git-commit-vandalism/refs/files-backend.c

4086 lines
111 KiB
C
Raw Normal View History

#include "../cache.h"
#include "../refs.h"
#include "refs-internal.h"
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
#include "../iterator.h"
#include "../dir-iterator.h"
#include "../lockfile.h"
#include "../object.h"
#include "../dir.h"
struct ref_lock {
char *ref_name;
struct lock_file *lk;
struct object_id old_oid;
};
struct ref_entry;
/*
* Information used (along with the information in ref_entry) to
* describe a single cached reference. This data structure only
* occurs embedded in a union in struct ref_entry, and only when
* (ref_entry->flag & REF_DIR) is zero.
*/
struct ref_value {
/*
* The name of the object to which this reference resolves
* (which may be a tag object). If REF_ISBROKEN, this is
* null. If REF_ISSYMREF, then this is the name of the object
* referred to by the last reference in the symlink chain.
*/
struct object_id oid;
/*
* If REF_KNOWS_PEELED, then this field holds the peeled value
* of this reference, or null if the reference is known not to
* be peelable. See the documentation for peel_ref() for an
* exact definition of "peelable".
*/
struct object_id peeled;
};
struct ref_cache;
/*
* Information used (along with the information in ref_entry) to
* describe a level in the hierarchy of references. This data
* structure only occurs embedded in a union in struct ref_entry, and
* only when (ref_entry.flag & REF_DIR) is set. In that case,
* (ref_entry.flag & REF_INCOMPLETE) determines whether the references
* in the directory have already been read:
*
* (ref_entry.flag & REF_INCOMPLETE) unset -- a directory of loose
* or packed references, already read.
*
* (ref_entry.flag & REF_INCOMPLETE) set -- a directory of loose
* references that hasn't been read yet (nor has any of its
* subdirectories).
*
* Entries within a directory are stored within a growable array of
* pointers to ref_entries (entries, nr, alloc). Entries 0 <= i <
* sorted are sorted by their component name in strcmp() order and the
* remaining entries are unsorted.
*
* Loose references are read lazily, one directory at a time. When a
* directory of loose references is read, then all of the references
* in that directory are stored, and REF_INCOMPLETE stubs are created
* for any subdirectories, but the subdirectories themselves are not
* read. The reading is triggered by get_ref_dir().
*/
struct ref_dir {
int nr, alloc;
/*
* Entries with index 0 <= i < sorted are sorted by name. New
* entries are appended to the list unsorted, and are sorted
* only when required; thus we avoid the need to sort the list
* after the addition of every reference.
*/
int sorted;
/* A pointer to the ref_cache that contains this ref_dir. */
struct ref_cache *ref_cache;
struct ref_entry **entries;
};
/*
* Bit values for ref_entry::flag. REF_ISSYMREF=0x01,
* REF_ISPACKED=0x02, REF_ISBROKEN=0x04 and REF_BAD_NAME=0x08 are
* public values; see refs.h.
*/
/*
* The field ref_entry->u.value.peeled of this value entry contains
* the correct peeled value for the reference, which might be
* null_sha1 if the reference is not a tag or if it is broken.
*/
#define REF_KNOWS_PEELED 0x10
/* ref_entry represents a directory of references */
#define REF_DIR 0x20
/*
* Entry has not yet been read from disk (used only for REF_DIR
* entries representing loose references)
*/
#define REF_INCOMPLETE 0x40
/*
* A ref_entry represents either a reference or a "subdirectory" of
* references.
*
* Each directory in the reference namespace is represented by a
* ref_entry with (flags & REF_DIR) set and containing a subdir member
* that holds the entries in that directory that have been read so
* far. If (flags & REF_INCOMPLETE) is set, then the directory and
* its subdirectories haven't been read yet. REF_INCOMPLETE is only
* used for loose reference directories.
*
* References are represented by a ref_entry with (flags & REF_DIR)
* unset and a value member that describes the reference's value. The
* flag member is at the ref_entry level, but it is also needed to
* interpret the contents of the value field (in other words, a
* ref_value object is not very much use without the enclosing
* ref_entry).
*
* Reference names cannot end with slash and directories' names are
* always stored with a trailing slash (except for the top-level
* directory, which is always denoted by ""). This has two nice
* consequences: (1) when the entries in each subdir are sorted
* lexicographically by name (as they usually are), the references in
* a whole tree can be generated in lexicographic order by traversing
* the tree in left-to-right, depth-first order; (2) the names of
* references and subdirectories cannot conflict, and therefore the
* presence of an empty subdirectory does not block the creation of a
* similarly-named reference. (The fact that reference names with the
* same leading components can conflict *with each other* is a
* separate issue that is regulated by verify_refname_available().)
*
* Please note that the name field contains the fully-qualified
* reference (or subdirectory) name. Space could be saved by only
* storing the relative names. But that would require the full names
* to be generated on the fly when iterating in do_for_each_ref(), and
* would break callback functions, who have always been able to assume
* that the name strings that they are passed will not be freed during
* the iteration.
*/
struct ref_entry {
unsigned char flag; /* ISSYMREF? ISPACKED? */
union {
struct ref_value value; /* if not (flags&REF_DIR) */
struct ref_dir subdir; /* if (flags&REF_DIR) */
} u;
/*
* The full name of the reference (e.g., "refs/heads/master")
* or the full name of the directory with a trailing slash
* (e.g., "refs/heads/"):
*/
char name[FLEX_ARRAY];
};
static void read_loose_refs(const char *dirname, struct ref_dir *dir);
static int search_ref_dir(struct ref_dir *dir, const char *refname, size_t len);
static struct ref_entry *create_dir_entry(struct ref_cache *ref_cache,
const char *dirname, size_t len,
int incomplete);
static void add_entry_to_dir(struct ref_dir *dir, struct ref_entry *entry);
static struct ref_dir *get_ref_dir(struct ref_entry *entry)
{
struct ref_dir *dir;
assert(entry->flag & REF_DIR);
dir = &entry->u.subdir;
if (entry->flag & REF_INCOMPLETE) {
read_loose_refs(entry->name, dir);
/*
* Manually add refs/bisect, which, being
* per-worktree, might not appear in the directory
* listing for refs/ in the main repo.
*/
if (!strcmp(entry->name, "refs/")) {
int pos = search_ref_dir(dir, "refs/bisect/", 12);
if (pos < 0) {
struct ref_entry *child_entry;
child_entry = create_dir_entry(dir->ref_cache,
"refs/bisect/",
12, 1);
add_entry_to_dir(dir, child_entry);
read_loose_refs("refs/bisect",
&child_entry->u.subdir);
}
}
entry->flag &= ~REF_INCOMPLETE;
}
return dir;
}
static struct ref_entry *create_ref_entry(const char *refname,
const unsigned char *sha1, int flag,
int check_name)
{
struct ref_entry *ref;
if (check_name &&
check_refname_format(refname, REFNAME_ALLOW_ONELEVEL))
die("Reference has invalid format: '%s'", refname);
FLEX_ALLOC_STR(ref, name, refname);
hashcpy(ref->u.value.oid.hash, sha1);
oidclr(&ref->u.value.peeled);
ref->flag = flag;
return ref;
}
static void clear_ref_dir(struct ref_dir *dir);
static void free_ref_entry(struct ref_entry *entry)
{
if (entry->flag & REF_DIR) {
/*
* Do not use get_ref_dir() here, as that might
* trigger the reading of loose refs.
*/
clear_ref_dir(&entry->u.subdir);
}
free(entry);
}
/*
* Add a ref_entry to the end of dir (unsorted). Entry is always
* stored directly in dir; no recursion into subdirectories is
* done.
*/
static void add_entry_to_dir(struct ref_dir *dir, struct ref_entry *entry)
{
ALLOC_GROW(dir->entries, dir->nr + 1, dir->alloc);
dir->entries[dir->nr++] = entry;
/* optimize for the case that entries are added in order */
if (dir->nr == 1 ||
(dir->nr == dir->sorted + 1 &&
strcmp(dir->entries[dir->nr - 2]->name,
dir->entries[dir->nr - 1]->name) < 0))
dir->sorted = dir->nr;
}
/*
* Clear and free all entries in dir, recursively.
*/
static void clear_ref_dir(struct ref_dir *dir)
{
int i;
for (i = 0; i < dir->nr; i++)
free_ref_entry(dir->entries[i]);
free(dir->entries);
dir->sorted = dir->nr = dir->alloc = 0;
dir->entries = NULL;
}
/*
* Create a struct ref_entry object for the specified dirname.
* dirname is the name of the directory with a trailing slash (e.g.,
* "refs/heads/") or "" for the top-level directory.
*/
static struct ref_entry *create_dir_entry(struct ref_cache *ref_cache,
const char *dirname, size_t len,
int incomplete)
{
struct ref_entry *direntry;
FLEX_ALLOC_MEM(direntry, name, dirname, len);
direntry->u.subdir.ref_cache = ref_cache;
direntry->flag = REF_DIR | (incomplete ? REF_INCOMPLETE : 0);
return direntry;
}
static int ref_entry_cmp(const void *a, const void *b)
{
struct ref_entry *one = *(struct ref_entry **)a;
struct ref_entry *two = *(struct ref_entry **)b;
return strcmp(one->name, two->name);
}
static void sort_ref_dir(struct ref_dir *dir);
struct string_slice {
size_t len;
const char *str;
};
static int ref_entry_cmp_sslice(const void *key_, const void *ent_)
{
const struct string_slice *key = key_;
const struct ref_entry *ent = *(const struct ref_entry * const *)ent_;
int cmp = strncmp(key->str, ent->name, key->len);
if (cmp)
return cmp;
return '\0' - (unsigned char)ent->name[key->len];
}
/*
* Return the index of the entry with the given refname from the
* ref_dir (non-recursively), sorting dir if necessary. Return -1 if
* no such entry is found. dir must already be complete.
*/
static int search_ref_dir(struct ref_dir *dir, const char *refname, size_t len)
{
struct ref_entry **r;
struct string_slice key;
if (refname == NULL || !dir->nr)
return -1;
sort_ref_dir(dir);
key.len = len;
key.str = refname;
r = bsearch(&key, dir->entries, dir->nr, sizeof(*dir->entries),
ref_entry_cmp_sslice);
if (r == NULL)
return -1;
return r - dir->entries;
}
/*
* Search for a directory entry directly within dir (without
* recursing). Sort dir if necessary. subdirname must be a directory
* name (i.e., end in '/'). If mkdir is set, then create the
* directory if it is missing; otherwise, return NULL if the desired
* directory cannot be found. dir must already be complete.
*/
static struct ref_dir *search_for_subdir(struct ref_dir *dir,
const char *subdirname, size_t len,
int mkdir)
{
int entry_index = search_ref_dir(dir, subdirname, len);
struct ref_entry *entry;
if (entry_index == -1) {
if (!mkdir)
return NULL;
/*
* Since dir is complete, the absence of a subdir
* means that the subdir really doesn't exist;
* therefore, create an empty record for it but mark
* the record complete.
*/
entry = create_dir_entry(dir->ref_cache, subdirname, len, 0);
add_entry_to_dir(dir, entry);
} else {
entry = dir->entries[entry_index];
}
return get_ref_dir(entry);
}
/*
* If refname is a reference name, find the ref_dir within the dir
* tree that should hold refname. If refname is a directory name
* (i.e., ends in '/'), then return that ref_dir itself. dir must
* represent the top-level directory and must already be complete.
* Sort ref_dirs and recurse into subdirectories as necessary. If
* mkdir is set, then create any missing directories; otherwise,
* return NULL if the desired directory cannot be found.
*/
static struct ref_dir *find_containing_dir(struct ref_dir *dir,
const char *refname, int mkdir)
{
const char *slash;
for (slash = strchr(refname, '/'); slash; slash = strchr(slash + 1, '/')) {
size_t dirnamelen = slash - refname + 1;
struct ref_dir *subdir;
subdir = search_for_subdir(dir, refname, dirnamelen, mkdir);
if (!subdir) {
dir = NULL;
break;
}
dir = subdir;
}
return dir;
}
/*
* Find the value entry with the given name in dir, sorting ref_dirs
* and recursing into subdirectories as necessary. If the name is not
* found or it corresponds to a directory entry, return NULL.
*/
static struct ref_entry *find_ref(struct ref_dir *dir, const char *refname)
{
int entry_index;
struct ref_entry *entry;
dir = find_containing_dir(dir, refname, 0);
if (!dir)
return NULL;
entry_index = search_ref_dir(dir, refname, strlen(refname));
if (entry_index == -1)
return NULL;
entry = dir->entries[entry_index];
return (entry->flag & REF_DIR) ? NULL : entry;
}
/*
* Remove the entry with the given name from dir, recursing into
* subdirectories as necessary. If refname is the name of a directory
* (i.e., ends with '/'), then remove the directory and its contents.
* If the removal was successful, return the number of entries
* remaining in the directory entry that contained the deleted entry.
* If the name was not found, return -1. Please note that this
* function only deletes the entry from the cache; it does not delete
* it from the filesystem or ensure that other cache entries (which
* might be symbolic references to the removed entry) are updated.
* Nor does it remove any containing dir entries that might be made
* empty by the removal. dir must represent the top-level directory
* and must already be complete.
*/
static int remove_entry(struct ref_dir *dir, const char *refname)
{
int refname_len = strlen(refname);
int entry_index;
struct ref_entry *entry;
int is_dir = refname[refname_len - 1] == '/';
if (is_dir) {
/*
* refname represents a reference directory. Remove
* the trailing slash; otherwise we will get the
* directory *representing* refname rather than the
* one *containing* it.
*/
char *dirname = xmemdupz(refname, refname_len - 1);
dir = find_containing_dir(dir, dirname, 0);
free(dirname);
} else {
dir = find_containing_dir(dir, refname, 0);
}
if (!dir)
return -1;
entry_index = search_ref_dir(dir, refname, refname_len);
if (entry_index == -1)
return -1;
entry = dir->entries[entry_index];
memmove(&dir->entries[entry_index],
&dir->entries[entry_index + 1],
(dir->nr - entry_index - 1) * sizeof(*dir->entries)
);
dir->nr--;
if (dir->sorted > entry_index)
dir->sorted--;
free_ref_entry(entry);
return dir->nr;
}
/*
* Add a ref_entry to the ref_dir (unsorted), recursing into
* subdirectories as necessary. dir must represent the top-level
* directory. Return 0 on success.
*/
static int add_ref(struct ref_dir *dir, struct ref_entry *ref)
{
dir = find_containing_dir(dir, ref->name, 1);
if (!dir)
return -1;
add_entry_to_dir(dir, ref);
return 0;
}
/*
* Emit a warning and return true iff ref1 and ref2 have the same name
* and the same sha1. Die if they have the same name but different
* sha1s.
*/
static int is_dup_ref(const struct ref_entry *ref1, const struct ref_entry *ref2)
{
if (strcmp(ref1->name, ref2->name))
return 0;
/* Duplicate name; make sure that they don't conflict: */
if ((ref1->flag & REF_DIR) || (ref2->flag & REF_DIR))
/* This is impossible by construction */
die("Reference directory conflict: %s", ref1->name);
if (oidcmp(&ref1->u.value.oid, &ref2->u.value.oid))
die("Duplicated ref, and SHA1s don't match: %s", ref1->name);
warning("Duplicated ref: %s", ref1->name);
return 1;
}
/*
* Sort the entries in dir non-recursively (if they are not already
* sorted) and remove any duplicate entries.
*/
static void sort_ref_dir(struct ref_dir *dir)
{
int i, j;
struct ref_entry *last = NULL;
/*
* This check also prevents passing a zero-length array to qsort(),
* which is a problem on some platforms.
*/
if (dir->sorted == dir->nr)
return;
qsort(dir->entries, dir->nr, sizeof(*dir->entries), ref_entry_cmp);
/* Remove any duplicates: */
for (i = 0, j = 0; j < dir->nr; j++) {
struct ref_entry *entry = dir->entries[j];
if (last && is_dup_ref(last, entry))
free_ref_entry(entry);
else
last = dir->entries[i++] = entry;
}
dir->sorted = dir->nr = i;
}
/*
* Return true if refname, which has the specified oid and flags, can
* be resolved to an object in the database. If the referred-to object
* does not exist, emit a warning and return false.
*/
static int ref_resolves_to_object(const char *refname,
const struct object_id *oid,
unsigned int flags)
{
if (flags & REF_ISBROKEN)
return 0;
if (!has_sha1_file(oid->hash)) {
error("%s does not point to a valid object!", refname);
return 0;
}
return 1;
}
/*
* Return true if the reference described by entry can be resolved to
* an object in the database; otherwise, emit a warning and return
* false.
*/
static int entry_resolves_to_object(struct ref_entry *entry)
{
return ref_resolves_to_object(entry->name,
&entry->u.value.oid, entry->flag);
}
typedef int each_ref_entry_fn(struct ref_entry *entry, void *cb_data);
/*
* Call fn for each reference in dir that has index in the range
* offset <= index < dir->nr. Recurse into subdirectories that are in
* that index range, sorting them before iterating. This function
* does not sort dir itself; it should be sorted beforehand. fn is
* called for all references, including broken ones.
*/
static int do_for_each_entry_in_dir(struct ref_dir *dir, int offset,
each_ref_entry_fn fn, void *cb_data)
{
int i;
assert(dir->sorted == dir->nr);
for (i = offset; i < dir->nr; i++) {
struct ref_entry *entry = dir->entries[i];
int retval;
if (entry->flag & REF_DIR) {
struct ref_dir *subdir = get_ref_dir(entry);
sort_ref_dir(subdir);
retval = do_for_each_entry_in_dir(subdir, 0, fn, cb_data);
} else {
retval = fn(entry, cb_data);
}
if (retval)
return retval;
}
return 0;
}
/*
* Load all of the refs from the dir into our in-memory cache. The hard work
* of loading loose refs is done by get_ref_dir(), so we just need to recurse
* through all of the sub-directories. We do not even need to care about
* sorting, as traversal order does not matter to us.
*/
static void prime_ref_dir(struct ref_dir *dir)
{
int i;
for (i = 0; i < dir->nr; i++) {
struct ref_entry *entry = dir->entries[i];
if (entry->flag & REF_DIR)
prime_ref_dir(get_ref_dir(entry));
}
}
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
/*
* A level in the reference hierarchy that is currently being iterated
* through.
