current path doesn't meet + one of these other criteria ... ### for now just fallback to + the old copy hunt algorithm. */ + remainder_path = svn_fspath__skip_ancestor(copy_dst, path); + + if (remainder_path) + { + /* If we get here, then our current path is the destination + of, or the child of the destination of, a copy. Fill + in the return values and get outta here. */ + SVN_ERR(svn_fs_fs__dag_get_copyfrom_rev(&src_rev, node)); + SVN_ERR(svn_fs_fs__dag_get_copyfrom_path(©_src, node)); + + dst_rev = copyroot_rev; + src_path = svn_fspath__join(copy_src, remainder_path, pool); + } + } + + /* If we calculated a copy source path and revision, we'll make a + 'copy-style' history object. */ + if (src_path && SVN_IS_VALID_REVNUM(src_rev)) + { + svn_boolean_t retry = FALSE; + + /* It's possible for us to find a copy location that is the same + as the history point we've just reported. If that happens, + we simply need to take another trip through this history + search. */ + if ((dst_rev == revision) && reported) + retry = TRUE; + + *prev_history = assemble_history(fs, apr_pstrdup(retpool, path), + dst_rev, ! retry, + src_path, src_rev, retpool); + } + else + { + *prev_history = assemble_history(fs, apr_pstrdup(retpool, commit_path), + commit_rev, TRUE, NULL, + SVN_INVALID_REVNUM, retpool); + } + + return SVN_NO_ERROR; +} + + +/* Implement svn_fs_history_prev, set *PREV_HISTORY_P to a new + svn_fs_history_t object that represents the predecessory of + HISTORY. If CROSS_COPIES is true, *PREV_HISTORY_P may be related + only through a copy operation. Perform all allocations in POOL. */ +static svn_error_t * +fs_history_prev(svn_fs_history_t **prev_history_p, + svn_fs_history_t *history, + svn_boolean_t cross_copies, + apr_pool_t *pool) +{ + svn_fs_history_t *prev_history = NULL; + fs_history_data_t *fhd = history->fsap_data; + svn_fs_t *fs = fhd->fs; + + /* Special case: the root directory changes in every single + revision, no exceptions. And, the root can't be the target (or + child of a target -- duh) of a copy. So, if that's our path, + then we need only decrement our revision by 1, and there you go. */ + if (strcmp(fhd->path, "/") == 0) + { + if (! fhd->is_interesting) + prev_history = assemble_history(fs, "/", fhd->revision, + 1, NULL, SVN_INVALID_REVNUM, pool); + else if (fhd->revision > 0) + prev_history = assemble_history(fs, "/", fhd->revision - 1, + 1, NULL, SVN_INVALID_REVNUM, pool); + } + else + { + struct history_prev_args args; + prev_history = history; + + while (1) + { + args.prev_history_p = &prev_history; + args.history = prev_history; + args.cross_copies = cross_copies; + args.pool = pool; + SVN_ERR(history_prev(&args, pool)); + + if (! prev_history) + break; + fhd = prev_history->fsap_data; + if (fhd->is_interesting) + break; + } + } + + *prev_history_p = prev_history; + return SVN_NO_ERROR; +} + + +/* Set *PATH and *REVISION to the path and revision for the HISTORY + object. Use POOL for all allocations. */ +static svn_error_t * +fs_history_location(const char **path, + svn_revnum_t *revision, + svn_fs_history_t *history, + apr_pool_t *pool) +{ + fs_history_data_t *fhd = history->fsap_data; + + *path = apr_pstrdup(pool, fhd->path); + *revision = fhd->revision; + return SVN_NO_ERROR; +} + +static history_vtable_t history_vtable = { + fs_history_prev, + fs_history_location +}; + +/* Return a new history object (marked as "interesting") for PATH and + REVISION, allocated