manifest: tigher manifest parsing and flag use

In the manifest line, flags are put directly after the hash, so the
parser has been guessing the presence of flags based on the length of
the hash. Replace this assumption by an enumeration of the valid flags
and removing them from the hash first as they are distinct input values.

Consistently handle the expected 256bit length of the SHA1-replacement
in the pure Python parser. Check that setting flags will use one of the
blessed values.

Extend write logic in the C version to handle 256bit hashes as well.

Verify that hashes always have exactly the expected length. Since
1070df141718 we should no longer depend on the old extra-byte hack.

Differential Revision: https://phab.mercurial-scm.org/D8679

mercurial/cext/manifest.c

@@ -1,1006 +1,1014 b''
 /*
  * manifest.c - manifest type that does on-demand parsing.
  *
  * Copyright 2015, Google Inc.
  *
  * This software may be used and distributed according to the terms of
  * the GNU General Public License, incorporated herein by reference.
  */
 #include <Python.h>
 
 #include <assert.h>
 #include <stdlib.h>
 #include <string.h>
 
 #include "charencode.h"
 #include "util.h"
 
 #define DEFAULT_LINES 100000
 
 typedef struct {
 	char *start;
 	Py_ssize_t len; /* length of line including terminal newline */
 	char hash_suffix;
 	bool from_malloc;
 	bool deleted;
 } line;
 
 typedef struct {
 	PyObject_HEAD
 	PyObject *pydata;
 	line *lines;
 	int numlines; /* number of line entries */
 	int livelines; /* number of non-deleted lines */
 	int maxlines; /* allocated number of lines */
 	bool dirty;
 } lazymanifest;
 
 #define MANIFEST_OOM -1
 #define MANIFEST_NOT_SORTED -2
 #define MANIFEST_MALFORMED -3
 #define MANIFEST_BOGUS_FILENAME -4
 #define MANIFEST_TOO_SHORT_LINE -5
 
 /* get the length of the path for a line */
 static Py_ssize_t pathlen(line *l)
 {
 	const char *end = memchr(l->start, '\0', l->len);
 	return (end) ? (Py_ssize_t)(end - l->start) : l->len;
 }
 
 /* get the node value of a single line */
-static PyObject *nodeof(line *l)
+static PyObject *nodeof(line *l, char *flag)
 {
 	char *s = l->start;
 	Py_ssize_t llen = pathlen(l);
 	Py_ssize_t hlen = l->len - llen - 2;
-	Py_ssize_t hlen_raw = 20;
+	Py_ssize_t hlen_raw;
 	PyObject *hash;
 	if (llen + 1 + 40 + 1 > l->len) { /* path '\0' hash '\n' */
 		PyErr_SetString(PyExc_ValueError, "manifest line too short");
 		return NULL;
 	}
+	/* Detect flags after the hash first. */
+	switch (s[llen + hlen]) {
+	case 'l':
+	case 't':
+	case 'x':
+		*flag = s[llen + hlen];
+		--hlen;
+		break;
+	default:
+		*flag = '\0';
+		break;
+	}
+
 	switch (hlen) {
 	case 40: /* sha1 */
-	case 41: /* sha1 with cruft for a merge */
+		hlen_raw = 20;
 		break;
 	case 64: /* new hash */
-	case 65: /* new hash with cruft for a merge */
 		hlen_raw = 32;
 		break;
 	default:
 		PyErr_SetString(PyExc_ValueError, "invalid node length in manifest");
 		return NULL;
 	}
 	hash = unhexlify(s + llen + 1, hlen_raw * 2);
 	if (!hash) {
 		return NULL;
 	}
 	if (l->hash_suffix != '\0') {
 		char newhash[33];
 		memcpy(newhash, PyBytes_AsString(hash), hlen_raw);
 		Py_DECREF(hash);
 		newhash[hlen_raw] = l->hash_suffix;
 		hash = PyBytes_FromStringAndSize(newhash, hlen_raw+1);
 	}
 	return hash;
 }
 
 /* get the node hash and flags of a line as a tuple */
 static PyObject *hashflags(line *l)
 {
-	char *s = l->start;
-	Py_ssize_t plen = pathlen(l);
-	PyObject *hash = nodeof(l);
+	char flag;
+	PyObject *hash = nodeof(l, &flag);
 	ssize_t hlen;
 	Py_ssize_t hplen, flen;
 	PyObject *flags;
 	PyObject *tup;
 
 	if (!hash)
 		return NULL;
-	/* hash is either 20 or 21 bytes for an old hash, so we use a
-	   ternary here to get the "real" hexlified sha length. */
-	hlen = PyBytes_GET_SIZE(hash) < 22 ? 40 : 64;
-	/* 1 for null byte, 1 for newline */
-	hplen = plen + hlen + 2;
-	flen = l->len - hplen;
-
-	flags = PyBytes_FromStringAndSize(s + hplen - 1, flen);
+	flags = PyBytes_FromStringAndSize(&flag, flag ? 1 : 0);
 	if (!flags) {
 		Py_DECREF(hash);
 		return NULL;
 	}
 	tup = PyTuple_Pack(2, hash, flags);
 	Py_DECREF(flags);
 	Py_DECREF(hash);
 	return tup;
 }
 