*/
struct cache_ref_iterator_level {
/*
* The ref_dir being iterated over at this level. The ref_dir
* is sorted before being stored here.
*/
struct ref_dir *dir;
/*
* The index of the current entry within dir (which might
* itself be a directory). If index == -1, then the iteration
* hasn't yet begun. If index == dir->nr, then the iteration
* through this level is over.
*/
int index;
};
/*
* Represent an iteration through a ref_dir in the memory cache. The
* iteration recurses through subdirectories.
*/
struct cache_ref_iterator {
struct ref_iterator base;
/*
* The number of levels currently on the stack. This is always
* at least 1, because when it becomes zero the iteration is
* ended and this struct is freed.
*/
size_t levels_nr;
/* The number of levels that have been allocated on the stack */
size_t levels_alloc;
/*
* A stack of levels. levels[0] is the uppermost level that is
* being iterated over in this iteration. (This is not
* necessary the top level in the references hierarchy. If we
* are iterating through a subtree, then levels[0] will hold
* the ref_dir for that subtree, and subsequent levels will go
* on from there.)
*/
struct cache_ref_iterator_level *levels;
};
static int cache_ref_iterator_advance(struct ref_iterator *ref_iterator)
{
struct cache_ref_iterator *iter =
(struct cache_ref_iterator *)ref_iterator;
while (1) {
struct cache_ref_iterator_level *level =
&iter->levels[iter->levels_nr - 1];
struct ref_dir *dir = level->dir;
struct ref_entry *entry;
if (level->index == -1)
sort_ref_dir(dir);
if (++level->index == level->dir->nr) {
/* This level is exhausted; pop up a level */
if (--iter->levels_nr == 0)
return ref_iterator_abort(ref_iterator);
continue;
}
entry = dir->entries[level->index];
if (entry->flag & REF_DIR) {
/* push down a level */
ALLOC_GROW(iter->levels, iter->levels_nr + 1,
iter->levels_alloc);
level = &iter->levels[iter->levels_nr++];
level->dir = get_ref_dir(entry);
level->index = -1;
} else {
iter->base.refname = entry->name;
iter->base.oid = &entry->u.value.oid;
iter->base.flags = entry->flag;
return ITER_OK;
}
}
}
static enum peel_status peel_entry(struct ref_entry *entry, int repeel);
static int cache_ref_iterator_peel(struct ref_iterator *ref_iterator,
struct object_id *peeled)
{
struct cache_ref_iterator *iter =
(struct cache_ref_iterator *)ref_iterator;
struct cache_ref_iterator_level *level;
struct ref_entry *entry;
level = &iter->levels[iter->levels_nr - 1];
if (level->index == -1)
die("BUG: peel called before advance for cache iterator");
entry = level->dir->entries[level->index];
if (peel_entry(entry, 0))
return -1;
hashcpy(peeled->hash, entry->u.value.peeled.hash);
return 0;
}
static int cache_ref_iterator_abort(struct ref_iterator *ref_iterator)
{
struct cache_ref_iterator *iter =
(struct cache_ref_iterator *)ref_iterator;
free(iter->levels);
base_ref_iterator_free(ref_iterator);
return ITER_DONE;
}
static struct ref_iterator_vtable cache_ref_iterator_vtable = {
cache_ref_iterator_advance,
cache_ref_iterator_peel,
cache_ref_iterator_abort
};
static struct ref_iterator *cache_ref_iterator_begin(struct ref_dir *dir)
{
struct cache_ref_iterator *iter;
struct ref_iterator *ref_iterator;
struct cache_ref_iterator_level *level;
iter = xcalloc(1, sizeof(*iter));
ref_iterator = &iter->base;
base_ref_iterator_init(ref_iterator, &cache_ref_iterator_vtable);
ALLOC_GROW(iter->levels, 10, iter->levels_alloc);
iter->levels_nr = 1;
level = &iter->levels[0];
level->index = -1;
level->dir = dir;
return ref_iterator;
}
struct nonmatching_ref_data {
const struct string_list *skip;
const char *conflicting_refname;
};
static int nonmatching_ref_fn(struct ref_entry *entry, void *vdata)
{
struct nonmatching_ref_data *data = vdata;
if (data->skip && string_list_has_string(data->skip, entry->name))
return 0;
data->conflicting_refname = entry->name;
return 1;
}
/*
* Return 0 if a reference named refname could be created without
* conflicting with the name of an existing reference in dir.
* See verify_refname_available for more information.
*/
static int verify_refname_available_dir(const char *refname,
const struct string_list *extras,
const struct string_list *skip,
struct ref_dir *dir,
struct strbuf *err)
{
const char *slash;
const char *extra_refname;
int pos;
struct strbuf dirname = STRBUF_INIT;
int ret = -1;
/*
* For the sake of comments in this function, suppose that
* refname is "refs/foo/bar".
*/
assert(err);
strbuf_grow(&dirname, strlen(refname) + 1);
for (slash = strchr(refname, '/'); slash; slash = strchr(slash + 1, '/')) {
/* Expand dirname to the new prefix, not including the trailing slash: */
strbuf_add(&dirname, refname + dirname.len, slash - refname - dirname.len);
/*
* We are still at a leading dir of the refname (e.g.,
* "refs/foo"; if there is a reference with that name,
* it is a conflict, *unless* it is in skip.
*/
if (dir) {
pos = search_ref_dir(dir, dirname.buf, dirname.len);
if (pos >= 0 &&
(!skip || !string_list_has_string(skip, dirname.buf))) {
/*
* We found a reference whose name is
* a proper prefix of refname; e.g.,
* "refs/foo", and is not in skip.
*/
strbuf_addf(err, "'%s' exists; cannot create '%s'",
dirname.buf, refname);
goto cleanup;
}
}
if (extras && string_list_has_string(extras, dirname.buf) &&
(!skip || !string_list_has_string(skip, dirname.buf))) {
strbuf_addf(err, "cannot process '%s' and '%s' at the same time",
refname, dirname.buf);
goto cleanup;
}
/*
* Otherwise, we can try to continue our search with
* the next component. So try to look up the
* directory, e.g., "refs/foo/". If we come up empty,
* we know there is nothing under this whole prefix,
* but even in that case we still have to continue the
* search for conflicts with extras.
*/
strbuf_addch(&dirname, '/');
if (dir) {
pos = search_ref_dir(dir, dirname.buf, dirname.len);
if (pos < 0) {
/*
* There was no directory "refs/foo/",
* so there is nothing under this
* whole prefix. So there is no need
* to continue looking for conflicting
* references. But we need to continue
* looking for conflicting extras.
*/
dir = NULL;
} else {
dir = get_ref_dir(dir->entries[pos]);
}
}
}
/*
* We are at the leaf of our refname (e.g., "refs/foo/bar").
* There is no point in searching for a reference with that
* name, because a refname isn't considered to conflict with
* itself. But we still need to check for references whose
* names are in the "refs/foo/bar/" namespace, because they
* *do* conflict.
*/
strbuf_addstr(&dirname, refname + dirname.len);
strbuf_addch(&dirname, '/');
if (dir) {
pos = search_ref_dir(dir, dirname.buf, dirname.len);
if (pos >= 0) {
/*
* We found a directory named "$refname/"
* (e.g., "refs/foo/bar/"). It is a problem
* iff it contains any ref that is not in
* "skip".
*/
struct nonmatching_ref_data data;
data.skip = skip;
data.conflicting_refname = NULL;
dir = get_ref_dir(dir->entries[pos]);
sort_ref_dir(dir);
if (do_for_each_entry_in_dir(dir, 0, nonmatching_ref_fn, &data)) {
strbuf_addf(err, "'%s' exists; cannot create '%s'",
data.conflicting_refname, refname);
goto cleanup;
}
}
}
extra_refname = find_descendant_ref(dirname.buf, extras, skip);
if (extra_refname)
strbuf_addf(err, "cannot process '%s' and '%s' at the same time",
refname, extra_refname);
else
ret = 0;
cleanup:
strbuf_release(&dirname);
return ret;
}
struct packed_ref_cache {
struct ref_entry *root;
/*
* Count of references to the data structure in this instance,
* including the pointer from ref_cache::packed if any. The
* data will not be freed as long as the reference count is
* nonzero.
*/
unsigned int referrers;
/*
* Iff the packed-refs file associated with this instance is
* currently locked for writing, this points at the associated
* lock (which is owned by somebody else). The referrer count
* is also incremented when the file is locked and decremented
* when it is unlocked.
*/
struct lock_file *lock;
/* The metadata from when this packed-refs cache was read */
struct stat_validity validity;
};
/*
* Future: need to be in "struct repository"
* when doing a full libification.
*/
static struct ref_cache {
struct ref_cache *next;
struct ref_entry *loose;
struct packed_ref_cache *packed;
/*
* The submodule name, or "" for the main repo. We allocate
* length 1 rather than FLEX_ARRAY so that the main ref_cache
* is initialized correctly.
*/
char name[1];
} ref_cache, *submodule_ref_caches;
/* Lock used for the main packed-refs file: */
static struct lock_file packlock;
/*
* Increment the reference count of *packed_refs.
*/
static void acquire_packed_ref_cache(struct packed_ref_cache *packed_refs)
{
packed_refs->referrers++;
}
/*
* Decrease the reference count of *packed_refs. If it goes to zero,
* free *packed_refs and return true; otherwise return false.
*/
static int release_packed_ref_cache(struct packed_ref_cache *packed_refs)
{
if (!--packed_refs->referrers) {
free_ref_entry(packed_refs->root);
stat_validity_clear(&packed_refs->validity);
free(packed_refs);
return 1;
} else {
return 0;
}
}
static void clear_packed_ref_cache(struct ref_cache *refs)
{
if (refs->packed) {
struct packed_ref_cache *packed_refs = refs->packed;
if (packed_refs->lock)
die("internal error: packed-ref cache cleared while locked");
refs->packed = NULL;
release_packed_ref_cache(packed_refs);
}
}
static void clear_loose_ref_cache(struct ref_cache *refs)
{
if (refs->loose) {
free_ref_entry(refs->loose);
refs->loose = NULL;
}
}
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
/*
* Create a new submodule ref cache and add it to the internal
* set of caches.
*/
static struct ref_cache *create_ref_cache(const char *submodule)
{
struct ref_cache *refs;
if (!submodule)
submodule = "";
FLEX_ALLOC_STR(refs, name, submodule);
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
refs->next = submodule_ref_caches;
submodule_ref_caches = refs;
return refs;
}
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
static struct ref_cache *lookup_ref_cache(const char *submodule)
{
struct ref_cache *refs;
if (!submodule || !*submodule)
return &ref_cache;
for (refs = submodule_ref_caches; refs; refs = refs->next)
if (!strcmp(submodule, refs->name))
return refs;
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
return NULL;
}
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
/*
* Return a pointer to a ref_cache for the specified submodule. For
* the main repository, use submodule==NULL; such a call cannot fail.
* For a submodule, the submodule must exist and be a nonbare
* repository, otherwise return NULL.
*
* The returned structure will be allocated and initialized but not
* necessarily populated; it should not be freed.
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
*/
static struct ref_cache *get_ref_cache(const char *submodule)
{
struct ref_cache *refs = lookup_ref_cache(submodule);
if (!refs) {
struct strbuf submodule_sb = STRBUF_INIT;
strbuf_addstr(&submodule_sb, submodule);
if (is_nonbare_repository_dir(&submodule_sb))
refs = create_ref_cache(submodule);
strbuf_release(&submodule_sb);
}
return refs;
}
/* The length of a peeled reference line in packed-refs, including EOL: */
#define PEELED_LINE_LENGTH 42
/*
* The packed-refs header line that we write out. Perhaps other
* traits will be added later. The trailing space is required.
*/
static const char PACKED_REFS_HEADER[] =
"# pack-refs with: peeled fully-peeled \n";
/*
* Parse one line from a packed-refs file. Write the SHA1 to sha1.
* Return a pointer to the refname within the line (null-terminated),
* or NULL if there was a problem.
*/
static const char *parse_ref_line(struct strbuf *line, unsigned char *sha1)
{
const char *ref;
/*
* 42: the answer to everything.
*
* In this case, it happens to be the answer to
* 40 (length of sha1 hex representation)
* +1 (space in between hex and name)
* +1 (newline at the end of the line)
*/
if (line->len <= 42)
return NULL;
if (get_sha1_hex(line->buf, sha1) < 0)
return NULL;
if (!isspace(line->buf[40]))
return NULL;
ref = line->buf + 41;
if (isspace(*ref))
return NULL;
if (line->buf[line->len - 1] != '\n')
return NULL;
line->buf[--line->len] = 0;
return ref;
}
/*
* Read f, which is a packed-refs file, into dir.
*
* A comment line of the form "# pack-refs with: " may contain zero or
* more traits. We interpret the traits as follows:
*
* No traits:
*
* Probably no references are peeled. But if the file contains a
* peeled value for a reference, we will use it.
*
* peeled:
*
* References under "refs/tags/", if they *can* be peeled, *are*
* peeled in this file. References outside of "refs/tags/" are
* probably not peeled even if they could have been, but if we find
* a peeled value for such a reference we will use it.
*
* fully-peeled:
*
* All references in the file that can be peeled are peeled.
* Inversely (and this is more important), any references in the
* file for which no peeled value is recorded is not peelable. This
* trait should typically be written alongside "peeled" for
* compatibility with older clients, but we do not require it
* (i.e., "peeled" is a no-op if "fully-peeled" is set).
*/
static void read_packed_refs(FILE *f, struct ref_dir *dir)
{
struct ref_entry *last = NULL;
struct strbuf line = STRBUF_INIT;
enum { PEELED_NONE, PEELED_TAGS, PEELED_FULLY } peeled = PEELED_NONE;
while (strbuf_getwholeline(&line, f, '\n') != EOF) {
unsigned char sha1[20];
const char *refname;
const char *traits;
if (skip_prefix(line.buf, "# pack-refs with:", &traits)) {
if (strstr(traits, " fully-peeled "))
peeled = PEELED_FULLY;
else if (strstr(traits, " peeled "))
peeled = PEELED_TAGS;
/* perhaps other traits later as well */
continue;
}
refname = parse_ref_line(&line, sha1);
if (refname) {
int flag = REF_ISPACKED;
if (check_refname_format(refname, REFNAME_ALLOW_ONELEVEL)) {
if (!refname_is_safe(refname))
die("packed refname is dangerous: %s", refname);
hashclr(sha1);
flag |= REF_BAD_NAME | REF_ISBROKEN;
}
last = create_ref_entry(refname, sha1, flag, 0);
if (peeled == PEELED_FULLY ||
(peeled == PEELED_TAGS && starts_with(refname, "refs/tags/")))
last->flag |= REF_KNOWS_PEELED;
add_ref(dir, last);
continue;
}
if (last &&
line.buf[0] == '^' &&
line.len == PEELED_LINE_LENGTH &&
line.buf[PEELED_LINE_LENGTH - 1] == '\n' &&
!get_sha1_hex(line.buf + 1, sha1)) {
hashcpy(last->u.value.peeled.hash, sha1);
/*
* Regardless of what the file header said,
* we definitely know the value of *this*
* reference:
*/
last->flag |= REF_KNOWS_PEELED;
}
}
strbuf_release(&line);
}
/*
* Get the packed_ref_cache for the specified ref_cache, creating it
* if necessary.
*/
static struct packed_ref_cache *get_packed_ref_cache(struct ref_cache *refs)
{
char *packed_refs_file;
if (*refs->name)
packed_refs_file = git_pathdup_submodule(refs->name, "packed-refs");
else
packed_refs_file = git_pathdup("packed-refs");
if (refs->packed &&
!stat_validity_check(&refs->packed->validity, packed_refs_file))
clear_packed_ref_cache(refs);
if (!refs->packed) {
FILE *f;
refs->packed = xcalloc(1, sizeof(*refs->packed));
acquire_packed_ref_cache(refs->packed);
refs->packed->root = create_dir_entry(refs, "", 0, 0);
f = fopen(packed_refs_file, "r");
if (f) {
stat_validity_update(&refs->packed->validity, fileno(f));
read_packed_refs(f, get_ref_dir(refs->packed->root));
fclose(f);
}
}
free(packed_refs_file);
return refs->packed;
}
static struct ref_dir *get_packed_ref_dir(struct packed_ref_cache *packed_ref_cache)
{
return get_ref_dir(packed_ref_cache->root);
}
static struct ref_dir *get_packed_refs(struct ref_cache *refs)
{
return get_packed_ref_dir(get_packed_ref_cache(refs));
}
/*
* Add a reference to the in-memory packed reference cache. This may
* only be called while the packed-refs file is locked (see
* lock_packed_refs()). To actually write the packed-refs file, call
* commit_packed_refs().
*/
static void add_packed_ref(const char *refname, const unsigned char *sha1)
{
struct packed_ref_cache *packed_ref_cache =
get_packed_ref_cache(&ref_cache);
if (!packed_ref_cache->lock)
die("internal error: packed refs not locked");
add_ref(get_packed_ref_dir(packed_ref_cache),
create_ref_entry(refname, sha1, REF_ISPACKED, 1));
}
/*
* Read the loose references from the namespace dirname into dir
* (without recursing). dirname must end with '/'. dir must be the
* directory entry corresponding to dirname.