in POOL, and with its members set to the values + of the parameters provided. Note that PATH and PATH_HINT are not + duped into POOL -- it is the responsibility of the caller to ensure + that this happens. */ +static svn_fs_history_t * +assemble_history(svn_fs_t *fs, + const char *path, + svn_revnum_t revision, + svn_boolean_t is_interesting, + const char *path_hint, + svn_revnum_t rev_hint, + apr_pool_t *pool) +{ + svn_fs_history_t *history = apr_pcalloc(pool, sizeof(*history)); + fs_history_data_t *fhd = apr_pcalloc(pool, sizeof(*fhd)); + fhd->path = svn_fs__canonicalize_abspath(path, pool); + fhd->revision = revision; + fhd->is_interesting = is_interesting; + fhd->path_hint = path_hint; + fhd->rev_hint = rev_hint; + fhd->fs = fs; + + history->vtable = &history_vtable; + history->fsap_data = fhd; + return history; +} + + +/* mergeinfo queries */ + + +/* DIR_DAG is a directory DAG node which has mergeinfo in its + descendants. This function iterates over its children. For each + child with immediate mergeinfo, it adds its mergeinfo to + RESULT_CATALOG. appropriate arguments. For each child with + descendants with mergeinfo, it recurses. Note that it does *not* + call the action on the path for DIR_DAG itself. + + POOL is used for temporary allocations, including the mergeinfo + hashes passed to actions; RESULT_POOL is used for the mergeinfo added + to RESULT_CATALOG. + */ +static svn_error_t * +crawl_directory_dag_for_mergeinfo(svn_fs_root_t *root, + const char *this_path, + dag_node_t *dir_dag, + svn_mergeinfo_catalog_t result_catalog, + apr_pool_t *result_pool, + apr_pool_t *scratch_pool) +{ + apr_hash_t *entries; + apr_hash_index_t *hi; + apr_pool_t *iterpool = svn_pool_create(scratch_pool); + + SVN_ERR(svn_fs_fs__dag_dir_entries(&entries, dir_dag, + scratch_pool)); + + for (hi = apr_hash_first(scratch_pool, entries); + hi; + hi = apr_hash_next(hi)) + { + svn_fs_dirent_t *dirent = svn__apr_hash_index_val(hi); + const char *kid_path; + dag_node_t *kid_dag; + svn_boolean_t has_mergeinfo, go_down; + + svn_pool_clear(iterpool); + + kid_path = svn_fspath__join(this_path, dirent->name, iterpool); + SVN_ERR(get_dag(&kid_dag, root, kid_path, TRUE, iterpool)); + + SVN_ERR(svn_fs_fs__dag_has_mergeinfo(&has_mergeinfo, kid_dag)); + SVN_ERR(svn_fs_fs__dag_has_descendants_with_mergeinfo(&go_down, kid_dag)); + + if (has_mergeinfo) + { + /* Save this particular node's mergeinfo. */ + apr_hash_t *proplist; + svn_mergeinfo_t kid_mergeinfo; + svn_string_t *mergeinfo_string; + svn_error_t *err; + + SVN_ERR(svn_fs_fs__dag_get_proplist(&proplist, kid_dag, iterpool)); + mergeinfo_string = svn_hash_gets(proplist, SVN_PROP_MERGEINFO); + if (!mergeinfo_string) + { + svn_string_t *idstr = svn_fs_fs__id_unparse(dirent->id, iterpool); + return svn_error_createf + (SVN_ERR_FS_CORRUPT, NULL, + _("Node-revision #'%s' claims to have mergeinfo but doesn't"), + idstr->data); + } + + /* Issue #3896: If a node has syntactically invalid mergeinfo, then + treat it as if no mergeinfo is present rather than raising a parse + error. */ + err = svn_mergeinfo_parse(&kid_mergeinfo, + mergeinfo_string->data, + result_pool); + if (err) + { + if (err->apr_err == SVN_ERR_MERGEINFO_PARSE_ERROR) + svn_error_clear(err); + else + return svn_error_trace(err); + } + else + { + svn_hash_sets(result_catalog, apr_pstrdup(result_pool, kid_path), + kid_mergeinfo); + } + } + + if (go_down) + SVN_ERR(crawl_directory_dag_for_mergeinfo(root, + kid_path, + kid_dag, + result_catalog, + result_pool, + iterpool)); + } + + svn_pool_destroy(iterpool); + return SVN_NO_ERROR; +} + +/* Return the cache key as a combination of REV_ROOT->REV, the inheritance + flags INHERIT and ADJUST_INHERITED_MERGEINFO, and the PATH. The result + will be allocated in POOL.. + */ +static const char * +mergeinfo_cache_key(const char *path, + svn_fs_root_t *rev_root, + svn_mergeinfo_inheritance_t inherit, + svn_boolean_t adjust_inherited_mergeinfo, + apr_pool_t *pool) +{ + apr_int64_t number = rev_root->rev; + number = number * 4 + + (inherit == svn_mergeinfo_nearest_ancestor ? 2 : 0) + + (adjust_inherited_mergeinfo ? 1 : 0); + + return svn_fs_fs__combine_number_and_string(number, path, pool); +} + +/* Calculates the mergeinfo for PATH under REV_ROOT using inheritance + type INHERIT. Returns it in *MERGEINFO, or NULL if there is none. + The result is allocated in RESULT_POOL; SCRATCH_POOL is + used for temporary allocations. + */ +static svn_error_t * +get_mergeinfo_for_path_internal(svn_mergeinfo_t *mergeinfo, + svn_fs_root_t *rev_root, + const char *path, + svn_mergeinfo_inheritance_t inherit, + svn_boolean_t adjust_inherited_mergeinfo, + apr_pool_t *result_pool, + apr_pool_t *scratch_pool) +{ + parent_path_t *parent_path, *nearest_ancestor; + apr_hash_t *proplist; + svn_string_t *mergeinfo_string; + + path = svn_fs__canonicalize_abspath(path, scratch_pool); + + SVN_ERR(open_path(&parent_path, rev_root, path, 0, NULL, scratch_pool)); + + if (inherit == svn_mergeinfo_nearest_ancestor && ! parent_path->parent) + return SVN_NO_ERROR; + + if (inherit == svn_mergeinfo_nearest_ancestor) + nearest_ancestor = parent_path->parent; + else + nearest_ancestor = parent_path; + + while (TRUE) + { + svn_boolean_t has_mergeinfo; + + SVN_ERR(svn_fs_fs__dag_has_mergeinfo(&has_mergeinfo, + nearest_ancestor->node)); + if (has_mergeinfo) + break; + + /* No need to loop if we're looking for explicit mergeinfo. */ + if (inherit == svn_mergeinfo_explicit) + { + return SVN_NO_ERROR; + } + + nearest_ancestor = nearest_ancestor->parent; + + /* Run out? There's no mergeinfo. */ + if (!nearest_ancestor) + { + return SVN_NO_ERROR; + } + } + + SVN_ERR(svn_fs_fs__dag_get_proplist(&proplist, nearest_ancestor->node, + scratch_pool)); + mergeinfo_string = svn_hash_gets(proplist, SVN_PROP_MERGEINFO); + if (!mergeinfo_string) + return svn_error_createf + (SVN_ERR_FS_CORRUPT, NULL, + _("Node-revision '%s@%ld' claims to have mergeinfo but doesn't"), + parent_path_path(nearest_ancestor, scratch_pool), rev_root->rev); + + /* Parse the mergeinfo; store the result in *MERGEINFO. */ + { + /* Issue #3896: If a node has syntactically invalid mergeinfo, then + treat it as if no mergeinfo is present rather than raising a parse + error. */ + svn_error_t *err = svn_mergeinfo_parse(mergeinfo, + mergeinfo_string->data, + result_pool); + if (err) + { + if (err->apr_err == SVN_ERR_MERGEINFO_PARSE_ERROR) + { + svn_error_clear(err); + err = NULL; + *mergeinfo = NULL; + } + return svn_error_trace(err); + } + } + + /* If our nearest ancestor is the very path we inquired about, we + can return the mergeinfo results directly. Otherwise, we're + inheriting the mergeinfo, so we need to a) remove non-inheritable + ranges and b) telescope the merged-from paths. */ + if (adjust_inherited_mergeinfo && (nearest_ancestor != parent_path)) + { + svn_mergeinfo_t