 /* if we're about to run out of space in the line index, add more */
 static bool realloc_if_full(lazymanifest *self)
 {
 	if (self->numlines == self->maxlines) {
 		self->maxlines *= 2;
 		self->lines = realloc(self->lines, self->maxlines * sizeof(line));
 	}
 	return !!self->lines;
 }
 
 /*
  * Find the line boundaries in the manifest that 'data' points to and store
  * information about each line in 'self'.
  */
 static int find_lines(lazymanifest *self, char *data, Py_ssize_t len)
 {
 	char *prev = NULL;
 	while (len > 0) {
 		line *l;
 		char *next;
 		if (*data == '\0') {
 			/* It's implausible there's no filename, don't
 			 * even bother looking for the newline. */
 			return MANIFEST_BOGUS_FILENAME;
 		}
 		next = memchr(data, '\n', len);
 		if (!next) {
 			return MANIFEST_MALFORMED;
 		}
 		if ((next - data) < 42) {
 			/* We should have at least 42 bytes in a line:
 			   1 byte filename
 			   1 NUL
 			   40 bytes of hash
 			   so we can give up here.
 			*/
 			return MANIFEST_TOO_SHORT_LINE;
 		}
 		next++; /* advance past newline */
 		if (prev && strcmp(prev, data) > -1) {
 			/* This data isn't sorted, so we have to abort. */
 			return MANIFEST_NOT_SORTED;
 		}
 		if (!realloc_if_full(self)) {
 			return MANIFEST_OOM; /* no memory */
 		}
 		l = self->lines + ((self->numlines)++);
 		l->start = data;
 		l->len = next - data;
 		l->hash_suffix = '\0';
 		l->from_malloc = false;
 		l->deleted = false;
 		len = len - l->len;
 		prev = data;
 		data = next;
 	}
 	self->livelines = self->numlines;
 	return 0;
 }
 
 static void lazymanifest_init_early(lazymanifest *self)
 {
 	self->pydata = NULL;
 	self->lines = NULL;
 	self->numlines = 0;
 	self->maxlines = 0;
 }
 
 static int lazymanifest_init(lazymanifest *self, PyObject *args)
 {
 	char *data;
 	Py_ssize_t len;
 	int err, ret;
 	PyObject *pydata;
 
 	lazymanifest_init_early(self);
 	if (!PyArg_ParseTuple(args, "S", &pydata)) {
 		return -1;
 	}
 	err = PyBytes_AsStringAndSize(pydata, &data, &len);
 
 	self->dirty = false;
 	if (err == -1)
 		return -1;
 	self->pydata = pydata;
 	Py_INCREF(self->pydata);
 	Py_BEGIN_ALLOW_THREADS
 	self->lines = malloc(DEFAULT_LINES * sizeof(line));
 	self->maxlines = DEFAULT_LINES;
 	self->numlines = 0;
 	if (!self->lines)
 		ret = MANIFEST_OOM;
 	else
 		ret = find_lines(self, data, len);
 	Py_END_ALLOW_THREADS
 	switch (ret) {
 	case 0:
 		break;
 	case MANIFEST_OOM:
 		PyErr_NoMemory();
 		break;
 	case MANIFEST_NOT_SORTED:
 		PyErr_Format(PyExc_ValueError,
 			     "Manifest lines not in sorted order.");
 		break;
 	case MANIFEST_MALFORMED:
 		PyErr_Format(PyExc_ValueError,
 			     "Manifest did not end in a newline.");
 		break;
 	case MANIFEST_BOGUS_FILENAME:
 		PyErr_Format(
 			PyExc_ValueError,
 			"Manifest had an entry with a zero-length filename.");
 		break;
 	case MANIFEST_TOO_SHORT_LINE:
 		PyErr_Format(
 			PyExc_ValueError,
 			"Manifest had implausibly-short line.");
 		break;
 	default:
 		PyErr_Format(PyExc_ValueError,
 			     "Unknown problem parsing manifest.");
 	}
 	return ret == 0 ? 0 : -1;
 }
 
 static void lazymanifest_dealloc(lazymanifest *self)
 {
 	/* free any extra lines we had to allocate */
 	int i;
 	for (i = 0; self->lines && (i < self->numlines); i++) {
 		if (self->lines[i].from_malloc) {
 			free(self->lines[i].start);
 		}
 	}
 	free(self->lines);
 	self->lines = NULL;
 	if (self->pydata) {
 		Py_DECREF(self->pydata);
 		self->pydata = NULL;
 	}
 	PyObject_Del(self);
 }
 