*/
static void read_loose_refs(const char *dirname, struct ref_dir *dir)
{
struct ref_cache *refs = dir->ref_cache;
DIR *d;
struct dirent *de;
int dirnamelen = strlen(dirname);
struct strbuf refname;
struct strbuf path = STRBUF_INIT;
size_t path_baselen;
allow do_submodule_path to work even if submodule isn't checked out Currently, do_submodule_path will attempt locating the .git directory by using read_gitfile on <path>/.git. If this fails it just assumes the <path>/.git is actually a git directory. This is good because it allows for handling submodules which were cloned in a regular manner first before being added to the superproject. Unfortunately this fails if the <path> is not actually checked out any longer, such as by removing the directory. Fix this by checking if the directory we found is actually a gitdir. In the case it is not, attempt to lookup the submodule configuration and find the name of where it is stored in the .git/modules/ directory of the superproject. If we can't locate the submodule configuration, this might occur because for example a submodule gitlink was added but the corresponding .gitmodules file was not properly updated. A die() here would not be pleasant to the users of submodule diff formats, so instead, modify do_submodule_path() to return an error code: - git_pathdup_submodule() returns NULL when we fail to find a path. - strbuf_git_path_submodule() propagates the error code to the caller. Modify the callers of these functions to check the error code and fail properly. This ensures we don't attempt to use a bad path that doesn't match the corresponding submodule. Because this change fixes add_submodule_odb() to work even if the submodule is not checked out, update the wording of the submodule log diff format to correctly display that the submodule is "not initialized" instead of "not checked out" Add tests to ensure this change works as expected. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-09-01 01:27:22 +02:00
int err = 0;
if (*refs->name)
allow do_submodule_path to work even if submodule isn't checked out Currently, do_submodule_path will attempt locating the .git directory by using read_gitfile on <path>/.git. If this fails it just assumes the <path>/.git is actually a git directory. This is good because it allows for handling submodules which were cloned in a regular manner first before being added to the superproject. Unfortunately this fails if the <path> is not actually checked out any longer, such as by removing the directory. Fix this by checking if the directory we found is actually a gitdir. In the case it is not, attempt to lookup the submodule configuration and find the name of where it is stored in the .git/modules/ directory of the superproject. If we can't locate the submodule configuration, this might occur because for example a submodule gitlink was added but the corresponding .gitmodules file was not properly updated. A die() here would not be pleasant to the users of submodule diff formats, so instead, modify do_submodule_path() to return an error code: - git_pathdup_submodule() returns NULL when we fail to find a path. - strbuf_git_path_submodule() propagates the error code to the caller. Modify the callers of these functions to check the error code and fail properly. This ensures we don't attempt to use a bad path that doesn't match the corresponding submodule. Because this change fixes add_submodule_odb() to work even if the submodule is not checked out, update the wording of the submodule log diff format to correctly display that the submodule is "not initialized" instead of "not checked out" Add tests to ensure this change works as expected. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-09-01 01:27:22 +02:00
err = strbuf_git_path_submodule(&path, refs->name, "%s", dirname);
else
strbuf_git_path(&path, "%s", dirname);
path_baselen = path.len;
allow do_submodule_path to work even if submodule isn't checked out Currently, do_submodule_path will attempt locating the .git directory by using read_gitfile on <path>/.git. If this fails it just assumes the <path>/.git is actually a git directory. This is good because it allows for handling submodules which were cloned in a regular manner first before being added to the superproject. Unfortunately this fails if the <path> is not actually checked out any longer, such as by removing the directory. Fix this by checking if the directory we found is actually a gitdir. In the case it is not, attempt to lookup the submodule configuration and find the name of where it is stored in the .git/modules/ directory of the superproject. If we can't locate the submodule configuration, this might occur because for example a submodule gitlink was added but the corresponding .gitmodules file was not properly updated. A die() here would not be pleasant to the users of submodule diff formats, so instead, modify do_submodule_path() to return an error code: - git_pathdup_submodule() returns NULL when we fail to find a path. - strbuf_git_path_submodule() propagates the error code to the caller. Modify the callers of these functions to check the error code and fail properly. This ensures we don't attempt to use a bad path that doesn't match the corresponding submodule. Because this change fixes add_submodule_odb() to work even if the submodule is not checked out, update the wording of the submodule log diff format to correctly display that the submodule is "not initialized" instead of "not checked out" Add tests to ensure this change works as expected. Signed-off-by: Jacob Keller <jacob.keller@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-09-01 01:27:22 +02:00
if (err) {
strbuf_release(&path);
return;
}
d = opendir(path.buf);
if (!d) {
strbuf_release(&path);
return;
}
strbuf_init(&refname, dirnamelen + 257);
strbuf_add(&refname, dirname, dirnamelen);
while ((de = readdir(d)) != NULL) {
unsigned char sha1[20];
struct stat st;
int flag;
if (de->d_name[0] == '.')
continue;
if (ends_with(de->d_name, ".lock"))
continue;
strbuf_addstr(&refname, de->d_name);
strbuf_addstr(&path, de->d_name);
if (stat(path.buf, &st) < 0) {
; /* silently ignore */
} else if (S_ISDIR(st.st_mode)) {
strbuf_addch(&refname, '/');
add_entry_to_dir(dir,
create_dir_entry(refs, refname.buf,
refname.len, 1));
} else {
int read_ok;
if (*refs->name) {
hashclr(sha1);
flag = 0;
read_ok = !resolve_gitlink_ref(refs->name,
refname.buf, sha1);
} else {
read_ok = !read_ref_full(refname.buf,
RESOLVE_REF_READING,
sha1, &flag);
}
if (!read_ok) {
hashclr(sha1);
flag |= REF_ISBROKEN;
} else if (is_null_sha1(sha1)) {
/*
* It is so astronomically unlikely
* that NULL_SHA1 is the SHA-1 of an
* actual object that we consider its
* appearance in a loose reference
* file to be repo corruption
* (probably due to a software bug).
*/
flag |= REF_ISBROKEN;
}
if (check_refname_format(refname.buf,
REFNAME_ALLOW_ONELEVEL)) {
if (!refname_is_safe(refname.buf))
die("loose refname is dangerous: %s", refname.buf);
hashclr(sha1);
flag |= REF_BAD_NAME | REF_ISBROKEN;
}
add_entry_to_dir(dir,
create_ref_entry(refname.buf, sha1, flag, 0));
}
strbuf_setlen(&refname, dirnamelen);
strbuf_setlen(&path, path_baselen);
}
strbuf_release(&refname);
strbuf_release(&path);
closedir(d);
}
static struct ref_dir *get_loose_refs(struct ref_cache *refs)
{
if (!refs->loose) {
/*
* Mark the top-level directory complete because we
* are about to read the only subdirectory that can
* hold references:
*/
refs->loose = create_dir_entry(refs, "", 0, 0);
/*
* Create an incomplete entry for "refs/":
*/
add_entry_to_dir(get_ref_dir(refs->loose),
create_dir_entry(refs, "refs/", 5, 1));
}
return get_ref_dir(refs->loose);
}
#define MAXREFLEN (1024)
/*
* Called by resolve_gitlink_ref_recursive() after it failed to read
* from the loose refs in ref_cache refs. Find <refname> in the
* packed-refs file for the submodule.
*/
static int resolve_gitlink_packed_ref(struct ref_cache *refs,
const char *refname, unsigned char *sha1)
{
struct ref_entry *ref;
struct ref_dir *dir = get_packed_refs(refs);
ref = find_ref(dir, refname);
if (ref == NULL)
return -1;
hashcpy(sha1, ref->u.value.oid.hash);
return 0;
}
static int resolve_gitlink_ref_recursive(struct ref_cache *refs,
const char *refname, unsigned char *sha1,
int recursion)
{
int fd, len;
char buffer[128], *p;
char *path;
if (recursion > SYMREF_MAXDEPTH || strlen(refname) > MAXREFLEN)
return -1;
path = *refs->name
? git_pathdup_submodule(refs->name, "%s", refname)
: git_pathdup("%s", refname);
fd = open(path, O_RDONLY);
free(path);
if (fd < 0)
return resolve_gitlink_packed_ref(refs, refname, sha1);
len = read(fd, buffer, sizeof(buffer)-1);
close(fd);
if (len < 0)
return -1;
while (len && isspace(buffer[len-1]))
len--;
buffer[len] = 0;
/* Was it a detached head or an old-fashioned symlink? */
if (!get_sha1_hex(buffer, sha1))
return 0;
/* Symref? */
if (strncmp(buffer, "ref:", 4))
return -1;
p = buffer + 4;
while (isspace(*p))
p++;
return resolve_gitlink_ref_recursive(refs, p, sha1, recursion+1);
}
int resolve_gitlink_ref(const char *path, const char *refname, unsigned char *sha1)
{
int len = strlen(path), retval;
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
struct strbuf submodule = STRBUF_INIT;
struct ref_cache *refs;
while (len && path[len-1] == '/')
len--;
if (!len)
return -1;
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
strbuf_add(&submodule, path, len);
refs = get_ref_cache(submodule.buf);
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
if (!refs) {
strbuf_release(&submodule);
return -1;
resolve_gitlink_ref: ignore non-repository paths When we want to look up a submodule ref, we use get_ref_cache(path) to find or auto-create its ref cache. But if we feed a path that isn't actually a git repository, we blindly create the ref cache, and then may die deeper in the code when we try to access it. This is a problem because many callers speculatively feed us a path that looks vaguely like a repository, and expect us to tell them when it is not. This patch teaches resolve_gitlink_ref to reject non-repository paths without creating a ref_cache. This avoids the die(), and also performs better if you have a large number of these faux-submodule directories (because the ref_cache lookup is linear, under the assumption that there won't be a large number of submodules). To accomplish this, we also break get_ref_cache into two pieces: the lookup and auto-creation (the latter is lumped into create_ref_cache). This lets us first cheaply ask our cache "is it a submodule we know about?" If so, we can avoid repeating our filesystem lookup. So lookups of real submodules are not penalized; they examine the submodule's .git directory only once. The test in t3000 demonstrates a case where this improves correctness (we used to just die). The new perf case in p7300 shows off the speed improvement in an admittedly pathological repository: Test HEAD^ HEAD ---------------------------------------------------------------- 7300.4: ls-files -o 66.97(66.15+0.87) 0.33(0.08+0.24) -99.5% Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-22 23:29:30 +01:00
}
strbuf_release(&submodule);
retval = resolve_gitlink_ref_recursive(refs, refname, sha1, 0);
return retval;
}
/*
* Return the ref_entry for the given refname from the packed
* references. If it does not exist, return NULL.
*/
static struct ref_entry *get_packed_ref(const char *refname)
{
return find_ref(get_packed_refs(&ref_cache), refname);
}
/*
* A loose ref file doesn't exist; check for a packed ref.
*/
static int resolve_missing_loose_ref(const char *refname,
unsigned char *sha1,
unsigned int *flags)
{
struct ref_entry *entry;
/*
* The loose reference file does not exist; check for a packed
* reference.
*/
entry = get_packed_ref(refname);
if (entry) {
hashcpy(sha1, entry->u.value.oid.hash);
*flags |= REF_ISPACKED;
return 0;
}
/* refname is not a packed reference. */
return -1;
}
int read_raw_ref(const char *refname, unsigned char *sha1,
struct strbuf *referent, unsigned int *type)
{
struct strbuf sb_contents = STRBUF_INIT;
struct strbuf sb_path = STRBUF_INIT;
const char *path;
const char *buf;
struct stat st;
int fd;
int ret = -1;
int save_errno;
*type = 0;
strbuf_reset(&sb_path);
strbuf_git_path(&sb_path, "%s", refname);
path = sb_path.buf;
stat_ref:
/*
* We might have to loop back here to avoid a race
* condition: first we lstat() the file, then we try
* to read it as a link or as a file. But if somebody
* changes the type of the file (file <-> directory
* <-> symlink) between the lstat() and reading, then
* we don't want to report that as an error but rather
* try again starting with the lstat().
*/
if (lstat(path, &st) < 0) {
if (errno != ENOENT)
goto out;
if (resolve_missing_loose_ref(refname, sha1, type)) {
errno = ENOENT;
goto out;
}
ret = 0;
goto out;
}
/* Follow "normalized" - ie "refs/.." symlinks by hand */
if (S_ISLNK(st.st_mode)) {
strbuf_reset(&sb_contents);
if (strbuf_readlink(&sb_contents, path, 0) < 0) {
if (errno == ENOENT || errno == EINVAL)
/* inconsistent with lstat; retry */
goto stat_ref;
else
goto out;
}
if (starts_with(sb_contents.buf, "refs/") &&
!check_refname_format(sb_contents.buf, 0)) {
strbuf_swap(&sb_contents, referent);
*type |= REF_ISSYMREF;
ret = 0;
goto out;
}
}
/* Is it a directory? */
if (S_ISDIR(st.st_mode)) {
/*
* Even though there is a directory where the loose
* ref is supposed to be, there could still be a
* packed ref:
*/
if (resolve_missing_loose_ref(refname, sha1, type)) {
errno = EISDIR;
goto out;
}
ret = 0;
goto out;
}
/*
* Anything else, just open it and try to use it as
* a ref
*/
fd = open(path, O_RDONLY);
if (fd < 0) {
if (errno == ENOENT)
/* inconsistent with lstat; retry */
goto stat_ref;
else
goto out;
}
strbuf_reset(&sb_contents);
if (strbuf_read(&sb_contents, fd, 256) < 0) {
int save_errno = errno;
close(fd);
errno = save_errno;
goto out;
}
close(fd);
strbuf_rtrim(&sb_contents);
buf = sb_contents.buf;
if (starts_with(buf, "ref:")) {
buf += 4;
while (isspace(*buf))
buf++;
strbuf_reset(referent);
strbuf_addstr(referent, buf);
*type |= REF_ISSYMREF;
ret = 0;
goto out;
}
/*
* Please note that FETCH_HEAD has additional
* data after the sha.
*/
if (get_sha1_hex(buf, sha1) ||
(buf[40] != '\0' && !isspace(buf[40]))) {
*type |= REF_ISBROKEN;
errno = EINVAL;
goto out;
}
ret = 0;
out:
save_errno = errno;
strbuf_release(&sb_path);
strbuf_release(&sb_contents);
errno = save_errno;
return ret;
}
static void unlock_ref(struct ref_lock *lock)
{
/* Do not free lock->lk -- atexit() still looks at them */
if (lock->lk)
rollback_lock_file(lock->lk);
free(lock->ref_name);
free(lock);
}
/*
* Lock refname, without following symrefs, and set *lock_p to point
* at a newly-allocated lock object. Fill in lock->old_oid, referent,
* and type similarly to read_raw_ref().
*
* The caller must verify that refname is a "safe" reference name (in
* the sense of refname_is_safe()) before calling this function.
*
* If the reference doesn't already exist, verify that refname doesn't
* have a D/F conflict with any existing references. extras and skip
* are passed to verify_refname_available_dir() for this check.
*
* If mustexist is not set and the reference is not found or is
* broken, lock the reference anyway but clear sha1.
*
* Return 0 on success. On failure, write an error message to err and
* return TRANSACTION_NAME_CONFLICT or TRANSACTION_GENERIC_ERROR.
*
* Implementation note: This function is basically
*
* lock reference
* read_raw_ref()
*
* but it includes a lot more code to
* - Deal with possible races with other processes
* - Avoid calling verify_refname_available_dir() when it can be
* avoided, namely if we were successfully able to read the ref
* - Generate informative error messages in the case of failure
*/
static int lock_raw_ref(const char *refname, int mustexist,
const struct string_list *extras,
const struct string_list *skip,
struct ref_lock **lock_p,
struct strbuf *referent,
unsigned int *type,
struct strbuf *err)
{
struct ref_lock *lock;
struct strbuf ref_file = STRBUF_INIT;
int attempts_remaining = 3;
int ret = TRANSACTION_GENERIC_ERROR;
assert(err);
*type = 0;
/* First lock the file so it can't change out from under us. */
*lock_p = lock = xcalloc(1, sizeof(*lock));
lock->ref_name = xstrdup(refname);
strbuf_git_path(&ref_file, "%s", refname);
retry:
switch (safe_create_leading_directories(ref_file.buf)) {
case SCLD_OK:
break; /* success */
case SCLD_EXISTS:
/*
* Suppose refname is "refs/foo/bar". We just failed
* to create the containing directory, "refs/foo",
* because there was a non-directory in the way. This
* indicates a D/F conflict, probably because of
* another reference such as "refs/foo". There is no
* reason to expect this error to be transitory.
*/
if (verify_refname_available(refname, extras, skip, err)) {
if (mustexist) {
/*
* To the user the relevant error is
* that the "mustexist" reference is
* missing:
*/
strbuf_reset(err);
strbuf_addf(err, "unable to resolve reference '%s'",
refname);
} else {
/*
* The error message set by
* verify_refname_available_dir() is OK.
*/
ret = TRANSACTION_NAME_CONFLICT;
}
} else {
/*
* The file that is in the way isn't a loose
* reference. Report it as a low-level
* failure.
*/
strbuf_addf(err, "unable to create lock file %s.lock; "
"non-directory in the way",
ref_file.buf);
}
goto error_return;
case SCLD_VANISHED:
/* Maybe another process was tidying up. Try again. */
if (--attempts_remaining > 0)
goto retry;
/* fall through */
default:
strbuf_addf(err, "unable to create directory for %s",
ref_file.buf);
goto error_return;
}
if (!lock->lk)
lock->lk = xcalloc(1, sizeof(struct lock_file));
if (hold_lock_file_for_update(lock->lk, ref_file.buf, LOCK_NO_DEREF) < 0) {
if (errno == ENOENT && --attempts_remaining > 0) {
/*
* Maybe somebody just deleted one of the
* directories leading to ref_file. Try
* again:
*/
goto retry;
} else {
unable_to_lock_message(ref_file.buf, errno, err);
goto error_return;
}
}
/*
* Now we hold the lock and can read the reference without
* fear that its value will change.
*/
if (read_raw_ref(refname, lock->old_oid.hash, referent, type)) {
if (errno == ENOENT) {
if (mustexist) {
/* Garden variety missing reference. */
strbuf_addf(err, "unable to resolve reference '%s'",
refname);
goto error_return;
} else {
/*
* Reference is missing, but that's OK. We
* know that there is not a conflict with
* another loose reference because
* (supposing that we are trying to lock
* reference "refs/foo/bar"):
*
* - We were successfully able to create
* the lockfile refs/foo/bar.lock, so we
* know there cannot be a loose reference
* named "refs/foo".
*
* - We got ENOENT and not EISDIR, so we
* know that there cannot be a loose
* reference named "refs/foo/bar/baz".
*/
}
} else if (errno == EISDIR) {
/*
* There is a directory in the way. It might have
* contained references that have been deleted. If
* we don't require that the reference already
* exists, try to remove the directory so that it
* doesn't cause trouble when we want to rename the
* lockfile into place later.