tmp_mergeinfo; + + SVN_ERR(svn_mergeinfo_inheritable2(&tmp_mergeinfo, *mergeinfo, + NULL, SVN_INVALID_REVNUM, + SVN_INVALID_REVNUM, TRUE, + scratch_pool, scratch_pool)); + SVN_ERR(svn_fs__append_to_merged_froms(mergeinfo, tmp_mergeinfo, + parent_path_relpath( + parent_path, nearest_ancestor, + scratch_pool), + result_pool)); + } + + return SVN_NO_ERROR; +} + +/* Caching wrapper around get_mergeinfo_for_path_internal(). + */ +static svn_error_t * +get_mergeinfo_for_path(svn_mergeinfo_t *mergeinfo, + svn_fs_root_t *rev_root, + const char *path, + svn_mergeinfo_inheritance_t inherit, + svn_boolean_t adjust_inherited_mergeinfo, + apr_pool_t *result_pool, + apr_pool_t *scratch_pool) +{ + fs_fs_data_t *ffd = rev_root->fs->fsap_data; + const char *cache_key; + svn_boolean_t found = FALSE; + svn_stringbuf_t *mergeinfo_exists; + + *mergeinfo = NULL; + + cache_key = mergeinfo_cache_key(path, rev_root, inherit, + adjust_inherited_mergeinfo, scratch_pool); + if (ffd->mergeinfo_existence_cache) + { + SVN_ERR(svn_cache__get((void **)&mergeinfo_exists, &found, + ffd->mergeinfo_existence_cache, + cache_key, result_pool)); + if (found && mergeinfo_exists->data[0] == '1') + SVN_ERR(svn_cache__get((void **)mergeinfo, &found, + ffd->mergeinfo_cache, + cache_key, result_pool)); + } + + if (! found) + { + SVN_ERR(get_mergeinfo_for_path_internal(mergeinfo, rev_root, path, + inherit, + adjust_inherited_mergeinfo, + result_pool, scratch_pool)); + if (ffd->mergeinfo_existence_cache) + { + mergeinfo_exists = svn_stringbuf_create(*mergeinfo ? "1" : "0", + scratch_pool); + SVN_ERR(svn_cache__set(ffd->mergeinfo_existence_cache, + cache_key, mergeinfo_exists, scratch_pool)); + if (*mergeinfo) + SVN_ERR(svn_cache__set(ffd->mergeinfo_cache, + cache_key, *mergeinfo, scratch_pool)); + } + } + + return SVN_NO_ERROR; +} + +/* Adds mergeinfo for each descendant of PATH (but not PATH itself) + under ROOT to RESULT_CATALOG. Returned values are allocated in + RESULT_POOL; temporary values in POOL. */ +static svn_error_t * +add_descendant_mergeinfo(svn_mergeinfo_catalog_t result_catalog, + svn_fs_root_t *root, + const char *path, + apr_pool_t *result_pool, + apr_pool_t *scratch_pool) +{ + dag_node_t *this_dag; + svn_boolean_t go_down; + + SVN_ERR(get_dag(&this_dag, root, path, TRUE, scratch_pool)); + SVN_ERR(svn_fs_fs__dag_has_descendants_with_mergeinfo(&go_down, + this_dag)); + if (go_down) + SVN_ERR(crawl_directory_dag_for_mergeinfo(root, + path, + this_dag, + result_catalog, + result_pool, + scratch_pool)); + return SVN_NO_ERROR; +} + + +/* Get the mergeinfo for a set of paths, returned in + *MERGEINFO_CATALOG. Returned values are allocated in + POOL, while temporary values are allocated in a sub-pool. */ +static svn_error_t * +get_mergeinfos_for_paths(svn_fs_root_t *root, + svn_mergeinfo_catalog_t *mergeinfo_catalog, + const apr_array_header_t *paths, + svn_mergeinfo_inheritance_t inherit, + svn_boolean_t include_descendants, + svn_boolean_t adjust_inherited_mergeinfo, + apr_pool_t *result_pool, + apr_pool_t *scratch_pool) +{ + svn_mergeinfo_catalog_t result_catalog = svn_hash__make(result_pool); + apr_pool_t *iterpool = svn_pool_create(scratch_pool); + int i; + + for (i = 0; i < paths->nelts; i++) + { + svn_error_t *err; + svn_mergeinfo_t path_mergeinfo; + const char *path = APR_ARRAY_IDX(paths, i, const char *); + + svn_pool_clear(iterpool); + + err = get_mergeinfo_for_path(&path_mergeinfo, root, path, + inherit, adjust_inherited_mergeinfo, + result_pool, iterpool); + if (err) + { + if (err->apr_err == SVN_ERR_MERGEINFO_PARSE_ERROR) + { + svn_error_clear(err); + err = NULL; + path_mergeinfo = NULL; + } + else + { + return svn_error_trace(err); + } + } + + if (path_mergeinfo) + svn_hash_sets(result_catalog, path, path_mergeinfo); + if (include_descendants) + SVN_ERR(add_descendant_mergeinfo(result_catalog, root, path, + result_pool, scratch_pool)); + } + svn_pool_destroy(iterpool); + + *mergeinfo_catalog = result_catalog; + return SVN_NO_ERROR; +} + + +/* Implements svn_fs_get_mergeinfo. */ +static svn_error_t * +fs_get_mergeinfo(svn_mergeinfo_catalog_t *catalog, + svn_fs_root_t *root, + const apr_array_header_t *paths, + svn_mergeinfo_inheritance_t inherit, + svn_boolean_t include_descendants, + svn_boolean_t adjust_inherited_mergeinfo, + apr_pool_t *result_pool, + apr_pool_t *scratch_pool) +{ + fs_fs_data_t *ffd = root->fs->fsap_data; + + /* We require a revision root. */ + if (root->is_txn_root) + return svn_error_create(SVN_ERR_FS_NOT_REVISION_ROOT, NULL, NULL); + + /* We have to actually be able to find the mergeinfo metadata! */ + if (! svn_fs_fs__fs_supports_mergeinfo(root->fs)) + return svn_error_createf + (SVN_ERR_UNSUPPORTED_FEATURE, NULL, + _("Querying mergeinfo requires version %d of the FSFS filesystem " + "schema; filesystem '%s' uses only version %d"), + SVN_FS_FS__MIN_MERGEINFO_FORMAT, root->fs->path, ffd->format); + + /* Retrieve a path -> mergeinfo hash mapping. */ + return get_mergeinfos_for_paths(root, catalog, paths, + inherit, + include_descendants, + adjust_inherited_mergeinfo, + result_pool, scratch_pool); +} + + +/* The vtable associated with root objects. */ +static root_vtable_t root_vtable = { + fs_paths_changed, + svn_fs_fs__check_path, + fs_node_history, + svn_fs_fs__node_id, + svn_fs_fs__node_created_rev, + fs_node_origin_rev, + fs_node_created_path, + fs_delete_node, + fs_copied_from, + fs_closest_copy, + fs_node_prop, + fs_node_proplist, + fs_change_node_prop, + fs_props_changed, + fs_dir_entries, + fs_make_dir, + fs_copy, + fs_revision_link, + fs_file_length, + fs_file_checksum, + fs_file_contents, + fs_try_process_file_contents, + fs_make_file, + fs_apply_textdelta, + fs_apply_text, + fs_contents_changed, + fs_get_file_delta_stream, + fs_merge, + fs_get_mergeinfo, +}; + +/* Construct a new root object in FS, allocated from POOL. */ +static svn_fs_root_t * +make_root(svn_fs_t *fs, + apr_pool_t *pool) +{ + svn_fs_root_t *root = apr_pcalloc(pool, sizeof(*root)); + + root->fs = fs; + root->pool = pool; + root->vtable = &root_vtable; + + return root; +} + + +/* Construct a root object referring to the root of REVISION in FS, + whose root directory is ROOT_DIR. Create the new root in POOL. */ +static svn_fs_root_t * +make_revision_root(svn_fs_t *fs, + svn_revnum_t rev, + dag_node_t *root_dir, + apr_pool_t *pool) +{ + svn_fs_root_t *root = make_root(fs, pool); + fs_rev_root_data_t *frd = apr_pcalloc(root->pool, sizeof(*frd)); + + root->is_txn_root = FALSE; + root->rev = rev; + + frd->root_dir = root_dir; + frd->copyfrom_cache = svn_hash__make(root->pool); + + root->fsap_data = frd; + + return root; +} + + +/* Construct a root object referring to the root of the transaction + named TXN and based on revision BASE_REV in FS, with FLAGS to + describe transaction's behavior. Create the new root in POOL. */ +static svn_error_t * +make_txn_root(svn_fs_root_t **root_p, + svn_fs_t *fs, + const char *txn, + svn_revnum_t