 /* iteration support */
 
 typedef struct {
 	PyObject_HEAD lazymanifest *m;
 	Py_ssize_t pos;
 } lmIter;
 
 static void lmiter_dealloc(PyObject *o)
 {
 	lmIter *self = (lmIter *)o;
 	Py_DECREF(self->m);
 	PyObject_Del(self);
 }
 
 static line *lmiter_nextline(lmIter *self)
 {
 	do {
 		self->pos++;
 		if (self->pos >= self->m->numlines) {
 			return NULL;
 		}
 		/* skip over deleted manifest entries */
 	} while (self->m->lines[self->pos].deleted);
 	return self->m->lines + self->pos;
 }
 
 static PyObject *lmiter_iterentriesnext(PyObject *o)
 {
 	Py_ssize_t pl;
 	line *l;
+	char flag;
 	Py_ssize_t consumed;
 	PyObject *ret = NULL, *path = NULL, *hash = NULL, *flags = NULL;
 	l = lmiter_nextline((lmIter *)o);
 	if (!l) {
 		goto done;
 	}
 	pl = pathlen(l);
 	path = PyBytes_FromStringAndSize(l->start, pl);
-	hash = nodeof(l);
+	hash = nodeof(l, &flag);
 	if (!path || !hash) {
 		goto done;
 	}
-	consumed = pl + 41;
-	flags = PyBytes_FromStringAndSize(l->start + consumed,
-					   l->len - consumed - 1);
+	flags = PyBytes_FromStringAndSize(&flag, flag ? 1 : 0);
 	if (!flags) {
 		goto done;
 	}
 	ret = PyTuple_Pack(3, path, hash, flags);
 done:
 	Py_XDECREF(path);
 	Py_XDECREF(hash);
 	Py_XDECREF(flags);
 	return ret;
 }
 
 #ifdef IS_PY3K
 #define LAZYMANIFESTENTRIESITERATOR_TPFLAGS Py_TPFLAGS_DEFAULT
 #else
 #define LAZYMANIFESTENTRIESITERATOR_TPFLAGS Py_TPFLAGS_DEFAULT \
 	| Py_TPFLAGS_HAVE_ITER
 #endif
 
 static PyTypeObject lazymanifestEntriesIterator = {
 	PyVarObject_HEAD_INIT(NULL, 0) /* header */
 	"parsers.lazymanifest.entriesiterator", /*tp_name */
 	sizeof(lmIter),                  /*tp_basicsize */
 	0,                               /*tp_itemsize */
 	lmiter_dealloc,                  /*tp_dealloc */
 	0,                               /*tp_print */
 	0,                               /*tp_getattr */
 	0,                               /*tp_setattr */
 	0,                               /*tp_compare */
 	0,                               /*tp_repr */
 	0,                               /*tp_as_number */
 	0,                               /*tp_as_sequence */
 	0,                               /*tp_as_mapping */
 	0,                               /*tp_hash */
 	0,                               /*tp_call */
 	0,                               /*tp_str */
 	0,                               /*tp_getattro */
 	0,                               /*tp_setattro */
 	0,                               /*tp_as_buffer */
 	LAZYMANIFESTENTRIESITERATOR_TPFLAGS, /* tp_flags */
 	"Iterator for 3-tuples in a lazymanifest.",  /* tp_doc */
 	0,                               /* tp_traverse */
 	0,                               /* tp_clear */
 	0,                               /* tp_richcompare */
 	0,                               /* tp_weaklistoffset */
 	PyObject_SelfIter,               /* tp_iter: __iter__() method */
 	lmiter_iterentriesnext,          /* tp_iternext: next() method */
 };
 
 static PyObject *lmiter_iterkeysnext(PyObject *o)
 {
 	Py_ssize_t pl;
 	line *l = lmiter_nextline((lmIter *)o);
 	if (!l) {
 		return NULL;
 	}
 	pl = pathlen(l);
 	return PyBytes_FromStringAndSize(l->start, pl);
 }
 
 #ifdef IS_PY3K
 #define LAZYMANIFESTKEYSITERATOR_TPFLAGS Py_TPFLAGS_DEFAULT
 #else
 #define LAZYMANIFESTKEYSITERATOR_TPFLAGS Py_TPFLAGS_DEFAULT \
 	| Py_TPFLAGS_HAVE_ITER
 #endif
 
 static PyTypeObject lazymanifestKeysIterator = {
 	PyVarObject_HEAD_INIT(NULL, 0) /* header */
 	"parsers.lazymanifest.keysiterator", /*tp_name */
 	sizeof(lmIter),                  /*tp_basicsize */
 	0,                               /*tp_itemsize */
 	lmiter_dealloc,                  /*tp_dealloc */
 	0,                               /*tp_print */
 	0,                               /*tp_getattr */
 	0,                               /*tp_setattr */
 	0,                               /*tp_compare */
 	0,                               /*tp_repr */
 	0,                               /*tp_as_number */
 	0,                               /*tp_as_sequence */
 	0,                               /*tp_as_mapping */
 	0,                               /*tp_hash */
 	0,                               /*tp_call */
 	0,                               /*tp_str */
 	0,                               /*tp_getattro */
 	0,                               /*tp_setattro */
 	0,                               /*tp_as_buffer */
 	LAZYMANIFESTKEYSITERATOR_TPFLAGS, /* tp_flags */
 	"Keys iterator for a lazymanifest.",  /* tp_doc */
 	0,                               /* tp_traverse */
 	0,                               /* tp_clear */
 	0,                               /* tp_richcompare */
 	0,                               /* tp_weaklistoffset */
 	PyObject_SelfIter,               /* tp_iter: __iter__() method */
 	lmiter_iterkeysnext,             /* tp_iternext: next() method */
 };
 