*/
if (mustexist) {
/* Garden variety missing reference. */
strbuf_addf(err, "unable to resolve reference '%s'",
refname);
goto error_return;
} else if (remove_dir_recursively(&ref_file,
REMOVE_DIR_EMPTY_ONLY)) {
if (verify_refname_available_dir(
refname, extras, skip,
get_loose_refs(&ref_cache),
err)) {
/*
* The error message set by
* verify_refname_available() is OK.
*/
ret = TRANSACTION_NAME_CONFLICT;
goto error_return;
} else {
/*
* We can't delete the directory,
* but we also don't know of any
* references that it should
* contain.
*/
strbuf_addf(err, "there is a non-empty directory '%s' "
"blocking reference '%s'",
ref_file.buf, refname);
goto error_return;
}
}
} else if (errno == EINVAL && (*type & REF_ISBROKEN)) {
strbuf_addf(err, "unable to resolve reference '%s': "
"reference broken", refname);
goto error_return;
} else {
strbuf_addf(err, "unable to resolve reference '%s': %s",
refname, strerror(errno));
goto error_return;
}
/*
* If the ref did not exist and we are creating it,
* make sure there is no existing packed ref whose
* name begins with our refname, nor a packed ref
* whose name is a proper prefix of our refname.
*/
if (verify_refname_available_dir(
refname, extras, skip,
get_packed_refs(&ref_cache),
err)) {
goto error_return;
}
}
ret = 0;
goto out;
error_return:
unlock_ref(lock);
*lock_p = NULL;
out:
strbuf_release(&ref_file);
return ret;
}
/*
* Peel the entry (if possible) and return its new peel_status. If
* repeel is true, re-peel the entry even if there is an old peeled
* value that is already stored in it.
*
* It is OK to call this function with a packed reference entry that
* might be stale and might even refer to an object that has since
* been garbage-collected. In such a case, if the entry has
* REF_KNOWS_PEELED then leave the status unchanged and return
* PEEL_PEELED or PEEL_NON_TAG; otherwise, return PEEL_INVALID.
*/
static enum peel_status peel_entry(struct ref_entry *entry, int repeel)
{
enum peel_status status;
if (entry->flag & REF_KNOWS_PEELED) {
if (repeel) {
entry->flag &= ~REF_KNOWS_PEELED;
oidclr(&entry->u.value.peeled);
} else {
return is_null_oid(&entry->u.value.peeled) ?
PEEL_NON_TAG : PEEL_PEELED;
}
}
if (entry->flag & REF_ISBROKEN)
return PEEL_BROKEN;
if (entry->flag & REF_ISSYMREF)
return PEEL_IS_SYMREF;
status = peel_object(entry->u.value.oid.hash, entry->u.value.peeled.hash);
if (status == PEEL_PEELED || status == PEEL_NON_TAG)
entry->flag |= REF_KNOWS_PEELED;
return status;
}
int peel_ref(const char *refname, unsigned char *sha1)
{
int flag;
unsigned char base[20];
do_for_each_ref(): reimplement using reference iteration Use the reference iterator interface to implement do_for_each_ref(). Delete a bunch of code supporting the old for_each_ref() implementation. And now that do_for_each_ref() is generic code (it is no longer tied to the files backend), move it to refs.c. The implementation is via a new function, do_for_each_ref_iterator(), which takes a reference iterator as argument and calls a callback function for each of the references in the iterator. This change requires the current_ref performance hack for peel_ref() to be implemented via ref_iterator_peel() rather than peel_entry() because we don't have a ref_entry handy (it is hidden under three layers: file_ref_iterator, merge_ref_iterator, and cache_ref_iterator). So: * do_for_each_ref_iterator() records the active iterator in current_ref_iter while it is running. * peel_ref() checks whether current_ref_iter is pointing at the requested reference. If so, it asks the iterator to peel the reference (which it can do efficiently via its "peel" virtual function). For extra safety, we do the optimization only if the refname *addresses* are the same, not only if the refname *strings* are the same, to forestall possible mixups between refnames that come from different ref_iterators. Please note that this optimization of peel_ref() is only available when iterating via do_for_each_ref_iterator() (including all of the for_each_ref() functions, which call it indirectly). It would be complicated to implement a similar optimization when iterating directly using a reference iterator, because multiple reference iterators can be in use at the same time, with interleaved calls to ref_iterator_advance(). (In fact we do exactly that in merge_ref_iterator.) But that is not necessary. peel_ref() is only called while iterating over references. Callers who iterate using the for_each_ref() functions benefit from the optimization described above. Callers who iterate using reference iterators directly have access to the ref_iterator, so they can call ref_iterator_peel() themselves to get an analogous optimization in a more straightforward manner. If we rewrite all callers to use the reference iteration API, then we can remove the current_ref_iter hack permanently. Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:16 +02:00
if (current_ref_iter && current_ref_iter->refname == refname) {
struct object_id peeled;
if (ref_iterator_peel(current_ref_iter, &peeled))
return -1;
do_for_each_ref(): reimplement using reference iteration Use the reference iterator interface to implement do_for_each_ref(). Delete a bunch of code supporting the old for_each_ref() implementation. And now that do_for_each_ref() is generic code (it is no longer tied to the files backend), move it to refs.c. The implementation is via a new function, do_for_each_ref_iterator(), which takes a reference iterator as argument and calls a callback function for each of the references in the iterator. This change requires the current_ref performance hack for peel_ref() to be implemented via ref_iterator_peel() rather than peel_entry() because we don't have a ref_entry handy (it is hidden under three layers: file_ref_iterator, merge_ref_iterator, and cache_ref_iterator). So: * do_for_each_ref_iterator() records the active iterator in current_ref_iter while it is running. * peel_ref() checks whether current_ref_iter is pointing at the requested reference. If so, it asks the iterator to peel the reference (which it can do efficiently via its "peel" virtual function). For extra safety, we do the optimization only if the refname *addresses* are the same, not only if the refname *strings* are the same, to forestall possible mixups between refnames that come from different ref_iterators. Please note that this optimization of peel_ref() is only available when iterating via do_for_each_ref_iterator() (including all of the for_each_ref() functions, which call it indirectly). It would be complicated to implement a similar optimization when iterating directly using a reference iterator, because multiple reference iterators can be in use at the same time, with interleaved calls to ref_iterator_advance(). (In fact we do exactly that in merge_ref_iterator.) But that is not necessary. peel_ref() is only called while iterating over references. Callers who iterate using the for_each_ref() functions benefit from the optimization described above. Callers who iterate using reference iterators directly have access to the ref_iterator, so they can call ref_iterator_peel() themselves to get an analogous optimization in a more straightforward manner. If we rewrite all callers to use the reference iteration API, then we can remove the current_ref_iter hack permanently. Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:16 +02:00
hashcpy(sha1, peeled.hash);
return 0;
}
if (read_ref_full(refname, RESOLVE_REF_READING, base, &flag))
return -1;
/*
* If the reference is packed, read its ref_entry from the
* cache in the hope that we already know its peeled value.
* We only try this optimization on packed references because
* (a) forcing the filling of the loose reference cache could
* be expensive and (b) loose references anyway usually do not
* have REF_KNOWS_PEELED.
*/
if (flag & REF_ISPACKED) {
struct ref_entry *r = get_packed_ref(refname);
if (r) {
if (peel_entry(r, 0))
return -1;
hashcpy(sha1, r->u.value.peeled.hash);
return 0;
}
}
return peel_object(base, sha1);
}
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
struct files_ref_iterator {
struct ref_iterator base;
struct packed_ref_cache *packed_ref_cache;
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
struct ref_iterator *iter0;
unsigned int flags;
};
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
static int files_ref_iterator_advance(struct ref_iterator *ref_iterator)
{
struct files_ref_iterator *iter =
(struct files_ref_iterator *)ref_iterator;
int ok;
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
while ((ok = ref_iterator_advance(iter->iter0)) == ITER_OK) {
if (!(iter->flags & DO_FOR_EACH_INCLUDE_BROKEN) &&
!ref_resolves_to_object(iter->iter0->refname,
iter->iter0->oid,
iter->iter0->flags))
continue;
iter->base.refname = iter->iter0->refname;
iter->base.oid = iter->iter0->oid;
iter->base.flags = iter->iter0->flags;
return ITER_OK;
}
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
iter->iter0 = NULL;
if (ref_iterator_abort(ref_iterator) != ITER_DONE)
ok = ITER_ERROR;
return ok;
}
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
static int files_ref_iterator_peel(struct ref_iterator *ref_iterator,
struct object_id *peeled)
{
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
struct files_ref_iterator *iter =
(struct files_ref_iterator *)ref_iterator;
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
return ref_iterator_peel(iter->iter0, peeled);
}
static int files_ref_iterator_abort(struct ref_iterator *ref_iterator)
{
struct files_ref_iterator *iter =
(struct files_ref_iterator *)ref_iterator;
int ok = ITER_DONE;
if (iter->iter0)
ok = ref_iterator_abort(iter->iter0);
release_packed_ref_cache(iter->packed_ref_cache);
base_ref_iterator_free(ref_iterator);
return ok;
}
static struct ref_iterator_vtable files_ref_iterator_vtable = {
files_ref_iterator_advance,
files_ref_iterator_peel,
files_ref_iterator_abort
};
struct ref_iterator *files_ref_iterator_begin(
const char *submodule,
const char *prefix, unsigned int flags)
{
struct ref_cache *refs = get_ref_cache(submodule);
struct ref_dir *loose_dir, *packed_dir;
struct ref_iterator *loose_iter, *packed_iter;
struct files_ref_iterator *iter;
struct ref_iterator *ref_iterator;
if (!refs)
return empty_ref_iterator_begin();
if (ref_paranoia < 0)
ref_paranoia = git_env_bool("GIT_REF_PARANOIA", 0);
if (ref_paranoia)
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
flags |= DO_FOR_EACH_INCLUDE_BROKEN;
iter = xcalloc(1, sizeof(*iter));
ref_iterator = &iter->base;
base_ref_iterator_init(ref_iterator, &files_ref_iterator_vtable);
/*
* We must make sure that all loose refs are read before
* accessing the packed-refs file; this avoids a race
* condition if loose refs are migrated to the packed-refs
* file by a simultaneous process, but our in-memory view is
* from before the migration. We ensure this as follows:
* First, we call prime_ref_dir(), which pre-reads the loose
* references for the subtree into the cache. (If they've
* already been read, that's OK; we only need to guarantee
* that they're read before the packed refs, not *how much*
* before.) After that, we call get_packed_ref_cache(), which
* internally checks whether the packed-ref cache is up to
* date with what is on disk, and re-reads it if not.
*/
loose_dir = get_loose_refs(refs);
if (prefix && *prefix)
loose_dir = find_containing_dir(loose_dir, prefix, 0);
refs: introduce an iterator interface Currently, the API for iterating over references is via a family of for_each_ref()-type functions that invoke a callback function for each selected reference. All of these eventually call do_for_each_ref(), which knows how to do one thing: iterate in parallel through two ref_caches, one for loose and one for packed refs, giving loose references precedence over packed refs. This is rather complicated code, and is quite specialized to the files backend. It also requires callers to encapsulate their work into a callback function, which often means that they have to define and use a "cb_data" struct to manage their context. The current design is already bursting at the seams, and will become even more awkward in the upcoming world of multiple reference storage backends: * Per-worktree vs. shared references are currently handled via a kludge in git_path() rather than iterating over each part of the reference namespace separately and merging the results. This kludge will cease to work when we have multiple reference storage backends. * The current scheme is inflexible. What if we sometimes want to bypass the ref_cache, or use it only for packed or only for loose refs? What if we want to store symbolic refs in one type of storage backend and non-symbolic ones in another? In the future, each reference backend will need to define its own way of iterating over references. The crux of the problem with the current design is that it is impossible to compose for_each_ref()-style iterations, because the flow of control is owned by the for_each_ref() function. There is nothing that a caller can do but iterate through all references in a single burst, so there is no way for it to interleave references from multiple backends and present the result to the rest of the world as a single compound backend. This commit introduces a new iteration primitive for references: a ref_iterator. A ref_iterator is a polymorphic object that a reference storage backend can be asked to instantiate. There are three functions that can be applied to a ref_iterator: * ref_iterator_advance(): move to the next reference in the iteration * ref_iterator_abort(): end the iteration before it is exhausted * ref_iterator_peel(): peel the reference currently being looked at Iterating using a ref_iterator leaves the flow of control in the hands of the caller, which means that ref_iterators from multiple sources (e.g., loose and packed refs) can be composed and presented to the world as a single compound ref_iterator. It also means that the backend code for implementing reference iteration will sometimes be more complicated. For example, the cache_ref_iterator (which iterates over a ref_cache) can't use the C stack to recurse; instead, it must manage its own stack internally as explicit data structures. There is also a lot of boilerplate connected with object-oriented programming in C. Eventually, end-user callers will be able to be written in a more natural way—managing their own flow of control rather than having to work via callbacks. Since there will only be a few reference backends but there are many consumers of this API, this is a good tradeoff. More importantly, we gain composability, and especially the possibility of writing interchangeable parts that can work with any ref_iterator. For example, merge_ref_iterator implements a generic way of merging the contents of any two ref_iterators. It is used to merge loose + packed refs as part of the implementation of the files_ref_iterator. But it will also be possible to use it to merge other pairs of reference sources (e.g., per-worktree vs. shared refs). Another example is prefix_ref_iterator, which can be used to trim a prefix off the front of reference names before presenting them to the caller (e.g., "refs/heads/master" -> "master"). In this patch, we introduce the iterator abstraction and many utilities, and implement a reference iterator for the files ref storage backend. (I've written several other obvious utilities, for example a generic way to filter references being iterated over. These will probably be useful in the future. But they are not needed for this patch series, so I am not including them at this time.) In a moment we will rewrite do_for_each_ref() to work via reference iterators (allowing some special-purpose code to be discarded), and do something similar for reflogs. In future patch series, we will expose the ref_iterator abstraction in the public refs API so that callers can use it directly. Implementation note: I tried abstracting this a layer further to allow generic iterators (over arbitrary types of objects) and generic utilities like a generic merge_iterator. But the implementation in C was very cumbersome, involving (in my opinion) too much boilerplate and too much unsafe casting, some of which would have had to be done on the caller side. However, I did put a few iterator-related constants in a top-level header file, iterator.h, as they will be useful in a moment to implement iteration over directory trees and possibly other types of iterators in the future. Signed-off-by: Ramsay Jones <ramsay@ramsayjones.plus.com> Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-06-18 06:15:15 +02:00
if (loose_dir) {
prime_ref_dir(loose_dir);
loose_iter = cache_ref_iterator_begin(loose_dir);
} else {
/* There's nothing to iterate over. */
loose_iter = empty_ref_iterator_begin();
}
iter->packed_ref_cache = get_packed_ref_cache(refs);
acquire_packed_ref_cache(iter->packed_ref_cache);
packed_dir = get_packed_ref_dir(iter->packed_ref_cache);
if (prefix && *prefix)
packed_dir = find_containing_dir(packed_dir, prefix, 0);
if (packed_dir) {
packed_iter = cache_ref_iterator_begin(packed_dir);
} else {
/* There's nothing to iterate over. */
packed_iter = empty_ref_iterator_begin();
}
iter->iter0 = overlay_ref_iterator_begin(loose_iter, packed_iter);
iter->flags = flags;
return ref_iterator;
}
/*
* Verify that the reference locked by lock has the value old_sha1.
* Fail if the reference doesn't exist and mustexist is set. Return 0
* on success. On error, write an error message to err, set errno, and
* return a negative value.