base_rev, + apr_uint32_t flags, + apr_pool_t *pool) +{ + svn_fs_root_t *root = make_root(fs, pool); + fs_txn_root_data_t *frd = apr_pcalloc(root->pool, sizeof(*frd)); + + root->is_txn_root = TRUE; + root->txn = apr_pstrdup(root->pool, txn); + root->txn_flags = flags; + root->rev = base_rev; + + frd->txn_id = txn; + + /* Because this cache actually tries to invalidate elements, keep + the number of elements per page down. + + Note that since dag_node_cache_invalidate uses svn_cache__iter, + this *cannot* be a memcache-based cache. */ + SVN_ERR(svn_cache__create_inprocess(&(frd->txn_node_cache), + svn_fs_fs__dag_serialize, + svn_fs_fs__dag_deserialize, + APR_HASH_KEY_STRING, + 32, 20, FALSE, + apr_pstrcat(pool, txn, ":TXN", + (char *)NULL), + root->pool)); + + /* Initialize transaction-local caches in FS. + + Note that we cannot put those caches in frd because that content + fs root object is not available where we would need it. */ + SVN_ERR(svn_fs_fs__initialize_txn_caches(fs, txn, pool)); + + root->fsap_data = frd; + + *root_p = root; + return SVN_NO_ERROR; +} + + + +/* Verify. */ +static APR_INLINE const char * +stringify_node(dag_node_t *node, + apr_pool_t *pool) +{ + /* ### TODO: print some PATH@REV to it, too. */ + return svn_fs_fs__id_unparse(svn_fs_fs__dag_get_id(node), pool)->data; +} + +/* Check metadata sanity on NODE, and on its children. Manually verify + information for DAG nodes in revision REV, and trust the metadata + accuracy for nodes belonging to older revisions. */ +static svn_error_t * +verify_node(dag_node_t *node, + svn_revnum_t rev, + apr_pool_t *pool) +{ + svn_boolean_t has_mergeinfo; + apr_int64_t mergeinfo_count; + const svn_fs_id_t *pred_id; + svn_fs_t *fs = svn_fs_fs__dag_get_fs(node); + int pred_count; + svn_node_kind_t kind; + apr_pool_t *iterpool = svn_pool_create(pool); + + /* Fetch some data. */ + SVN_ERR(svn_fs_fs__dag_has_mergeinfo(&has_mergeinfo, node)); + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&mergeinfo_count, node)); + SVN_ERR(svn_fs_fs__dag_get_predecessor_id(&pred_id, node)); + SVN_ERR(svn_fs_fs__dag_get_predecessor_count(&pred_count, node)); + kind = svn_fs_fs__dag_node_kind(node); + + /* Sanity check. */ + if (mergeinfo_count < 0) + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "Negative mergeinfo-count %" APR_INT64_T_FMT + " on node '%s'", + mergeinfo_count, stringify_node(node, iterpool)); + + /* Issue #4129. (This check will explicitly catch non-root instances too.) */ + if (pred_id) + { + dag_node_t *pred; + int pred_pred_count; + SVN_ERR(svn_fs_fs__dag_get_node(&pred, fs, pred_id, iterpool)); + SVN_ERR(svn_fs_fs__dag_get_predecessor_count(&pred_pred_count, pred)); + if (pred_pred_count+1 != pred_count) + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "Predecessor count mismatch: " + "%s has %d, but %s has %d", + stringify_node(node, iterpool), pred_count, + stringify_node(pred, iterpool), + pred_pred_count); + } + + /* Kind-dependent verifications. */ + if (kind == svn_node_none) + { + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "Node '%s' has kind 'none'", + stringify_node(node, iterpool)); + } + if (kind == svn_node_file) + { + if (has_mergeinfo != mergeinfo_count) /* comparing int to bool */ + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "File node '%s' has inconsistent mergeinfo: " + "has_mergeinfo=%d, " + "mergeinfo_count=%" APR_INT64_T_FMT, + stringify_node(node, iterpool), + has_mergeinfo, mergeinfo_count); + } + if (kind == svn_node_dir) + { + apr_hash_t *entries; + apr_hash_index_t *hi; + apr_int64_t children_mergeinfo = 0; + + SVN_ERR(svn_fs_fs__dag_dir_entries(&entries, node, pool)); + + /* Compute CHILDREN_MERGEINFO. */ + for (hi = apr_hash_first(pool, entries); + hi; + hi = apr_hash_next(hi)) + { + svn_fs_dirent_t *dirent = svn__apr_hash_index_val(hi); + dag_node_t *child; + svn_revnum_t child_rev; + apr_int64_t child_mergeinfo; + + svn_pool_clear(iterpool); + + /* Compute CHILD_REV. */ + SVN_ERR(svn_fs_fs__dag_get_node(&child, fs, dirent->id, iterpool)); + SVN_ERR(svn_fs_fs__dag_get_revision(&child_rev, child, iterpool)); + + if (child_rev == rev) + SVN_ERR(verify_node(child, rev, iterpool)); + + SVN_ERR(svn_fs_fs__dag_get_mergeinfo_count(&child_mergeinfo, child)); + children_mergeinfo += child_mergeinfo; + } + + /* Side-effect of issue #4129. */ + if (children_mergeinfo+has_mergeinfo != mergeinfo_count) + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "Mergeinfo-count discrepancy on '%s': " + "expected %" APR_INT64_T_FMT "+%d, " + "counted %" APR_INT64_T_FMT, + stringify_node(node, iterpool), + mergeinfo_count, has_mergeinfo, + children_mergeinfo); + } + + svn_pool_destroy(iterpool); + return SVN_NO_ERROR; +} + +svn_error_t * +svn_fs_fs__verify_root(svn_fs_root_t *root, + apr_pool_t *pool) +{ + svn_fs_t *fs = root->fs; + dag_node_t *root_dir; + + /* Issue #4129: bogus pred-counts and minfo-cnt's on the root node-rev + (and elsewhere). This code makes more thorough checks than the + commit-time checks in validate_root_noderev(). */ + + /* Callers should disable caches by setting SVN_FS_CONFIG_FSFS_CACHE_NS; + see r1462436. + + When this code is called in the library, we want to ensure we + use the on-disk data --- rather than some data that was read + in the possibly-distance past and cached since. */ + + if (root->is_txn_root) + { + fs_txn_root_data_t *frd = root->fsap_data; + SVN_ERR(svn_fs_fs__dag_txn_root(&root_dir, fs, frd->txn_id, pool)); + } + else + { + fs_rev_root_data_t *frd = root->fsap_data; + root_dir = frd->root_dir; + } + + /* Recursively verify ROOT_DIR. */ + SVN_ERR(verify_node(root_dir, root->rev, pool)); + + /* Verify explicitly the predecessor of the root. */ + { + const svn_fs_id_t *pred_id; + + /* Only r0 should have no predecessor. */ + SVN_ERR(svn_fs_fs__dag_get_predecessor_id(&pred_id, root_dir)); + if (! root->is_txn_root && !!pred_id != !!root->rev) + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "r%ld's root node's predecessor is " + "unexpectedly '%s'", + root->rev, + (pred_id + ? svn_fs_fs__id_unparse(pred_id, pool)->data + : "(null)")); + if (root->is_txn_root && !pred_id) + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "Transaction '%s''s root node's predecessor is " + "unexpectedly NULL", + root->txn); + + /* Check the predecessor's revision. */ + if (pred_id) + { + svn_revnum_t pred_rev = svn_fs_fs__id_rev(pred_id); + if (! root->is_txn_root && pred_rev+1 != root->rev) + /* Issue #4129. */ + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "r%ld's root node's predecessor is r%ld" + " but should be r%ld", + root->rev, pred_rev, root->rev - 1); + if (root->is_txn_root && pred_rev != root->rev) + return svn_error_createf(SVN_ERR_FS_CORRUPT, NULL, + "Transaction '%s''s root node's predecessor" + " is r%ld" + " but should be r%ld", + root->txn, pred_rev, root->rev); + } + } + + return SVN_NO_ERROR; +} |