 static lazymanifest *lazymanifest_copy(lazymanifest *self);
 
 static PyObject *lazymanifest_getentriesiter(lazymanifest *self)
 {
 	lmIter *i = NULL;
 	lazymanifest *t = lazymanifest_copy(self);
 	if (!t) {
 		PyErr_NoMemory();
 		return NULL;
 	}
 	i = PyObject_New(lmIter, &lazymanifestEntriesIterator);
 	if (i) {
 		i->m = t;
 		i->pos = -1;
 	} else {
 		Py_DECREF(t);
 		PyErr_NoMemory();
 	}
 	return (PyObject *)i;
 }
 
 static PyObject *lazymanifest_getkeysiter(lazymanifest *self)
 {
 	lmIter *i = NULL;
 	lazymanifest *t = lazymanifest_copy(self);
 	if (!t) {
 		PyErr_NoMemory();
 		return NULL;
 	}
 	i = PyObject_New(lmIter, &lazymanifestKeysIterator);
 	if (i) {
 		i->m = t;
 		i->pos = -1;
 	} else {
 		Py_DECREF(t);
 		PyErr_NoMemory();
 	}
 	return (PyObject *)i;
 }
 
 /* __getitem__ and __setitem__ support */
 
 static Py_ssize_t lazymanifest_size(lazymanifest *self)
 {
 	return self->livelines;
 }
 
 static int linecmp(const void *left, const void *right)
 {
 	return strcmp(((const line *)left)->start,
 		      ((const line *)right)->start);
 }
 
 static PyObject *lazymanifest_getitem(lazymanifest *self, PyObject *key)
 {
 	line needle;
 	line *hit;
 	if (!PyBytes_Check(key)) {
 		PyErr_Format(PyExc_TypeError,
 			     "getitem: manifest keys must be a string.");
 		return NULL;
 	}
 	needle.start = PyBytes_AsString(key);
 	hit = bsearch(&needle, self->lines, self->numlines, sizeof(line),
 		      &linecmp);
 	if (!hit || hit->deleted) {
 		PyErr_Format(PyExc_KeyError, "No such manifest entry.");
 		return NULL;
 	}
 	return hashflags(hit);
 }
 
 static int lazymanifest_delitem(lazymanifest *self, PyObject *key)
 {
 	line needle;
 	line *hit;
 	if (!PyBytes_Check(key)) {
 		PyErr_Format(PyExc_TypeError,
 			     "delitem: manifest keys must be a string.");
 		return -1;
 	}
 	needle.start = PyBytes_AsString(key);
 	hit = bsearch(&needle, self->lines, self->numlines, sizeof(line),
 		      &linecmp);
 	if (!hit || hit->deleted) {
 		PyErr_Format(PyExc_KeyError,
 			     "Tried to delete nonexistent manifest entry.");
 		return -1;
 	}
 	self->dirty = true;
 	hit->deleted = true;
 	self->livelines--;
 	return 0;
 }
 
 /* Do a binary search for the insertion point for new, creating the
  * new entry if needed. */
 static int internalsetitem(lazymanifest *self, line *new)
 {
 	int start = 0, end = self->numlines;
 	while (start < end) {
 		int pos = start + (end - start) / 2;
 		int c = linecmp(new, self->lines + pos);
 		if (c < 0)
 			end = pos;
 		else if (c > 0)
 			start = pos + 1;
 		else {
 			if (self->lines[pos].deleted)
 				self->livelines++;
 			if (self->lines[pos].from_malloc)
 				free(self->lines[pos].start);
 			start = pos;
 			goto finish;
 		}
 	}
 	/* being here means we need to do an insert */
 	if (!realloc_if_full(self)) {
 		PyErr_NoMemory();
 		return -1;
 	}
 	memmove(self->lines + start + 1, self->lines + start,
 		(self->numlines - start) * sizeof(line));
 	self->numlines++;
 	self->livelines++;
 finish:
 	self->lines[start] = *new;
 	self->dirty = true;
 	return 0;
 }
 
 static int lazymanifest_setitem(
 	lazymanifest *self, PyObject *key, PyObject *value)
 {
 	char *path;
 	Py_ssize_t plen;
 	PyObject *pyhash;
 	Py_ssize_t hlen;
 	char *hash;
 	PyObject *pyflags;
 	char *flags;
 	Py_ssize_t flen;
 	Py_ssize_t dlen;
 	char *dest;
 	int i;
 	line new;
 	if (!PyBytes_Check(key)) {
 		PyErr_Format(PyExc_TypeError,
 			     "setitem: manifest keys must be a string.");
 		return -1;
 	}
 	if (!value) {
 		return lazymanifest_delitem(self, key);
 	}
 	if (!PyTuple_Check(value) || PyTuple_Size(value) != 2) {
 		PyErr_Format(PyExc_TypeError,
 			     "Manifest values must be a tuple of (node, flags).");
 		return -1;
 	}
 	if (PyBytes_AsStringAndSize(key, &path, &plen) == -1) {
 		return -1;
 	}
 