*/
static int verify_lock(struct ref_lock *lock,
const unsigned char *old_sha1, int mustexist,
struct strbuf *err)
{
assert(err);
if (read_ref_full(lock->ref_name,
mustexist ? RESOLVE_REF_READING : 0,
lock->old_oid.hash, NULL)) {
lock_ref_sha1_basic: always fill old_oid while holding lock Our basic strategy for taking a ref lock is: 1. Create $ref.lock to take the lock 2. Read the ref again while holding the lock (during which time we know that nobody else can be updating it). 3. Compare the value we read to the expected "old_sha1" The value we read in step (2) is returned to the caller via the lock->old_oid field, who may use it for other purposes (such as writing a reflog). If we have no "old_sha1" (i.e., we are unconditionally taking the lock), then we obviously must omit step 3. But we _also_ omit step 2. This seems like a nice optimization, but it means that the caller sees only whatever was left in lock->old_oid from previous calls to resolve_ref_unsafe(), which happened outside of the lock. We can demonstrate this race pretty easily. Imagine you have three commits, $one, $two, and $three. One script just flips between $one and $two, without providing an old-sha1: while true; do git update-ref -m one refs/heads/foo $one git update-ref -m two refs/heads/foo $two done Meanwhile, another script tries to set the value to $three, also not using an old-sha1: while true; do git update-ref -m three refs/heads/foo $three done If these run simultaneously, we'll see a lot of lock contention, but each of the writes will succeed some of the time. The reflog may record movements between any of the three refs, but we would expect it to provide a consistent log: the "from" field of each log entry should be the same as the "to" field of the previous one. But if we check this: perl -alne ' print "mismatch on line $." if defined $last && $F[0] ne $last; $last = $F[1]; ' .git/logs/refs/heads/foo we'll see many mismatches. Why? Because sometimes, in the time between lock_ref_sha1_basic filling lock->old_oid via resolve_ref_unsafe() and it taking the lock, there may be a complete write by another process. And the "from" field in our reflog entry will be wrong, and will refer to an older value. This is probably quite rare in practice. It requires writers which do not provide an old-sha1 value, and it is a very quick race. However, it is easy to fix: we simply perform step (2), the read-under-lock, whether we have an old-sha1 or not. Then the value we hand back to the caller is always atomic. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:44:39 +01:00
if (old_sha1) {
int save_errno = errno;
strbuf_addf(err, "can't verify ref '%s'", lock->ref_name);
lock_ref_sha1_basic: always fill old_oid while holding lock Our basic strategy for taking a ref lock is: 1. Create $ref.lock to take the lock 2. Read the ref again while holding the lock (during which time we know that nobody else can be updating it). 3. Compare the value we read to the expected "old_sha1" The value we read in step (2) is returned to the caller via the lock->old_oid field, who may use it for other purposes (such as writing a reflog). If we have no "old_sha1" (i.e., we are unconditionally taking the lock), then we obviously must omit step 3. But we _also_ omit step 2. This seems like a nice optimization, but it means that the caller sees only whatever was left in lock->old_oid from previous calls to resolve_ref_unsafe(), which happened outside of the lock. We can demonstrate this race pretty easily. Imagine you have three commits, $one, $two, and $three. One script just flips between $one and $two, without providing an old-sha1: while true; do git update-ref -m one refs/heads/foo $one git update-ref -m two refs/heads/foo $two done Meanwhile, another script tries to set the value to $three, also not using an old-sha1: while true; do git update-ref -m three refs/heads/foo $three done If these run simultaneously, we'll see a lot of lock contention, but each of the writes will succeed some of the time. The reflog may record movements between any of the three refs, but we would expect it to provide a consistent log: the "from" field of each log entry should be the same as the "to" field of the previous one. But if we check this: perl -alne ' print "mismatch on line $." if defined $last && $F[0] ne $last; $last = $F[1]; ' .git/logs/refs/heads/foo we'll see many mismatches. Why? Because sometimes, in the time between lock_ref_sha1_basic filling lock->old_oid via resolve_ref_unsafe() and it taking the lock, there may be a complete write by another process. And the "from" field in our reflog entry will be wrong, and will refer to an older value. This is probably quite rare in practice. It requires writers which do not provide an old-sha1 value, and it is a very quick race. However, it is easy to fix: we simply perform step (2), the read-under-lock, whether we have an old-sha1 or not. Then the value we hand back to the caller is always atomic. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:44:39 +01:00
errno = save_errno;
return -1;
} else {
oidclr(&lock->old_oid);
lock_ref_sha1_basic: always fill old_oid while holding lock Our basic strategy for taking a ref lock is: 1. Create $ref.lock to take the lock 2. Read the ref again while holding the lock (during which time we know that nobody else can be updating it). 3. Compare the value we read to the expected "old_sha1" The value we read in step (2) is returned to the caller via the lock->old_oid field, who may use it for other purposes (such as writing a reflog). If we have no "old_sha1" (i.e., we are unconditionally taking the lock), then we obviously must omit step 3. But we _also_ omit step 2. This seems like a nice optimization, but it means that the caller sees only whatever was left in lock->old_oid from previous calls to resolve_ref_unsafe(), which happened outside of the lock. We can demonstrate this race pretty easily. Imagine you have three commits, $one, $two, and $three. One script just flips between $one and $two, without providing an old-sha1: while true; do git update-ref -m one refs/heads/foo $one git update-ref -m two refs/heads/foo $two done Meanwhile, another script tries to set the value to $three, also not using an old-sha1: while true; do git update-ref -m three refs/heads/foo $three done If these run simultaneously, we'll see a lot of lock contention, but each of the writes will succeed some of the time. The reflog may record movements between any of the three refs, but we would expect it to provide a consistent log: the "from" field of each log entry should be the same as the "to" field of the previous one. But if we check this: perl -alne ' print "mismatch on line $." if defined $last && $F[0] ne $last; $last = $F[1]; ' .git/logs/refs/heads/foo we'll see many mismatches. Why? Because sometimes, in the time between lock_ref_sha1_basic filling lock->old_oid via resolve_ref_unsafe() and it taking the lock, there may be a complete write by another process. And the "from" field in our reflog entry will be wrong, and will refer to an older value. This is probably quite rare in practice. It requires writers which do not provide an old-sha1 value, and it is a very quick race. However, it is easy to fix: we simply perform step (2), the read-under-lock, whether we have an old-sha1 or not. Then the value we hand back to the caller is always atomic. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:44:39 +01:00
return 0;
}
}
lock_ref_sha1_basic: always fill old_oid while holding lock Our basic strategy for taking a ref lock is: 1. Create $ref.lock to take the lock 2. Read the ref again while holding the lock (during which time we know that nobody else can be updating it). 3. Compare the value we read to the expected "old_sha1" The value we read in step (2) is returned to the caller via the lock->old_oid field, who may use it for other purposes (such as writing a reflog). If we have no "old_sha1" (i.e., we are unconditionally taking the lock), then we obviously must omit step 3. But we _also_ omit step 2. This seems like a nice optimization, but it means that the caller sees only whatever was left in lock->old_oid from previous calls to resolve_ref_unsafe(), which happened outside of the lock. We can demonstrate this race pretty easily. Imagine you have three commits, $one, $two, and $three. One script just flips between $one and $two, without providing an old-sha1: while true; do git update-ref -m one refs/heads/foo $one git update-ref -m two refs/heads/foo $two done Meanwhile, another script tries to set the value to $three, also not using an old-sha1: while true; do git update-ref -m three refs/heads/foo $three done If these run simultaneously, we'll see a lot of lock contention, but each of the writes will succeed some of the time. The reflog may record movements between any of the three refs, but we would expect it to provide a consistent log: the "from" field of each log entry should be the same as the "to" field of the previous one. But if we check this: perl -alne ' print "mismatch on line $." if defined $last && $F[0] ne $last; $last = $F[1]; ' .git/logs/refs/heads/foo we'll see many mismatches. Why? Because sometimes, in the time between lock_ref_sha1_basic filling lock->old_oid via resolve_ref_unsafe() and it taking the lock, there may be a complete write by another process. And the "from" field in our reflog entry will be wrong, and will refer to an older value. This is probably quite rare in practice. It requires writers which do not provide an old-sha1 value, and it is a very quick race. However, it is easy to fix: we simply perform step (2), the read-under-lock, whether we have an old-sha1 or not. Then the value we hand back to the caller is always atomic. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:44:39 +01:00
if (old_sha1 && hashcmp(lock->old_oid.hash, old_sha1)) {
strbuf_addf(err, "ref '%s' is at %s but expected %s",
lock->ref_name,
oid_to_hex(&lock->old_oid),
sha1_to_hex(old_sha1));
errno = EBUSY;
return -1;
}
return 0;
}
static int remove_empty_directories(struct strbuf *path)
{
/*
* we want to create a file but there is a directory there;
* if that is an empty directory (or a directory that contains
* only empty directories), remove them.
*/
return remove_dir_recursively(path, REMOVE_DIR_EMPTY_ONLY);
}
/*
* Locks a ref returning the lock on success and NULL on failure.
* On failure errno is set to something meaningful.
*/
static struct ref_lock *lock_ref_sha1_basic(const char *refname,
const unsigned char *old_sha1,
const struct string_list *extras,
const struct string_list *skip,
unsigned int flags, int *type,
struct strbuf *err)
{
struct strbuf ref_file = STRBUF_INIT;
struct ref_lock *lock;
int last_errno = 0;
int lflags = LOCK_NO_DEREF;
int mustexist = (old_sha1 && !is_null_sha1(old_sha1));
int resolve_flags = RESOLVE_REF_NO_RECURSE;
int attempts_remaining = 3;
int resolved;
assert(err);
lock = xcalloc(1, sizeof(struct ref_lock));
if (mustexist)
resolve_flags |= RESOLVE_REF_READING;
lock_ref_sha1_basic: handle REF_NODEREF with invalid refs We sometimes call lock_ref_sha1_basic with REF_NODEREF to operate directly on a symbolic ref. This is used, for example, to move to a detached HEAD, or when updating the contents of HEAD via checkout or symbolic-ref. However, the first step of the function is to resolve the refname to get the "old" sha1, and we do so without telling resolve_ref_unsafe() that we are only interested in the symref. As a result, we may detect a problem there not with the symref itself, but with something it points to. The real-world example I found (and what is used in the test suite) is a HEAD pointing to a ref that cannot exist, because it would cause a directory/file conflict with other existing refs. This situation is somewhat broken, of course, as trying to _commit_ on that HEAD would fail. But it's not explicitly forbidden, and we should be able to move away from it. However, neither "git checkout" nor "git symbolic-ref" can do so. We try to take the lock on HEAD, which is pointing to a non-existent ref. We bail from resolve_ref_unsafe() with errno set to EISDIR, and the lock code thinks we are attempting to create a d/f conflict. Of course we're not. The problem is that the lock code has no idea what level we were at when we got EISDIR, so trying to diagnose or remove empty directories for HEAD is not useful. To make things even more complicated, we only get EISDIR in the loose-ref case. If the refs are packed, the resolution may "succeed", giving us the pointed-to ref in "refname", but a null oid. Later, we say "ah, the null oid means we are creating; let's make sure there is room for it", but mistakenly check against the _resolved_ refname, not the original. We can fix this by making two tweaks: 1. Call resolve_ref_unsafe() with RESOLVE_REF_NO_RECURSE when REF_NODEREF is set. This means any errors we get will be from the orig_refname, and we can act accordingly. We already do this in the REF_DELETING case, but we should do it for update, too. 2. If we do get a "refname" return from resolve_ref_unsafe(), even with RESOLVE_REF_NO_RECURSE it may be the name of the ref pointed-to by a symref. We already normalize this back to orig_refname before taking the lockfile, but we need to do so before the null_oid check. While we're rearranging the REF_NODEREF handling, we can also bump the initialization of lflags to the top of the function, where we are setting up other flags. This saves us from having yet another conditional block on REF_NODEREF just to set it later. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:45:09 +01:00
if (flags & REF_DELETING)
resolve_flags |= RESOLVE_REF_ALLOW_BAD_NAME;
strbuf_git_path(&ref_file, "%s", refname);
resolved = !!resolve_ref_unsafe(refname, resolve_flags,
lock->old_oid.hash, type);
if (!resolved && errno == EISDIR) {
/*
* we are trying to lock foo but we used to
* have foo/bar which now does not exist;
* it is normal for the empty directory 'foo'
* to remain.
*/
if (remove_empty_directories(&ref_file)) {
last_errno = errno;
if (!verify_refname_available_dir(refname, extras, skip,
get_loose_refs(&ref_cache), err))
strbuf_addf(err, "there are still refs under '%s'",
refname);
goto error_return;
}
resolved = !!resolve_ref_unsafe(refname, resolve_flags,
lock->old_oid.hash, type);
}
if (!resolved) {
last_errno = errno;
if (last_errno != ENOTDIR ||
!verify_refname_available_dir(refname, extras, skip,
get_loose_refs(&ref_cache), err))
strbuf_addf(err, "unable to resolve reference '%s': %s",
refname, strerror(last_errno));
goto error_return;
}
lock_ref_sha1_basic: handle REF_NODEREF with invalid refs We sometimes call lock_ref_sha1_basic with REF_NODEREF to operate directly on a symbolic ref. This is used, for example, to move to a detached HEAD, or when updating the contents of HEAD via checkout or symbolic-ref. However, the first step of the function is to resolve the refname to get the "old" sha1, and we do so without telling resolve_ref_unsafe() that we are only interested in the symref. As a result, we may detect a problem there not with the symref itself, but with something it points to. The real-world example I found (and what is used in the test suite) is a HEAD pointing to a ref that cannot exist, because it would cause a directory/file conflict with other existing refs. This situation is somewhat broken, of course, as trying to _commit_ on that HEAD would fail. But it's not explicitly forbidden, and we should be able to move away from it. However, neither "git checkout" nor "git symbolic-ref" can do so. We try to take the lock on HEAD, which is pointing to a non-existent ref. We bail from resolve_ref_unsafe() with errno set to EISDIR, and the lock code thinks we are attempting to create a d/f conflict. Of course we're not. The problem is that the lock code has no idea what level we were at when we got EISDIR, so trying to diagnose or remove empty directories for HEAD is not useful. To make things even more complicated, we only get EISDIR in the loose-ref case. If the refs are packed, the resolution may "succeed", giving us the pointed-to ref in "refname", but a null oid. Later, we say "ah, the null oid means we are creating; let's make sure there is room for it", but mistakenly check against the _resolved_ refname, not the original. We can fix this by making two tweaks: 1. Call resolve_ref_unsafe() with RESOLVE_REF_NO_RECURSE when REF_NODEREF is set. This means any errors we get will be from the orig_refname, and we can act accordingly. We already do this in the REF_DELETING case, but we should do it for update, too. 2. If we do get a "refname" return from resolve_ref_unsafe(), even with RESOLVE_REF_NO_RECURSE it may be the name of the ref pointed-to by a symref. We already normalize this back to orig_refname before taking the lockfile, but we need to do so before the null_oid check. While we're rearranging the REF_NODEREF handling, we can also bump the initialization of lflags to the top of the function, where we are setting up other flags. This saves us from having yet another conditional block on REF_NODEREF just to set it later. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:45:09 +01:00
/*
* If the ref did not exist and we are creating it, make sure
* there is no existing packed ref whose name begins with our
* refname, nor a packed ref whose name is a proper prefix of
* our refname.
*/
if (is_null_oid(&lock->old_oid) &&
verify_refname_available_dir(refname, extras, skip,
get_packed_refs(&ref_cache), err)) {
last_errno = ENOTDIR;
goto error_return;
}
lock->lk = xcalloc(1, sizeof(struct lock_file));
lock->ref_name = xstrdup(refname);
retry:
switch (safe_create_leading_directories_const(ref_file.buf)) {
case SCLD_OK:
break; /* success */
case SCLD_VANISHED:
if (--attempts_remaining > 0)
goto retry;
/* fall through */
default:
last_errno = errno;
strbuf_addf(err, "unable to create directory for '%s'",
ref_file.buf);
goto error_return;
}
if (hold_lock_file_for_update(lock->lk, ref_file.buf, lflags) < 0) {
last_errno = errno;
if (errno == ENOENT && --attempts_remaining > 0)
/*
* Maybe somebody just deleted one of the
* directories leading to ref_file. Try
* again:
*/
goto retry;
else {
unable_to_lock_message(ref_file.buf, errno, err);
goto error_return;
}
}
lock_ref_sha1_basic: always fill old_oid while holding lock Our basic strategy for taking a ref lock is: 1. Create $ref.lock to take the lock 2. Read the ref again while holding the lock (during which time we know that nobody else can be updating it). 3. Compare the value we read to the expected "old_sha1" The value we read in step (2) is returned to the caller via the lock->old_oid field, who may use it for other purposes (such as writing a reflog). If we have no "old_sha1" (i.e., we are unconditionally taking the lock), then we obviously must omit step 3. But we _also_ omit step 2. This seems like a nice optimization, but it means that the caller sees only whatever was left in lock->old_oid from previous calls to resolve_ref_unsafe(), which happened outside of the lock. We can demonstrate this race pretty easily. Imagine you have three commits, $one, $two, and $three. One script just flips between $one and $two, without providing an old-sha1: while true; do git update-ref -m one refs/heads/foo $one git update-ref -m two refs/heads/foo $two done Meanwhile, another script tries to set the value to $three, also not using an old-sha1: while true; do git update-ref -m three refs/heads/foo $three done If these run simultaneously, we'll see a lot of lock contention, but each of the writes will succeed some of the time. The reflog may record movements between any of the three refs, but we would expect it to provide a consistent log: the "from" field of each log entry should be the same as the "to" field of the previous one. But if we check this: perl -alne ' print "mismatch on line $." if defined $last && $F[0] ne $last; $last = $F[1]; ' .git/logs/refs/heads/foo we'll see many mismatches. Why? Because sometimes, in the time between lock_ref_sha1_basic filling lock->old_oid via resolve_ref_unsafe() and it taking the lock, there may be a complete write by another process. And the "from" field in our reflog entry will be wrong, and will refer to an older value. This is probably quite rare in practice. It requires writers which do not provide an old-sha1 value, and it is a very quick race. However, it is easy to fix: we simply perform step (2), the read-under-lock, whether we have an old-sha1 or not. Then the value we hand back to the caller is always atomic. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2016-01-12 22:44:39 +01:00
if (verify_lock(lock, old_sha1, mustexist, err)) {
last_errno = errno;
goto error_return;
}
goto out;
error_return:
unlock_ref(lock);
lock = NULL;
out:
strbuf_release(&ref_file);
errno = last_errno;
return lock;
}
/*
* Write an entry to the packed-refs file for the specified refname.
* If peeled is non-NULL, write it as the entry's peeled value.
*/
static void write_packed_entry(FILE *fh, char *refname, unsigned char *sha1,
unsigned char *peeled)
{
fprintf_or_die(fh, "%s %s\n", sha1_to_hex(sha1), refname);
if (peeled)
fprintf_or_die(fh, "^%s\n", sha1_to_hex(peeled));
}
/*
* An each_ref_entry_fn that writes the entry to a packed-refs file.
*/
static int write_packed_entry_fn(struct ref_entry *entry, void *cb_data)
{
enum peel_status peel_status = peel_entry(entry, 0);
if (peel_status != PEEL_PEELED && peel_status != PEEL_NON_TAG)
error("internal error: %s is not a valid packed reference!",
entry->name);
write_packed_entry(cb_data, entry->name, entry->u.value.oid.hash,
peel_status == PEEL_PEELED ?
entry->u.value.peeled.hash : NULL);
return 0;
}
/*
* Lock the packed-refs file for writing. Flags is passed to
* hold_lock_file_for_update(). Return 0 on success. On errors, set
* errno appropriately and return a nonzero value.
*/
static int lock_packed_refs(int flags)
{
static int timeout_configured = 0;
static int timeout_value = 1000;
struct packed_ref_cache *packed_ref_cache;
if (!timeout_configured) {
git_config_get_int("core.packedrefstimeout", &timeout_value);
timeout_configured = 1;
}
if (hold_lock_file_for_update_timeout(
&packlock, git_path("packed-refs"),
flags, timeout_value) < 0)
return -1;
/*
* Get the current packed-refs while holding the lock. If the
* packed-refs file has been modified since we last read it,
* this will automatically invalidate the cache and re-read
* the packed-refs file.
*/
packed_ref_cache = get_packed_ref_cache(&ref_cache);
packed_ref_cache->lock = &packlock;
/* Increment the reference count to prevent it from being freed: */
acquire_packed_ref_cache(packed_ref_cache);
return 0;
}
/*
* Write the current version of the packed refs cache from memory to
* disk. The packed-refs file must already be locked for writing (see
* lock_packed_refs()). Return zero on success. On errors, set errno
* and return a nonzero value
*/
static int commit_packed_refs(void)
{
struct packed_ref_cache *packed_ref_cache =
get_packed_ref_cache(&ref_cache);
int error = 0;
int save_errno = 0;
FILE *out;
if (!packed_ref_cache->lock)
die("internal error: packed-refs not locked");
out = fdopen_lock_file(packed_ref_cache->lock, "w");
if (!out)
die_errno("unable to fdopen packed-refs descriptor");
fprintf_or_die(out, "%s", PACKED_REFS_HEADER);
do_for_each_entry_in_dir(get_packed_ref_dir(packed_ref_cache),
0, write_packed_entry_fn, out);
if (commit_lock_file(packed_ref_cache->lock)) {
save_errno = errno;
error = -1;
}
packed_ref_cache->lock = NULL;
release_packed_ref_cache(packed_ref_cache);
errno = save_errno;
return error;
}
/*
* Rollback the lockfile for the packed-refs file, and discard the
* in-memory packed reference cache. (The packed-refs file will be
* read anew if it is needed again after this function is called.)
*/
static void rollback_packed_refs(void)
{
struct packed_ref_cache *packed_ref_cache =
get_packed_ref_cache(&ref_cache);
if (!packed_ref_cache->lock)
die("internal error: packed-refs not locked");
rollback_lock_file(packed_ref_cache->lock);
packed_ref_cache->lock = NULL;
release_packed_ref_cache(packed_ref_cache);
clear_packed_ref_cache(&ref_cache);
}
struct ref_to_prune {
struct ref_to_prune *next;
unsigned char sha1[20];
char name[FLEX_ARRAY];
};
struct pack_refs_cb_data {
unsigned int flags;
struct ref_dir *packed_refs;
struct ref_to_prune *ref_to_prune;
};
/*
* An each_ref_entry_fn that is run over loose references only. If
* the loose reference can be packed, add an entry in the packed ref
* cache. If the reference should be pruned, also add it to
* ref_to_prune in the pack_refs_cb_data.
*/
static int pack_if_possible_fn(struct ref_entry *entry, void *cb_data)
{
struct pack_refs_cb_data *cb = cb_data;
enum peel_status peel_status;
struct ref_entry *packed_entry;
int is_tag_ref = starts_with(entry->name, "refs/tags/");
/* Do not pack per-worktree refs: */
if (ref_type(entry->name) != REF_TYPE_NORMAL)
return 0;
/* ALWAYS pack tags */
if (!(cb->flags & PACK_REFS_ALL) && !is_tag_ref)
return 0;
/* Do not pack symbolic or broken refs: */
if ((entry->flag & REF_ISSYMREF) || !entry_resolves_to_object(entry))
return 0;
/* Add a packed ref cache entry equivalent to the loose entry. */
peel_status = peel_entry(entry, 1);
if (peel_status != PEEL_PEELED && peel_status != PEEL_NON_TAG)
die("internal error peeling reference %s (%s)",
entry->name, oid_to_hex(&entry->u.value.oid));
packed_entry = find_ref(cb->packed_refs, entry->name);
if (packed_entry) {
/* Overwrite existing packed entry with info from loose entry */
packed_entry->flag = REF_ISPACKED | REF_KNOWS_PEELED;
oidcpy(&packed_entry->u.value.oid, &entry->u.value.oid);
} else {
packed_entry = create_ref_entry(entry->name, entry->u.value.oid.hash,
REF_ISPACKED | REF_KNOWS_PEELED, 0);
add_ref(cb->packed_refs, packed_entry);
}
oidcpy(&packed_entry->u.value.peeled, &entry->u.value.peeled);
/* Schedule the loose reference for pruning if requested. */
if ((cb->flags & PACK_REFS_PRUNE)) {
struct ref_to_prune *n;
FLEX_ALLOC_STR(n, name, entry->name);
hashcpy(n->sha1, entry->u.value.oid.hash);
n->next = cb->ref_to_prune;
cb->ref_to_prune = n;
}
return 0;
}
/*
* Remove empty parents, but spare refs/ and immediate subdirs.
* Note: munges *name.
*/
static void try_remove_empty_parents(char *name)
{
char *p, *q;
int i;
p = name;
for (i = 0; i < 2; i++) { /* refs/{heads,tags,...}/ */
while (*p && *p != '/')
p++;
/* tolerate duplicate slashes; see check_refname_format() */
while (*p == '/')
p++;
}
for (q = p; *q; q++)
;
while (1) {
while (q > p && *q != '/')
q--;
while (q > p && *(q-1) == '/')
q--;
if (q == p)
break;
*q = '\0';
if (rmdir(git_path("%s", name)))
break;
}
}
/* make sure nobody touched the ref, and unlink */
static void prune_ref(struct ref_to_prune *r)
{
struct ref_transaction *transaction;
struct strbuf err = STRBUF_INIT;
if (check_refname_format(r->name, 0))
return;
transaction = ref_transaction_begin(&err);
if (!transaction ||
ref_transaction_delete(transaction, r->name, r->sha1,
REF_ISPRUNING | REF_NODEREF, NULL, &err) ||
ref_transaction_commit(transaction, &err)) {
ref_transaction_free(transaction);
error("%s", err.buf);
strbuf_release(&err);
return;
}
ref_transaction_free(transaction);
strbuf_release(&err);
try_remove_empty_parents(r->name);
}
static void prune_refs(struct ref_to_prune *r)
{
while (r) {
prune_ref(r);
r = r->next;
}
}
int pack_refs(unsigned int flags)
{
struct pack_refs_cb_data cbdata;
memset(&cbdata, 0, sizeof(cbdata));
cbdata.flags = flags;
lock_packed_refs(LOCK_DIE_ON_ERROR);
cbdata.packed_refs = get_packed_refs(&ref_cache);
do_for_each_entry_in_dir(get_loose_refs(&ref_cache), 0,
pack_if_possible_fn, &cbdata);
if (commit_packed_refs())
die_errno("unable to overwrite old ref-pack file");
prune_refs(cbdata.ref_to_prune);
return 0;
}
/*
* Rewrite the packed-refs file, omitting any refs listed in
* 'refnames'. On error, leave packed-refs unchanged, write an error
* message to 'err', and return a nonzero value.
*
* The refs in 'refnames' needn't be sorted. `err` must not be NULL.
*/
static int repack_without_refs(struct string_list *refnames, struct strbuf *err)
{
struct ref_dir *packed;
struct string_list_item *refname;
int ret, needs_repacking = 0, removed = 0;
assert(err);
/* Look for a packed ref */
for_each_string_list_item(refname, refnames) {
if (get_packed_ref(refname->string)) {
needs_repacking = 1;
break;
}
}
/* Avoid locking if we have nothing to do */
if (!needs_repacking)
return 0; /* no refname exists in packed refs */
if (lock_packed_refs(0)) {
unable_to_lock_message(git_path("packed-refs"), errno, err);
return -1;
}
packed = get_packed_refs(&ref_cache);
/* Remove refnames from the cache */
for_each_string_list_item(refname, refnames)
if (remove_entry(packed, refname->string) != -1)
removed = 1;
if (!removed) {
/*
* All packed entries disappeared while we were
* acquiring the lock.
*/
rollback_packed_refs();
return 0;
}
/* Write what remains */
ret = commit_packed_refs();
if (ret)
strbuf_addf(err, "unable to overwrite old ref-pack file: %s",
strerror(errno));
return ret;
}
static int delete_ref_loose(struct ref_lock *lock, int flag, struct strbuf *err)
{
assert(err);
if (!(flag & REF_ISPACKED) || flag & REF_ISSYMREF) {
/*
* loose. The loose file name is the same as the
* lockfile name, minus ".lock":
*/
char *loose_filename = get_locked_file_path(lock->lk);
int res = unlink_or_msg(loose_filename, err);
free(loose_filename);
if (res)
return 1;
}
return 0;
}
int delete_refs(struct string_list *refnames, unsigned int flags)
{
struct strbuf err = STRBUF_INIT;
int i, result = 0;
if (!refnames->nr)
return 0;
result = repack_without_refs(refnames, &err);
if (result) {
/*
* If we failed to rewrite the packed-refs file, then
* it is unsafe to try to remove loose refs, because
* doing so might expose an obsolete packed value for
* a reference that might even point at an object that
* has been garbage collected.
*/
if (refnames->nr == 1)
error(_("could not delete reference %s: %s"),
refnames->items[0].string, err.buf);
else
error(_("could not delete references: %s"), err.buf);
goto out;
}
for (i = 0; i < refnames->nr; i++) {
const char *refname = refnames->items[i].string;
if (delete_ref(refname, NULL, flags))
result |= error(_("could not remove reference %s"), refname);
}
out:
strbuf_release(&err);
return result;
}
/*
* People using contrib's git-new-workdir have .git/logs/refs ->
* /some/other/path/.git/logs/refs, and that may live on another device.
*
* IOW, to avoid cross device rename errors, the temporary renamed log must
* live into logs/refs.
*/
#define TMP_RENAMED_LOG "logs/refs/.tmp-renamed-log"
static int rename_tmp_log(const char *newrefname)
{
int attempts_remaining = 4;
struct strbuf path = STRBUF_INIT;
int ret = -1;
retry:
strbuf_reset(&path);
strbuf_git_path(&path, "logs/%s", newrefname);
switch (safe_create_leading_directories_const(path.buf)) {
case SCLD_OK:
break; /* success */
case SCLD_VANISHED:
if (--attempts_remaining > 0)
goto retry;
/* fall through */
default:
error("unable to create directory for %s", newrefname);
goto out;
}
if (rename(git_path(TMP_RENAMED_LOG), path.buf)) {
if ((errno==EISDIR || errno==ENOTDIR) && --attempts_remaining > 0) {
/*
* rename(a, b) when b is an existing
* directory ought to result in ISDIR, but
* Solaris 5.8 gives ENOTDIR. Sheesh.
*/
if (remove_empty_directories(&path)) {
error("Directory not empty: logs/%s", newrefname);
goto out;
}
goto retry;
} else if (errno == ENOENT && --attempts_remaining > 0) {
/*
* Maybe another process just deleted one of
* the directories in the path to newrefname.
* Try again from the beginning.
*/
goto retry;
} else {
error("unable to move logfile "TMP_RENAMED_LOG" to logs/%s: %s",
newrefname, strerror(errno));
goto out;
}
}
ret = 0;
out:
strbuf_release(&path);
return ret;
}
int verify_refname_available(const char *newname,
const struct string_list *extras,
const struct string_list *skip,
struct strbuf *err)
{
struct ref_dir *packed_refs = get_packed_refs(&ref_cache);
struct ref_dir *loose_refs = get_loose_refs(&ref_cache);
if (verify_refname_available_dir(newname, extras, skip,
packed_refs, err) ||
verify_refname_available_dir(newname, extras, skip,
loose_refs, err))
return -1;
return 0;
}
static int write_ref_to_lockfile(struct ref_lock *lock,
const unsigned char *sha1, struct strbuf *err);
static int commit_ref_update(struct ref_lock *lock,
const unsigned char *sha1, const char *logmsg,
struct strbuf *err);
int rename_ref(const char *oldrefname, const char *newrefname, const char *logmsg)
{
unsigned char sha1[20], orig_sha1[20];
int flag = 0, logmoved = 0;
struct ref_lock *lock;
struct stat loginfo;
int log = !lstat(git_path("logs/%s", oldrefname), &loginfo);
struct strbuf err = STRBUF_INIT;
if (log && S_ISLNK(loginfo.st_mode))
return error("reflog for %s is a symlink", oldrefname);
if (!resolve_ref_unsafe(oldrefname, RESOLVE_REF_READING | RESOLVE_REF_NO_RECURSE,
orig_sha1, &flag))
return error("refname %s not found", oldrefname);
if (flag & REF_ISSYMREF)
return error("refname %s is a symbolic ref, renaming it is not supported",
oldrefname);
if (!rename_ref_available(oldrefname, newrefname))
return 1;
if (log && rename(git_path("logs/%s", oldrefname), git_path(TMP_RENAMED_LOG)))
return error("unable to move logfile logs/%s to "TMP_RENAMED_LOG": %s",
oldrefname, strerror(errno));
if (delete_ref(oldrefname, orig_sha1, REF_NODEREF)) {
error("unable to delete old %s", oldrefname);
goto rollback;
}
/*
* Since we are doing a shallow lookup, sha1 is not the
* correct value to pass to delete_ref as old_sha1. But that
* doesn't matter, because an old_sha1 check wouldn't add to
* the safety anyway; we want to delete the reference whatever
* its current value.
*/
if (!read_ref_full(newrefname, RESOLVE_REF_READING | RESOLVE_REF_NO_RECURSE,
sha1, NULL) &&
delete_ref(newrefname, NULL, REF_NODEREF)) {
if (errno==EISDIR) {
struct strbuf path = STRBUF_INIT;
int result;
strbuf_git_path(&path, "%s", newrefname);
result = remove_empty_directories(&path);
strbuf_release(&path);
if (result) {
error("Directory not empty: %s", newrefname);
goto rollback;
}
} else {
error("unable to delete existing %s", newrefname);
goto rollback;
}
}
if (log && rename_tmp_log(newrefname))
goto rollback;
logmoved = log;
lock = lock_ref_sha1_basic(newrefname, NULL, NULL, NULL, REF_NODEREF,
NULL, &err);
if (!lock) {
error("unable to rename '%s' to '%s': %s", oldrefname, newrefname, err.buf);
strbuf_release(&err);
goto rollback;
}
hashcpy(lock->old_oid.hash, orig_sha1);
if (write_ref_to_lockfile(lock, orig_sha1, &err) ||
commit_ref_update(lock, orig_sha1, logmsg, &err)) {
error("unable to write current sha1 into %s: %s", newrefname, err.buf);
strbuf_release(&err);
goto rollback;
}
return 0;
rollback:
lock = lock_ref_sha1_basic(oldrefname, NULL, NULL, NULL, REF_NODEREF,
NULL, &err);
if (!lock) {
error("unable to lock %s for rollback: %s", oldrefname, err.buf);
strbuf_release(&err);
goto rollbacklog;
}
flag = log_all_ref_updates;
log_all_ref_updates = 0;
if (write_ref_to_lockfile(lock, orig_sha1, &err) ||
commit_ref_update(lock, orig_sha1, NULL, &err)) {
error("unable to write current sha1 into %s: %s", oldrefname, err.buf);
strbuf_release(&err);
}
log_all_ref_updates = flag;
rollbacklog:
if (logmoved && rename(git_path("logs/%s", newrefname), git_path("logs/%s", oldrefname)))
error("unable to restore logfile %s from %s: %s",
oldrefname, newrefname, strerror(errno));
if (!logmoved && log &&
rename(git_path(TMP_RENAMED_LOG), git_path("logs/%s", oldrefname)))
error("unable to restore logfile %s from "TMP_RENAMED_LOG": %s",
oldrefname, strerror(errno));
return 1;
}
static int close_ref(struct ref_lock *lock)
{
if (close_lock_file(lock->lk))
return -1;
return 0;
}
static int commit_ref(struct ref_lock *lock)
{
char *path = get_locked_file_path(lock->lk);
struct stat st;
if (!lstat(path, &st) && S_ISDIR(st.st_mode)) {
/*
* There is a directory at the path we want to rename
* the lockfile to. Hopefully it is empty; try to
* delete it.
*/
size_t len = strlen(path);
struct strbuf sb_path = STRBUF_INIT;
strbuf_attach(&sb_path, path, len, len);
/*
* If this fails, commit_lock_file() will also fail
* and will report the problem.
*/
remove_empty_directories(&sb_path);
strbuf_release(&sb_path);
} else {
free(path);
}
if (commit_lock_file(lock->lk))
return -1;
return 0;
}
/*
* Create a reflog for a ref. If force_create = 0, the reflog will
* only be created for certain refs (those for which
* should_autocreate_reflog returns non-zero. Otherwise, create it
* regardless of the ref name. Fill in *err and return -1 on failure.
*/
static int log_ref_setup(const char *refname, struct strbuf *logfile, struct strbuf *err, int force_create)
{
int logfd, oflags = O_APPEND | O_WRONLY;
strbuf_git_path(logfile, "logs/%s", refname);
if (force_create || should_autocreate_reflog(refname)) {
if (safe_create_leading_directories(logfile->buf) < 0) {
strbuf_addf(err, "unable to create directory for '%s': "
"%s", logfile->buf, strerror(errno));
return -1;
}
oflags |= O_CREAT;
}
logfd = open(logfile->buf, oflags, 0666);
if (logfd < 0) {
if (!(oflags & O_CREAT) && (errno == ENOENT || errno == EISDIR))
return 0;
if (errno == EISDIR) {
if (remove_empty_directories(logfile)) {
strbuf_addf(err, "there are still logs under "
"'%s'", logfile->buf);
return -1;
}
logfd = open(logfile->buf, oflags, 0666);
}
if (logfd < 0) {
strbuf_addf(err, "unable to append to '%s': %s",
logfile->buf, strerror(errno));
return -1;
}
}
adjust_shared_perm(logfile->buf);
close(logfd);
return 0;
}
int safe_create_reflog(const char *refname, int force_create, struct strbuf *err)
{
int ret;
struct strbuf sb = STRBUF_INIT;
ret = log_ref_setup(refname, &sb, err, force_create);
strbuf_release(&sb);
return ret;
}
static int log_ref_write_fd(int fd, const unsigned char *old_sha1,
const unsigned char *new_sha1,
const char *committer, const char *msg)
{
int msglen, written;
unsigned maxlen, len;
char *logrec;
msglen = msg ? strlen(msg) : 0;
maxlen = strlen(committer) + msglen + 100;
logrec = xmalloc(maxlen);
len = xsnprintf(logrec, maxlen, "%s %s %s\n",
sha1_to_hex(old_sha1),
sha1_to_hex(new_sha1),
committer);
if (msglen)
len += copy_reflog_msg(logrec + len - 1, msg) - 1;
written = len <= maxlen ? write_in_full(fd, logrec, len) : -1;
free(logrec);
if (written != len)
return -1;
return 0;
}
static int log_ref_write_1(const char *refname, const unsigned char *old_sha1,
const unsigned char *new_sha1, const char *msg,
struct strbuf *logfile, int flags,
struct strbuf *err)
{
int logfd, result, oflags = O_APPEND | O_WRONLY;
if (log_all_ref_updates < 0)
log_all_ref_updates = !is_bare_repository();
result = log_ref_setup(refname, logfile, err, flags & REF_FORCE_CREATE_REFLOG);
if (result)
return result;
logfd = open(logfile->buf, oflags);
if (logfd < 0)
return 0;
result = log_ref_write_fd(logfd, old_sha1, new_sha1,
git_committer_info(0), msg);
if (result) {
strbuf_addf(err, "unable to append to '%s': %s", logfile->buf,
strerror(errno));
close(logfd);
return -1;
}
if (close(logfd)) {
strbuf_addf(err, "unable to append to '%s': %s", logfile->buf,
strerror(errno));
return -1;
}
return 0;
}
static int log_ref_write(const char *refname, const unsigned char *old_sha1,
const unsigned char *new_sha1, const char *msg,
int flags, struct strbuf *err)
{
return files_log_ref_write(refname, old_sha1, new_sha1, msg, flags,
err);
}
int files_log_ref_write(const char *refname, const unsigned char *old_sha1,
const unsigned char *new_sha1, const char *msg,
int flags, struct strbuf *err)
{
struct strbuf sb = STRBUF_INIT;
int ret = log_ref_write_1(refname, old_sha1, new_sha1, msg, &sb, flags,
err);
strbuf_release(&sb);
return ret;
}
/*
* Write sha1 into the open lockfile, then close the lockfile. On
* errors, rollback the lockfile, fill in *err and
* return -1.