 	pyhash = PyTuple_GetItem(value, 0);
 	if (!PyBytes_Check(pyhash)) {
 		PyErr_Format(PyExc_TypeError,
-			     "node must be a 20-byte string");
+			     "node must be a 20 or 32 bytes string");
 		return -1;
 	}
 	hlen = PyBytes_Size(pyhash);
-	/* Some parts of the codebase try and set 21 or 22
-	 * byte "hash" values in order to perturb things for
-	 * status. We have to preserve at least the 21st
-	 * byte. Sigh. If there's a 22nd byte, we drop it on
-	 * the floor, which works fine.
-	 */
-	if (hlen != 20 && hlen != 21 && hlen != 22) {
+	if (hlen != 20 && hlen != 32) {
 		PyErr_Format(PyExc_TypeError,
-			     "node must be a 20-byte string");
+			     "node must be a 20 or 32 bytes string");
 		return -1;
 	}
 	hash = PyBytes_AsString(pyhash);
 
 	pyflags = PyTuple_GetItem(value, 1);
 	if (!PyBytes_Check(pyflags) || PyBytes_Size(pyflags) > 1) {
 		PyErr_Format(PyExc_TypeError,
-			     "flags must a 0 or 1 byte string");
+			     "flags must a 0 or 1 bytes string");
 		return -1;
 	}
 	if (PyBytes_AsStringAndSize(pyflags, &flags, &flen) == -1) {
 		return -1;
 	}
+	if (flen == 1) {
+		switch (*flags) {
+		case 'l':
+		case 't':
+		case 'x':
+			break;
+		default:
+			PyErr_Format(PyExc_TypeError, "invalid manifest flag");
+			return -1;
+		}
+	}
 	/* one null byte and one newline */
-	dlen = plen + 41 + flen + 1;
+	dlen = plen + hlen * 2 + 1 + flen + 1;
 	dest = malloc(dlen);
 	if (!dest) {
 		PyErr_NoMemory();
 		return -1;
 	}
 	memcpy(dest, path, plen + 1);
-	for (i = 0; i < 20; i++) {
+	for (i = 0; i < hlen; i++) {
 		/* Cast to unsigned, so it will not get sign-extended when promoted
 		 * to int (as is done when passing to a variadic function)
 		 */
 		sprintf(dest + plen + 1 + (i * 2), "%02x", (unsigned char)hash[i]);
 	}
-	memcpy(dest + plen + 41, flags, flen);
-	dest[plen + 41 + flen] = '\n';
+	memcpy(dest + plen + 2 * hlen + 1, flags, flen);
+	dest[plen + 2 * hlen + 1 + flen] = '\n';
 	new.start = dest;
 	new.len = dlen;
 	new.hash_suffix = '\0';
 	if (hlen > 20) {
 		new.hash_suffix = hash[20];
 	}
 	new.from_malloc = true;     /* is `start` a pointer we allocated? */
 	new.deleted = false;        /* is this entry deleted? */
 	if (internalsetitem(self, &new)) {
 		return -1;
 	}
 	return 0;
 }
 
 static PyMappingMethods lazymanifest_mapping_methods = {
 	(lenfunc)lazymanifest_size,             /* mp_length */
 	(binaryfunc)lazymanifest_getitem,       /* mp_subscript */
 	(objobjargproc)lazymanifest_setitem,    /* mp_ass_subscript */
 };
 
 /* sequence methods (important or __contains__ builds an iterator) */
 
 static int lazymanifest_contains(lazymanifest *self, PyObject *key)
 {
 	line needle;
 	line *hit;
 	if (!PyBytes_Check(key)) {
 		/* Our keys are always strings, so if the contains
 		 * check is for a non-string, just return false. */
 		return 0;
 	}
 	needle.start = PyBytes_AsString(key);
 	hit = bsearch(&needle, self->lines, self->numlines, sizeof(line),
 		      &linecmp);
 	if (!hit || hit->deleted) {
 		return 0;
 	}
 	return 1;
 }
 
 static PySequenceMethods lazymanifest_seq_meths = {
 	(lenfunc)lazymanifest_size, /* sq_length */
 	0, /* sq_concat */
 	0, /* sq_repeat */
 	0, /* sq_item */
 	0, /* sq_slice */
 	0, /* sq_ass_item */
 	0, /* sq_ass_slice */
 	(objobjproc)lazymanifest_contains, /* sq_contains */
 	0, /* sq_inplace_concat */
 	0, /* sq_inplace_repeat */
 };
 
 
 /* Other methods (copy, diff, etc) */
 static PyTypeObject lazymanifestType;
 