*/
static int write_ref_to_lockfile(struct ref_lock *lock,
const unsigned char *sha1, struct strbuf *err)
{
static char term = '\n';
struct object *o;
int fd;
o = parse_object(sha1);
if (!o) {
strbuf_addf(err,
"trying to write ref '%s' with nonexistent object %s",
lock->ref_name, sha1_to_hex(sha1));
unlock_ref(lock);
return -1;
}
if (o->type != OBJ_COMMIT && is_branch(lock->ref_name)) {
strbuf_addf(err,
"trying to write non-commit object %s to branch '%s'",
sha1_to_hex(sha1), lock->ref_name);
unlock_ref(lock);
return -1;
}
fd = get_lock_file_fd(lock->lk);
if (write_in_full(fd, sha1_to_hex(sha1), 40) != 40 ||
write_in_full(fd, &term, 1) != 1 ||
close_ref(lock) < 0) {
strbuf_addf(err,
"couldn't write '%s'", get_lock_file_path(lock->lk));
unlock_ref(lock);
return -1;
}
return 0;
}
/*
* Commit a change to a loose reference that has already been written
* to the loose reference lockfile. Also update the reflogs if
* necessary, using the specified lockmsg (which can be NULL).
*/
static int commit_ref_update(struct ref_lock *lock,
const unsigned char *sha1, const char *logmsg,
struct strbuf *err)
{
clear_loose_ref_cache(&ref_cache);
if (log_ref_write(lock->ref_name, lock->old_oid.hash, sha1, logmsg, 0, err)) {
char *old_msg = strbuf_detach(err, NULL);
strbuf_addf(err, "cannot update the ref '%s': %s",
lock->ref_name, old_msg);
free(old_msg);
unlock_ref(lock);
return -1;
}
if (strcmp(lock->ref_name, "HEAD") != 0) {
/*
* Special hack: If a branch is updated directly and HEAD
* points to it (may happen on the remote side of a push
* for example) then logically the HEAD reflog should be
* updated too.
* A generic solution implies reverse symref information,
* but finding all symrefs pointing to the given branch
* would be rather costly for this rare event (the direct
* update of a branch) to be worth it. So let's cheat and
* check with HEAD only which should cover 99% of all usage
* scenarios (even 100% of the default ones).
*/
unsigned char head_sha1[20];
int head_flag;
const char *head_ref;
head_ref = resolve_ref_unsafe("HEAD", RESOLVE_REF_READING,
head_sha1, &head_flag);
if (head_ref && (head_flag & REF_ISSYMREF) &&
!strcmp(head_ref, lock->ref_name)) {
struct strbuf log_err = STRBUF_INIT;
if (log_ref_write("HEAD", lock->old_oid.hash, sha1,
logmsg, 0, &log_err)) {
error("%s", log_err.buf);
strbuf_release(&log_err);
}
}
}
if (commit_ref(lock)) {
strbuf_addf(err, "couldn't set '%s'", lock->ref_name);
unlock_ref(lock);
return -1;
}
unlock_ref(lock);
return 0;
}
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
static int create_ref_symlink(struct ref_lock *lock, const char *target)
{
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
int ret = -1;
#ifndef NO_SYMLINK_HEAD
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
char *ref_path = get_locked_file_path(lock->lk);
unlink(ref_path);
ret = symlink(target, ref_path);
free(ref_path);
if (ret)
fprintf(stderr, "no symlink - falling back to symbolic ref\n");
#endif
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
return ret;
}
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
static void update_symref_reflog(struct ref_lock *lock, const char *refname,
const char *target, const char *logmsg)
{
struct strbuf err = STRBUF_INIT;
unsigned char new_sha1[20];
if (logmsg && !read_ref(target, new_sha1) &&
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
log_ref_write(refname, lock->old_oid.hash, new_sha1, logmsg, 0, &err)) {
error("%s", err.buf);
strbuf_release(&err);
}
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
}
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
static int create_symref_locked(struct ref_lock *lock, const char *refname,
const char *target, const char *logmsg)
{
if (prefer_symlink_refs && !create_ref_symlink(lock, target)) {
update_symref_reflog(lock, refname, target, logmsg);
return 0;
}
if (!fdopen_lock_file(lock->lk, "w"))
return error("unable to fdopen %s: %s",
lock->lk->tempfile.filename.buf, strerror(errno));
update_symref_reflog(lock, refname, target, logmsg);
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
/* no error check; commit_ref will check ferror */
fprintf(lock->lk->tempfile.fp, "ref: %s\n", target);
if (commit_ref(lock) < 0)
return error("unable to write symref for %s: %s", refname,
strerror(errno));
return 0;
}
create_symref: use existing ref-lock code The create_symref() function predates the existence of "struct lock_file", let alone the more recent "struct ref_lock". Instead, it just does its own manual dot-locking. Besides being more code, this has a few downsides: - if git is interrupted while holding the lock, we don't clean up the lockfile - we don't do the usual directory/filename conflict check. So you can sometimes create a symref "refs/heads/foo/bar", even if "refs/heads/foo" exists (namely, if the refs are packed and we do not hit the d/f conflict in the filesystem). This patch refactors create_symref() to use the "struct ref_lock" interface, which handles both of these things. There are a few bonus cleanups that come along with it: - we leaked ref_path in some error cases - the symref contents were stored in a fixed-size buffer, putting an artificial (albeit large) limitation on the length of the refname. We now write through fprintf, and handle refnames of any size. - we called adjust_shared_perm only after the file was renamed into place, creating a potential race with readers in a shared repository. The lockfile code now handles this when creating the lockfile, making it atomic. - the legacy prefer_symlink_refs path did not do any locking at all. Admittedly, it is not atomic from a reader's perspective (as it unlinks and re-creates the symlink to overwrite), but at least it cannot conflict with other writers now. - the result of this patch is hopefully more readable. It eliminates three goto labels. Two were for error checking that is now simplified, and the third was to reach shared code that has been pulled into its own function. Signed-off-by: Jeff King <peff@peff.net> Reviewed-by: Michael Haggerty <mhagger@alum.mit.edu> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2015-12-29 06:57:01 +01:00
int create_symref(const char *refname, const char *target, const char *logmsg)
{
struct strbuf err = STRBUF_INIT;
struct ref_lock *lock;
int ret;
lock = lock_ref_sha1_basic(refname, NULL, NULL, NULL, REF_NODEREF, NULL,
&err);
if (!lock) {
error("%s", err.buf);
strbuf_release(&err);
return -1;
}
ret = create_symref_locked(lock, refname, target, logmsg);
unlock_ref(lock);
return ret;
}
int set_worktree_head_symref(const char *gitdir, const char *target)
{
static struct lock_file head_lock;
struct ref_lock *lock;
struct strbuf head_path = STRBUF_INIT;
const char *head_rel;
int ret;
strbuf_addf(&head_path, "%s/HEAD", absolute_path(gitdir));
if (hold_lock_file_for_update(&head_lock, head_path.buf,
LOCK_NO_DEREF) < 0) {
struct strbuf err = STRBUF_INIT;
unable_to_lock_message(head_path.buf, errno, &err);
error("%s", err.buf);
strbuf_release(&err);
strbuf_release(&head_path);
return -1;
}
/* head_rel will be "HEAD" for the main tree, "worktrees/wt/HEAD" for
linked trees */
head_rel = remove_leading_path(head_path.buf,
absolute_path(get_git_common_dir()));
/* to make use of create_symref_locked(), initialize ref_lock */
lock = xcalloc(1, sizeof(struct ref_lock));
lock->lk = &head_lock;
lock->ref_name = xstrdup(head_rel);
ret = create_symref_locked(lock, head_rel, target, NULL);
unlock_ref(lock); /* will free lock */
strbuf_release(&head_path);
return ret;
}
int reflog_exists(const char *refname)
{
struct stat st;
return !lstat(git_path("logs/%s", refname), &st) &&
S_ISREG(st.st_mode);
}
int delete_reflog(const char *refname)
{
return remove_path(git_path("logs/%s", refname));
}
static int show_one_reflog_ent(struct strbuf *sb, each_reflog_ent_fn fn, void *cb_data)
{
unsigned char osha1[20], nsha1[20];
char *email_end, *message;
unsigned long timestamp;
int tz;
/* old SP new SP name <email> SP time TAB msg LF */
if (sb->len < 83 || sb->buf[sb->len - 1] != '\n' ||
get_sha1_hex(sb->buf, osha1) || sb->buf[40] != ' ' ||
get_sha1_hex(sb->buf + 41, nsha1) || sb->buf[81] != ' ' ||
!(email_end = strchr(sb->buf + 82, '>')) ||
email_end[1] != ' ' ||
!(timestamp = strtoul(email_end + 2, &message, 10)) ||
!message || message[0] != ' ' ||
(message[1] != '+' && message[1] != '-') ||
!isdigit(message[2]) || !isdigit(message[3]) ||
!isdigit(message[4]) || !isdigit(message[5]))
return 0; /* corrupt? */
email_end[1] = '\0';
tz = strtol(message + 1, NULL, 10);
if (message[6] != '\t')
message += 6;
else
message += 7;
return fn(osha1, nsha1, sb->buf + 82, timestamp, tz, message, cb_data);
}
static char *find_beginning_of_line(char *bob, char *scan)
{
while (bob < scan && *(--scan) != '\n')
; /* keep scanning backwards */
/*
* Return either beginning of the buffer, or LF at the end of
* the previous line.
*/
return scan;
}
int for_each_reflog_ent_reverse(const char *refname, each_reflog_ent_fn fn, void *cb_data)
{
struct strbuf sb = STRBUF_INIT;
FILE *logfp;
long pos;
int ret = 0, at_tail = 1;
logfp = fopen(git_path("logs/%s", refname), "r");
if (!logfp)
return -1;
/* Jump to the end */
if (fseek(logfp, 0, SEEK_END) < 0)
return error("cannot seek back reflog for %s: %s",
refname, strerror(errno));
pos = ftell(logfp);
while (!ret && 0 < pos) {
int cnt;
size_t nread;
char buf[BUFSIZ];
char *endp, *scanp;
/* Fill next block from the end */
cnt = (sizeof(buf) < pos) ? sizeof(buf) : pos;
if (fseek(logfp, pos - cnt, SEEK_SET))
return error("cannot seek back reflog for %s: %s",
refname, strerror(errno));
nread = fread(buf, cnt, 1, logfp);
if (nread != 1)
return error("cannot read %d bytes from reflog for %s: %s",
cnt, refname, strerror(errno));
pos -= cnt;
scanp = endp = buf + cnt;
if (at_tail && scanp[-1] == '\n')
/* Looking at the final LF at the end of the file */
scanp--;
at_tail = 0;
while (buf < scanp) {
/*
* terminating LF of the previous line, or the beginning
* of the buffer.
*/
char *bp;
bp = find_beginning_of_line(buf, scanp);
if (*bp == '\n') {
/*
* The newline is the end of the previous line,
* so we know we have complete line starting
* at (bp + 1). Prefix it onto any prior data
* we collected for the line and process it.
*/
strbuf_splice(&sb, 0, 0, bp + 1, endp - (bp + 1));
scanp = bp;
endp = bp + 1;
ret = show_one_reflog_ent(&sb, fn, cb_data);
strbuf_reset(&sb);
if (ret)
break;
} else if (!pos) {
/*
* We are at the start of the buffer, and the
* start of the file; there is no previous
* line, and we have everything for this one.
* Process it, and we can end the loop.
*/
strbuf_splice(&sb, 0, 0, buf, endp - buf);
ret = show_one_reflog_ent(&sb, fn, cb_data);
strbuf_reset(&sb);
break;
}
if (bp == buf) {
/*
* We are at the start of the buffer, and there
* is more file to read backwards. Which means
* we are in the middle of a line. Note that we
* may get here even if *bp was a newline; that
* just means we are at the exact end of the
* previous line, rather than some spot in the
* middle.
*
* Save away what we have to be combined with
* the data from the next read.
*/
strbuf_splice(&sb, 0, 0, buf, endp - buf);
break;
}
}
}
if (!ret && sb.len)
die("BUG: reverse reflog parser had leftover data");
fclose(logfp);
strbuf_release(&sb);
return ret;
}
int for_each_reflog_ent(const char *refname, each_reflog_ent_fn fn, void *cb_data)
{
FILE *logfp;
struct strbuf sb = STRBUF_INIT;
int ret = 0;
logfp = fopen(git_path("logs/%s", refname), "r");
if (!logfp)
return -1;
while (!ret && !strbuf_getwholeline(&sb, logfp, '\n'))
ret = show_one_reflog_ent(&sb, fn, cb_data);
fclose(logfp);
strbuf_release(&sb);
return ret;
}
struct files_reflog_iterator {
struct ref_iterator base;
struct dir_iterator *dir_iterator;
struct object_id oid;
};
static int files_reflog_iterator_advance(struct ref_iterator *ref_iterator)
{
struct files_reflog_iterator *iter =
(struct files_reflog_iterator *)ref_iterator;
struct dir_iterator *diter = iter->dir_iterator;
int ok;
while ((ok = dir_iterator_advance(diter)) == ITER_OK) {
int flags;
if (!S_ISREG(diter->st.st_mode))
continue;
if (diter->basename[0] == '.')
continue;
if (ends_with(diter->basename, ".lock"))
continue;
if (read_ref_full(diter->relative_path, 0,
iter->oid.hash, &flags)) {
error("bad ref for %s", diter->path.buf);
continue;
}
iter->base.refname = diter->relative_path;
iter->base.oid = &iter->oid;
iter->base.flags = flags;
return ITER_OK;
}
iter->dir_iterator = NULL;
if (ref_iterator_abort(ref_iterator) == ITER_ERROR)
ok = ITER_ERROR;
return ok;
}
static int files_reflog_iterator_peel(struct ref_iterator *ref_iterator,
struct object_id *peeled)
{
die("BUG: ref_iterator_peel() called for reflog_iterator");
}
static int files_reflog_iterator_abort(struct ref_iterator *ref_iterator)
{
struct files_reflog_iterator *iter =
(struct files_reflog_iterator *)ref_iterator;
int ok = ITER_DONE;
if (iter->dir_iterator)
ok = dir_iterator_abort(iter->dir_iterator);
base_ref_iterator_free(ref_iterator);
return ok;
}
static struct ref_iterator_vtable files_reflog_iterator_vtable = {
files_reflog_iterator_advance,
files_reflog_iterator_peel,
files_reflog_iterator_abort
};
struct ref_iterator *files_reflog_iterator_begin(void)
{
struct files_reflog_iterator *iter = xcalloc(1, sizeof(*iter));
struct ref_iterator *ref_iterator = &iter->base;
base_ref_iterator_init(ref_iterator, &files_reflog_iterator_vtable);
iter->dir_iterator = dir_iterator_begin(git_path("logs"));
return ref_iterator;
}
int for_each_reflog(each_ref_fn fn, void *cb_data)
{
return do_for_each_ref_iterator(files_reflog_iterator_begin(),
fn, cb_data);
}
static int ref_update_reject_duplicates(struct string_list *refnames,
struct strbuf *err)
{
int i, n = refnames->nr;
assert(err);
for (i = 1; i < n; i++)
if (!strcmp(refnames->items[i - 1].string, refnames->items[i].string)) {
strbuf_addf(err,
"multiple updates for ref '%s' not allowed.",
refnames->items[i].string);
return 1;
}
return 0;
}
/*
* If update is a direct update of head_ref (the reference pointed to
* by HEAD), then add an extra REF_LOG_ONLY update for HEAD.
*/
static int split_head_update(struct ref_update *update,
struct ref_transaction *transaction,
const char *head_ref,
struct string_list *affected_refnames,
struct strbuf *err)
{
struct string_list_item *item;
struct ref_update *new_update;
if ((update->flags & REF_LOG_ONLY) ||
(update->flags & REF_ISPRUNING) ||
(update->flags & REF_UPDATE_VIA_HEAD))
return 0;
if (strcmp(update->refname, head_ref))
return 0;
/*
* First make sure that HEAD is not already in the
* transaction. This insertion is O(N) in the transaction
* size, but it happens at most once per transaction.
*/
item = string_list_insert(affected_refnames, "HEAD");
if (item->util) {
/* An entry already existed */
strbuf_addf(err,
"multiple updates for 'HEAD' (including one "
"via its referent '%s') are not allowed",
update->refname);
return TRANSACTION_NAME_CONFLICT;
}
new_update = ref_transaction_add_update(
transaction, "HEAD",
update->flags | REF_LOG_ONLY | REF_NODEREF,
update->new_sha1, update->old_sha1,
update->msg);
item->util = new_update;
return 0;
}
/*
* update is for a symref that points at referent and doesn't have
* REF_NODEREF set. Split it into two updates:
* - The original update, but with REF_LOG_ONLY and REF_NODEREF set
* - A new, separate update for the referent reference
* Note that the new update will itself be subject to splitting when
* the iteration gets to it.