 /* If the manifest has changes, build the new manifest text and reindex it. */
 static int compact(lazymanifest *self)
 {
 	int i;
 	ssize_t need = 0;
 	char *data;
 	line *src, *dst;
 	PyObject *pydata;
 	if (!self->dirty)
 		return 0;
 	for (i = 0; i < self->numlines; i++) {
 		if (!self->lines[i].deleted) {
 			need += self->lines[i].len;
 		}
 	}
 	pydata = PyBytes_FromStringAndSize(NULL, need);
 	if (!pydata)
 		return -1;
 	data = PyBytes_AsString(pydata);
 	if (!data) {
 		return -1;
 	}
 	src = self->lines;
 	dst = self->lines;
 	for (i = 0; i < self->numlines; i++, src++) {
 		char *tofree = NULL;
 		if (src->from_malloc) {
 			tofree = src->start;
 		}
 		if (!src->deleted) {
 			memcpy(data, src->start, src->len);
 			*dst = *src;
 			dst->start = data;
 			dst->from_malloc = false;
 			data += dst->len;
 			dst++;
 		}
 		free(tofree);
 	}
 	Py_DECREF(self->pydata);
 	self->pydata = pydata;
 	self->numlines = self->livelines;
 	self->dirty = false;
 	return 0;
 }
 
 static PyObject *lazymanifest_text(lazymanifest *self)
 {
 	if (compact(self) != 0) {
 		PyErr_NoMemory();
 		return NULL;
 	}
 	Py_INCREF(self->pydata);
 	return self->pydata;
 }
 
 static lazymanifest *lazymanifest_copy(lazymanifest *self)
 {
 	lazymanifest *copy = NULL;
 	if (compact(self) != 0) {
 		goto nomem;
 	}
 	copy = PyObject_New(lazymanifest, &lazymanifestType);
 	if (!copy) {
 		goto nomem;
 	}
 	lazymanifest_init_early(copy);
 	copy->numlines = self->numlines;
 	copy->livelines = self->livelines;
 	copy->dirty = false;
 	copy->lines = malloc(self->maxlines *sizeof(line));
 	if (!copy->lines) {
 		goto nomem;
 	}
 	memcpy(copy->lines, self->lines, self->numlines * sizeof(line));
 	copy->maxlines = self->maxlines;
 	copy->pydata = self->pydata;
 	Py_INCREF(copy->pydata);
 	return copy;
 nomem:
 	PyErr_NoMemory();
 	Py_XDECREF(copy);
 	return NULL;
 }
 
 static lazymanifest *lazymanifest_filtercopy(
 	lazymanifest *self, PyObject *matchfn)
 {
 	lazymanifest *copy = NULL;
 	int i;
 	if (!PyCallable_Check(matchfn)) {
 		PyErr_SetString(PyExc_TypeError, "matchfn must be callable");
 		return NULL;
 	}
 	/* compact ourselves first to avoid double-frees later when we
 	 * compact tmp so that it doesn't have random pointers to our
 	 * underlying from_malloc-data (self->pydata is safe) */
 	if (compact(self) != 0) {
 		goto nomem;
 	}
 	copy = PyObject_New(lazymanifest, &lazymanifestType);
 	if (!copy) {
 		goto nomem;
 	}
 	lazymanifest_init_early(copy);
 	copy->dirty = true;
 	copy->lines = malloc(self->maxlines * sizeof(line));
 	if (!copy->lines) {
 		goto nomem;
 	}
 	copy->maxlines = self->maxlines;
 	copy->numlines = 0;
 	copy->pydata = self->pydata;
 	Py_INCREF(copy->pydata);
 	for (i = 0; i < self->numlines; i++) {
 		PyObject *arglist = NULL, *result = NULL;
 		arglist = Py_BuildValue(PY23("(s)", "(y)"),
 					self->lines[i].start);
 		if (!arglist) {
 			goto bail;
 		}
 		result = PyObject_CallObject(matchfn, arglist);
 		Py_DECREF(arglist);
 		/* if the callback raised an exception, just let it
 		 * through and give up */
 		if (!result) {
 			goto bail;
 		}
 		if (PyObject_IsTrue(result)) {
 			assert(!(self->lines[i].from_malloc));
 			copy->lines[copy->numlines++] = self->lines[i];
 		}
 		Py_DECREF(result);
 	}
 	copy->livelines = copy->numlines;
 	return copy;
 nomem:
 	PyErr_NoMemory();
 bail:
 	Py_XDECREF(copy);
 	return NULL;
 }
 