*/
static int split_symref_update(struct ref_update *update,
const char *referent,
struct ref_transaction *transaction,
struct string_list *affected_refnames,
struct strbuf *err)
{
struct string_list_item *item;
struct ref_update *new_update;
unsigned int new_flags;
/*
* First make sure that referent is not already in the
* transaction. This insertion is O(N) in the transaction
* size, but it happens at most once per symref in a
* transaction.
*/
item = string_list_insert(affected_refnames, referent);
if (item->util) {
/* An entry already existed */
strbuf_addf(err,
"multiple updates for '%s' (including one "
"via symref '%s') are not allowed",
referent, update->refname);
return TRANSACTION_NAME_CONFLICT;
}
new_flags = update->flags;
if (!strcmp(update->refname, "HEAD")) {
/*
* Record that the new update came via HEAD, so that
* when we process it, split_head_update() doesn't try
* to add another reflog update for HEAD. Note that
* this bit will be propagated if the new_update
* itself needs to be split.
*/
new_flags |= REF_UPDATE_VIA_HEAD;
}
new_update = ref_transaction_add_update(
transaction, referent, new_flags,
update->new_sha1, update->old_sha1,
update->msg);
new_update->parent_update = update;
/*
* Change the symbolic ref update to log only. Also, it
* doesn't need to check its old SHA-1 value, as that will be
* done when new_update is processed.
*/
update->flags |= REF_LOG_ONLY | REF_NODEREF;
update->flags &= ~REF_HAVE_OLD;
item->util = new_update;
return 0;
}
/*
* Return the refname under which update was originally requested.
*/
static const char *original_update_refname(struct ref_update *update)
{
while (update->parent_update)
update = update->parent_update;
return update->refname;
}
/*
* Check whether the REF_HAVE_OLD and old_oid values stored in update
* are consistent with oid, which is the reference's current value. If
* everything is OK, return 0; otherwise, write an error message to
* err and return -1.
*/
static int check_old_oid(struct ref_update *update, struct object_id *oid,
struct strbuf *err)
{
if (!(update->flags & REF_HAVE_OLD) ||
!hashcmp(oid->hash, update->old_sha1))
return 0;
if (is_null_sha1(update->old_sha1))
strbuf_addf(err, "cannot lock ref '%s': "
"reference already exists",
original_update_refname(update));
else if (is_null_oid(oid))
strbuf_addf(err, "cannot lock ref '%s': "
"reference is missing but expected %s",
original_update_refname(update),
sha1_to_hex(update->old_sha1));
else
strbuf_addf(err, "cannot lock ref '%s': "
"is at %s but expected %s",
original_update_refname(update),
oid_to_hex(oid),
sha1_to_hex(update->old_sha1));
return -1;
}
/*
* Prepare for carrying out update:
* - Lock the reference referred to by update.
* - Read the reference under lock.
* - Check that its old SHA-1 value (if specified) is correct, and in
* any case record it in update->lock->old_oid for later use when
* writing the reflog.
* - If it is a symref update without REF_NODEREF, split it up into a
* REF_LOG_ONLY update of the symref and add a separate update for
* the referent to transaction.
* - If it is an update of head_ref, add a corresponding REF_LOG_ONLY
* update of HEAD.
*/
static int lock_ref_for_update(struct ref_update *update,
struct ref_transaction *transaction,
const char *head_ref,
struct string_list *affected_refnames,
struct strbuf *err)
{
struct strbuf referent = STRBUF_INIT;
int mustexist = (update->flags & REF_HAVE_OLD) &&
!is_null_sha1(update->old_sha1);
int ret;
struct ref_lock *lock;
if ((update->flags & REF_HAVE_NEW) && is_null_sha1(update->new_sha1))
update->flags |= REF_DELETING;
if (head_ref) {
ret = split_head_update(update, transaction, head_ref,
affected_refnames, err);
if (ret)
return ret;
}
ret = lock_raw_ref(update->refname, mustexist,
affected_refnames, NULL,
&update->lock, &referent,
&update->type, err);
if (ret) {
char *reason;
reason = strbuf_detach(err, NULL);
strbuf_addf(err, "cannot lock ref '%s': %s",
original_update_refname(update), reason);
free(reason);
return ret;
}
lock = update->lock;
if (update->type & REF_ISSYMREF) {
if (update->flags & REF_NODEREF) {
/*
* We won't be reading the referent as part of
* the transaction, so we have to read it here
* to record and possibly check old_sha1:
*/
if (read_ref_full(referent.buf, 0,
lock->old_oid.hash, NULL)) {
if (update->flags & REF_HAVE_OLD) {
strbuf_addf(err, "cannot lock ref '%s': "
"error reading reference",
original_update_refname(update));
return -1;
}
} else if (check_old_oid(update, &lock->old_oid, err)) {
return TRANSACTION_GENERIC_ERROR;
}
} else {
/*
* Create a new update for the reference this
* symref is pointing at. Also, we will record
* and verify old_sha1 for this update as part
* of processing the split-off update, so we
* don't have to do it here.
*/
ret = split_symref_update(update, referent.buf, transaction,
affected_refnames, err);
if (ret)
return ret;
}
} else {
struct ref_update *parent_update;
if (check_old_oid(update, &lock->old_oid, err))
return TRANSACTION_GENERIC_ERROR;
/*
* If this update is happening indirectly because of a
* symref update, record the old SHA-1 in the parent
* update:
*/
for (parent_update = update->parent_update;
parent_update;
parent_update = parent_update->parent_update) {
oidcpy(&parent_update->lock->old_oid, &lock->old_oid);
}
}
if ((update->flags & REF_HAVE_NEW) &&
!(update->flags & REF_DELETING) &&
!(update->flags & REF_LOG_ONLY)) {
if (!(update->type & REF_ISSYMREF) &&
!hashcmp(lock->old_oid.hash, update->new_sha1)) {
/*
* The reference already has the desired
* value, so we don't need to write it.
*/
} else if (write_ref_to_lockfile(lock, update->new_sha1,
err)) {
char *write_err = strbuf_detach(err, NULL);
/*
* The lock was freed upon failure of
* write_ref_to_lockfile():
*/
update->lock = NULL;
strbuf_addf(err,
"cannot update ref '%s': %s",
update->refname, write_err);
free(write_err);
return TRANSACTION_GENERIC_ERROR;
} else {
update->flags |= REF_NEEDS_COMMIT;
}
}
if (!(update->flags & REF_NEEDS_COMMIT)) {
/*
* We didn't call write_ref_to_lockfile(), so
* the lockfile is still open. Close it to
* free up the file descriptor:
*/
if (close_ref(lock)) {
strbuf_addf(err, "couldn't close '%s.lock'",
update->refname);
return TRANSACTION_GENERIC_ERROR;
}
}
return 0;
}
int ref_transaction_commit(struct ref_transaction *transaction,
struct strbuf *err)
{
int ret = 0, i;
struct string_list refs_to_delete = STRING_LIST_INIT_NODUP;
struct string_list_item *ref_to_delete;
struct string_list affected_refnames = STRING_LIST_INIT_NODUP;
char *head_ref = NULL;
int head_type;
struct object_id head_oid;
assert(err);
if (transaction->state != REF_TRANSACTION_OPEN)
die("BUG: commit called for transaction that is not open");
if (!transaction->nr) {
transaction->state = REF_TRANSACTION_CLOSED;
return 0;
}
/*
* Fail if a refname appears more than once in the
* transaction. (If we end up splitting up any updates using
* split_symref_update() or split_head_update(), those
* functions will check that the new updates don't have the
* same refname as any existing ones.)
*/
for (i = 0; i < transaction->nr; i++) {
struct ref_update *update = transaction->updates[i];
struct string_list_item *item =
string_list_append(&affected_refnames, update->refname);
/*
* We store a pointer to update in item->util, but at
* the moment we never use the value of this field
* except to check whether it is non-NULL.
*/
item->util = update;
}
string_list_sort(&affected_refnames);
if (ref_update_reject_duplicates(&affected_refnames, err)) {
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
/*
* Special hack: If a branch is updated directly and HEAD
* points to it (may happen on the remote side of a push
* for example) then logically the HEAD reflog should be
* updated too.
*
* A generic solution would require reverse symref lookups,
* but finding all symrefs pointing to a given branch would be
* rather costly for this rare event (the direct update of a
* branch) to be worth it. So let's cheat and check with HEAD
* only, which should cover 99% of all usage scenarios (even
* 100% of the default ones).
*
* So if HEAD is a symbolic reference, then record the name of
* the reference that it points to. If we see an update of
* head_ref within the transaction, then split_head_update()
* arranges for the reflog of HEAD to be updated, too.
*/
head_ref = resolve_refdup("HEAD", RESOLVE_REF_NO_RECURSE,
head_oid.hash, &head_type);
if (head_ref && !(head_type & REF_ISSYMREF)) {
free(head_ref);
head_ref = NULL;
}
/*
* Acquire all locks, verify old values if provided, check
* that new values are valid, and write new values to the
* lockfiles, ready to be activated. Only keep one lockfile
* open at a time to avoid running out of file descriptors.
*/
for (i = 0; i < transaction->nr; i++) {
struct ref_update *update = transaction->updates[i];
ret = lock_ref_for_update(update, transaction, head_ref,
&affected_refnames, err);
if (ret)
goto cleanup;
}
/* Perform updates first so live commits remain referenced */
for (i = 0; i < transaction->nr; i++) {
struct ref_update *update = transaction->updates[i];
struct ref_lock *lock = update->lock;
if (update->flags & REF_NEEDS_COMMIT ||
update->flags & REF_LOG_ONLY) {
if (log_ref_write(lock->ref_name, lock->old_oid.hash,
update->new_sha1,
update->msg, update->flags, err)) {
char *old_msg = strbuf_detach(err, NULL);
strbuf_addf(err, "cannot update the ref '%s': %s",
lock->ref_name, old_msg);
free(old_msg);
unlock_ref(lock);
update->lock = NULL;
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
}
if (update->flags & REF_NEEDS_COMMIT) {
clear_loose_ref_cache(&ref_cache);
if (commit_ref(lock)) {
strbuf_addf(err, "couldn't set '%s'", lock->ref_name);
unlock_ref(lock);
update->lock = NULL;
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
}
}
/* Perform deletes now that updates are safely completed */
for (i = 0; i < transaction->nr; i++) {
struct ref_update *update = transaction->updates[i];
if (update->flags & REF_DELETING &&
!(update->flags & REF_LOG_ONLY)) {
if (delete_ref_loose(update->lock, update->type, err)) {
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
if (!(update->flags & REF_ISPRUNING))
string_list_append(&refs_to_delete,
update->lock->ref_name);
}
}
if (repack_without_refs(&refs_to_delete, err)) {
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
for_each_string_list_item(ref_to_delete, &refs_to_delete)
unlink_or_warn(git_path("logs/%s", ref_to_delete->string));
clear_loose_ref_cache(&ref_cache);
cleanup:
transaction->state = REF_TRANSACTION_CLOSED;
for (i = 0; i < transaction->nr; i++)
if (transaction->updates[i]->lock)
unlock_ref(transaction->updates[i]->lock);
string_list_clear(&refs_to_delete, 0);
free(head_ref);
string_list_clear(&affected_refnames, 0);
return ret;
}
static int ref_present(const char *refname,
const struct object_id *oid, int flags, void *cb_data)
{
struct string_list *affected_refnames = cb_data;
return string_list_has_string(affected_refnames, refname);
}
int initial_ref_transaction_commit(struct ref_transaction *transaction,
struct strbuf *err)
{
int ret = 0, i;
struct string_list affected_refnames = STRING_LIST_INIT_NODUP;
assert(err);
if (transaction->state != REF_TRANSACTION_OPEN)
die("BUG: commit called for transaction that is not open");
/* Fail if a refname appears more than once in the transaction: */
for (i = 0; i < transaction->nr; i++)
string_list_append(&affected_refnames,
transaction->updates[i]->refname);
string_list_sort(&affected_refnames);
if (ref_update_reject_duplicates(&affected_refnames, err)) {
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
/*
* It's really undefined to call this function in an active
* repository or when there are existing references: we are
* only locking and changing packed-refs, so (1) any
* simultaneous processes might try to change a reference at
* the same time we do, and (2) any existing loose versions of
* the references that we are setting would have precedence
* over our values. But some remote helpers create the remote
* "HEAD" and "master" branches before calling this function,
* so here we really only check that none of the references
* that we are creating already exists.
*/
if (for_each_rawref(ref_present, &affected_refnames))
die("BUG: initial ref transaction called with existing refs");
for (i = 0; i < transaction->nr; i++) {
struct ref_update *update = transaction->updates[i];
if ((update->flags & REF_HAVE_OLD) &&
!is_null_sha1(update->old_sha1))
die("BUG: initial ref transaction with old_sha1 set");
if (verify_refname_available(update->refname,
&affected_refnames, NULL,
err)) {
ret = TRANSACTION_NAME_CONFLICT;
goto cleanup;
}
}
if (lock_packed_refs(0)) {
strbuf_addf(err, "unable to lock packed-refs file: %s",
strerror(errno));
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
for (i = 0; i < transaction->nr; i++) {
struct ref_update *update = transaction->updates[i];
if ((update->flags & REF_HAVE_NEW) &&
!is_null_sha1(update->new_sha1))
add_packed_ref(update->refname, update->new_sha1);
}
if (commit_packed_refs()) {
strbuf_addf(err, "unable to commit packed-refs file: %s",
strerror(errno));
ret = TRANSACTION_GENERIC_ERROR;
goto cleanup;
}
cleanup:
transaction->state = REF_TRANSACTION_CLOSED;
string_list_clear(&affected_refnames, 0);
return ret;
}
struct expire_reflog_cb {
unsigned int flags;
reflog_expiry_should_prune_fn *should_prune_fn;
void *policy_cb;
FILE *newlog;
unsigned char last_kept_sha1[20];
};
static int expire_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
const char *email, unsigned long timestamp, int tz,
const char *message, void *cb_data)
{
struct expire_reflog_cb *cb = cb_data;
struct expire_reflog_policy_cb *policy_cb = cb->policy_cb;
if (cb->flags & EXPIRE_REFLOGS_REWRITE)
osha1 = cb->last_kept_sha1;
if ((*cb->should_prune_fn)(osha1, nsha1, email, timestamp, tz,
message, policy_cb)) {
if (!cb->newlog)
printf("would prune %s", message);
else if (cb->flags & EXPIRE_REFLOGS_VERBOSE)
printf("prune %s", message);
} else {
if (cb->newlog) {
fprintf(cb->newlog, "%s %s %s %lu %+05d\t%s",
sha1_to_hex(osha1), sha1_to_hex(nsha1),
email, timestamp, tz, message);
hashcpy(cb->last_kept_sha1, nsha1);
}
if (cb->flags & EXPIRE_REFLOGS_VERBOSE)
printf("keep %s", message);
}
return 0;
}
int reflog_expire(const char *refname, const unsigned char *sha1,
unsigned int flags,
reflog_expiry_prepare_fn prepare_fn,
reflog_expiry_should_prune_fn should_prune_fn,
reflog_expiry_cleanup_fn cleanup_fn,
void *policy_cb_data)
{
static struct lock_file reflog_lock;
struct expire_reflog_cb cb;
struct ref_lock *lock;
char *log_file;
int status = 0;
int type;
struct strbuf err = STRBUF_INIT;
memset(&cb, 0, sizeof(cb));
cb.flags = flags;
cb.policy_cb = policy_cb_data;
cb.should_prune_fn = should_prune_fn;
/*
* The reflog file is locked by holding the lock on the
* reference itself, plus we might need to update the
* reference if --updateref was specified:
*/
lock = lock_ref_sha1_basic(refname, sha1, NULL, NULL, REF_NODEREF,
&type, &err);
if (!lock) {
error("cannot lock ref '%s': %s", refname, err.buf);
strbuf_release(&err);
return -1;
}
if (!reflog_exists(refname)) {
unlock_ref(lock);
return 0;
}
log_file = git_pathdup("logs/%s", refname);
if (!(flags & EXPIRE_REFLOGS_DRY_RUN)) {
/*
* Even though holding $GIT_DIR/logs/$reflog.lock has
* no locking implications, we use the lock_file
* machinery here anyway because it does a lot of the
* work we need, including cleaning up if the program
* exits unexpectedly.
*/
if (hold_lock_file_for_update(&reflog_lock, log_file, 0) < 0) {
struct strbuf err = STRBUF_INIT;
unable_to_lock_message(log_file, errno, &err);
error("%s", err.buf);
strbuf_release(&err);
goto failure;
}
cb.newlog = fdopen_lock_file(&reflog_lock, "w");
if (!cb.newlog) {
error("cannot fdopen %s (%s)",
get_lock_file_path(&reflog_lock), strerror(errno));
goto failure;
}
}
(*prepare_fn)(refname, sha1, cb.policy_cb);
for_each_reflog_ent(refname, expire_reflog_ent, &cb);
(*cleanup_fn)(cb.policy_cb);
if (!(flags & EXPIRE_REFLOGS_DRY_RUN)) {
/*
* It doesn't make sense to adjust a reference pointed
* to by a symbolic ref based on expiring entries in
* the symbolic reference's reflog. Nor can we update
* a reference if there are no remaining reflog
* entries.
*/
int update = (flags & EXPIRE_REFLOGS_UPDATE_REF) &&
!(type & REF_ISSYMREF) &&
!is_null_sha1(cb.last_kept_sha1);
if (close_lock_file(&reflog_lock)) {
status |= error("couldn't write %s: %s", log_file,
strerror(errno));
} else if (update &&
(write_in_full(get_lock_file_fd(lock->lk),
sha1_to_hex(cb.last_kept_sha1), 40) != 40 ||
write_str_in_full(get_lock_file_fd(lock->lk), "\n") != 1 ||
close_ref(lock) < 0)) {
status |= error("couldn't write %s",
get_lock_file_path(lock->lk));
rollback_lock_file(&reflog_lock);
} else if (commit_lock_file(&reflog_lock)) {
status |= error("unable to write reflog '%s' (%s)",
log_file, strerror(errno));
} else if (update && commit_ref(lock)) {
status |= error("couldn't set %s", lock->ref_name);
}
}
free(log_file);
unlock_ref(lock);
return status;
failure:
rollback_lock_file(&reflog_lock);
free(log_file);
unlock_ref(lock);
return -1;
}