 static PyObject *lazymanifest_diff(lazymanifest *self, PyObject *args)
 {
 	lazymanifest *other;
 	PyObject *pyclean = NULL;
 	bool listclean;
 	PyObject *emptyTup = NULL, *ret = NULL;
 	PyObject *es;
 	int sneedle = 0, oneedle = 0;
 	if (!PyArg_ParseTuple(args, "O!|O", &lazymanifestType, &other, &pyclean)) {
 		return NULL;
 	}
 	listclean = (!pyclean) ? false : PyObject_IsTrue(pyclean);
 	es = PyBytes_FromString("");
 	if (!es) {
 		goto nomem;
 	}
 	emptyTup = PyTuple_Pack(2, Py_None, es);
 	Py_DECREF(es);
 	if (!emptyTup) {
 		goto nomem;
 	}
 	ret = PyDict_New();
 	if (!ret) {
 		goto nomem;
 	}
 	while (sneedle != self->numlines || oneedle != other->numlines) {
 		line *left = self->lines + sneedle;
 		line *right = other->lines + oneedle;
 		int result;
 		PyObject *key;
 		PyObject *outer;
 		/* If we're looking at a deleted entry and it's not
 		 * the end of the manifest, just skip it. */
 		if (sneedle < self->numlines && left->deleted) {
 			sneedle++;
 			continue;
 		}
 		if (oneedle < other->numlines && right->deleted) {
 			oneedle++;
 			continue;
 		}
 		/* if we're at the end of either manifest, then we
 		 * know the remaining items are adds so we can skip
 		 * the strcmp. */
 		if (sneedle == self->numlines) {
 			result = 1;
 		} else if (oneedle == other->numlines) {
 			result = -1;
 		} else {
 			result = linecmp(left, right);
 		}
 		key = result <= 0 ?
 			PyBytes_FromString(left->start) :
 			PyBytes_FromString(right->start);
 		if (!key)
 			goto nomem;
 		if (result < 0) {
 			PyObject *l = hashflags(left);
 			if (!l) {
 				goto nomem;
 			}
 			outer = PyTuple_Pack(2, l, emptyTup);
 			Py_DECREF(l);
 			if (!outer) {
 				goto nomem;
 			}
 			PyDict_SetItem(ret, key, outer);
 			Py_DECREF(outer);
 			sneedle++;
 		} else if (result > 0) {
 			PyObject *r = hashflags(right);
 			if (!r) {
 				goto nomem;
 			}
 			outer = PyTuple_Pack(2, emptyTup, r);
 			Py_DECREF(r);
 			if (!outer) {
 				goto nomem;
 			}
 			PyDict_SetItem(ret, key, outer);
 			Py_DECREF(outer);
 			oneedle++;
 		} else {
 			/* file exists in both manifests */
 			if (left->len != right->len
 			    || memcmp(left->start, right->start, left->len)
 			    || left->hash_suffix != right->hash_suffix) {
 				PyObject *l = hashflags(left);
 				PyObject *r;
 				if (!l) {
 					goto nomem;
 				}
 				r = hashflags(right);
 				if (!r) {
 					Py_DECREF(l);
 					goto nomem;
 				}
 				outer = PyTuple_Pack(2, l, r);
 				Py_DECREF(l);
 				Py_DECREF(r);
 				if (!outer) {
 					goto nomem;
 				}
 				PyDict_SetItem(ret, key, outer);
 				Py_DECREF(outer);
 			} else if (listclean) {
 				PyDict_SetItem(ret, key, Py_None);
 			}
 			sneedle++;
 			oneedle++;
 		}
 		Py_DECREF(key);
 	}
 	Py_DECREF(emptyTup);
 	return ret;
 nomem:
 	PyErr_NoMemory();
 	Py_XDECREF(ret);
 	Py_XDECREF(emptyTup);
 	return NULL;
 }
 
 static PyMethodDef lazymanifest_methods[] = {
 	{"iterkeys", (PyCFunction)lazymanifest_getkeysiter, METH_NOARGS,
 	 "Iterate over file names in this lazymanifest."},
 	{"iterentries", (PyCFunction)lazymanifest_getentriesiter, METH_NOARGS,
 	 "Iterate over (path, nodeid, flags) tuples in this lazymanifest."},
 	{"copy", (PyCFunction)lazymanifest_copy, METH_NOARGS,
 	 "Make a copy of this lazymanifest."},
 	{"filtercopy", (PyCFunction)lazymanifest_filtercopy, METH_O,
 	 "Make a copy of this manifest filtered by matchfn."},
 	{"diff", (PyCFunction)lazymanifest_diff, METH_VARARGS,
 	 "Compare this lazymanifest to another one."},
 	{"text", (PyCFunction)lazymanifest_text, METH_NOARGS,
 	 "Encode this manifest to text."},
 	{NULL},
 };
 
 #ifdef IS_PY3K
 #define LAZYMANIFEST_TPFLAGS Py_TPFLAGS_DEFAULT
 #else
 #define LAZYMANIFEST_TPFLAGS Py_TPFLAGS_DEFAULT | Py_TPFLAGS_HAVE_SEQUENCE_IN
 #endif
 
 static PyTypeObject lazymanifestType = {
 	PyVarObject_HEAD_INIT(NULL, 0) /* header */
 	"parsers.lazymanifest",                           /* tp_name */
 	sizeof(lazymanifest),                             /* tp_basicsize */
 	0,                                                /* tp_itemsize */
 	(destructor)lazymanifest_dealloc,                 /* tp_dealloc */
 	0,                                                /* tp_print */
 	0,                                                /* tp_getattr */
 	0,                                                /* tp_setattr */
 	0,                                                /* tp_compare */
 	0,                                                /* tp_repr */
 	0,                                                /* tp_as_number */
 	&lazymanifest_seq_meths,                          /* tp_as_sequence */
 	&lazymanifest_mapping_methods,                    /* tp_as_mapping */
 	0,                                                /* tp_hash */
 	0,                                                /* tp_call */
 	0,                                                /* tp_str */
 	0,                                                /* tp_getattro */
 	0,                                                /* tp_setattro */
 	0,                                                /* tp_as_buffer */
 	LAZYMANIFEST_TPFLAGS,                             /* tp_flags */
 	"TODO(augie)",                                    /* tp_doc */
 	0,                                                /* tp_traverse */
 	0,                                                /* tp_clear */
 	0,                                                /* tp_richcompare */
 	0,                                             /* tp_weaklistoffset */
 	(getiterfunc)lazymanifest_getkeysiter,                /* tp_iter */
 	0,                                                /* tp_iternext */
 	lazymanifest_methods,                             /* tp_methods */
 	0,                                                /* tp_members */
 	0,                                                /* tp_getset */
 	0,                                                /* tp_base */
 	0,                                                /* tp_dict */
 	0,                                                /* tp_descr_get */
 	0,                                                /* tp_descr_set */
 	0,                                                /* tp_dictoffset */
 	(initproc)lazymanifest_init,                      /* tp_init */
 	0,                                                /* tp_alloc */
 };
 
 void manifest_module_init(PyObject * mod)
 {
 	lazymanifestType.tp_new = PyType_GenericNew;
 	if (PyType_Ready(&lazymanifestType) < 0)
 		return;
 	Py_INCREF(&lazymanifestType);
 
 	PyModule_AddObject(mod, "lazymanifest",
 			   (PyObject *)&lazymanifestType);
 }

mercurial/manifest.py

@@ -1,2329 +1,2340 b''
 # manifest.py - manifest revision class for mercurial
 #
 # Copyright 2005-2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import heapq
 import itertools
 import struct
 import weakref
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullid,
     nullrev,
 )
 from .pycompat import getattr
 from . import (
     encoding,
     error,
     match as matchmod,
     mdiff,
     pathutil,
     policy,
     pycompat,
     revlog,
     util,
 )
 from .interfaces import (
     repository,
     util as interfaceutil,
 )
 
 parsers = policy.importmod('parsers')
 propertycache = util.propertycache
 
 # Allow tests to more easily test the alternate path in manifestdict.fastdelta()
 FASTDELTA_TEXTDIFF_THRESHOLD = 1000
 
 
 def _parse(data):
     # This method does a little bit of excessive-looking
     # precondition checking. This is so that the behavior of this
     # class exactly matches its C counterpart to try and help
     # prevent surprise breakage for anyone that develops against
     # the pure version.
     if data and data[-1:] != b'\n':
         raise ValueError(b'Manifest did not end in a newline.')
     prev = None
     for l in data.splitlines():
         if prev is not None and prev > l:
             raise ValueError(b'Manifest lines not in sorted order.')
         prev = l
         f, n = l.split(b'\0')
         nl = len(n)
         if 64 < nl:
             # modern hash, full width
             yield f, bin(n[:64]), n[64:]
         elif 40 < nl < 45:
             # legacy hash, always sha1
             yield f, bin(n[:40]), n[40:]
         else:
             yield f, bin(n), b''
 
 
 def _text(it):
     files = []
     lines = []
     for f, n, fl in it:
         files.append(f)
         # if this is changed to support newlines in filenames,
         # be sure to check the templates/ dir again (especially *-raw.tmpl)
         lines.append(b"%s\0%s%s\n" % (f, hex(n), fl))
 
     _checkforbidden(files)
     return b''.join(lines)
 
 
 class lazymanifestiter(object):
     def __init__(self, lm):
         self.pos = 0
         self.lm = lm
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data[0]
         self.pos += 1
         zeropos = data.find(b'\x00', pos)
         return data[pos:zeropos]
 
     __next__ = next
 
 
 class lazymanifestiterentries(object):
     def __init__(self, lm):
         self.lm = lm
         self.pos = 0
 
     def __iter__(self):
         return self
 
     def next(self):
         try:
             data, pos = self.lm._get(self.pos)
         except IndexError:
             raise StopIteration
         if pos == -1:
             self.pos += 1
             return data
         zeropos = data.find(b'\x00', pos)
-        hashval = unhexlify(data, self.lm.extrainfo[self.pos], zeropos + 1, 40)
-        flags = self.lm._getflags(data, self.pos, zeropos)
+        nlpos = data.find(b'\n', pos)
+        if zeropos == -1 or nlpos == -1 or nlpos < zeropos:
+            raise error.StorageError(b'Invalid manifest line')
+        flags = data[nlpos - 1 : nlpos]
+        if flags in _manifestflags:
+            hlen = nlpos - zeropos - 2
+        else:
+            hlen = nlpos - zeropos - 1
+            flags = b''
+        if hlen not in (40, 64):
+            raise error.StorageError(b'Invalid manifest line')
+        hashval = unhexlify(
+            data, self.lm.extrainfo[self.pos], zeropos + 1, hlen
+        )
         self.pos += 1
         return (data[pos:zeropos], hashval, flags)
 
     __next__ = next
 
 
 def unhexlify(data, extra, pos, length):
     s = bin(data[pos : pos + length])
     if extra:
         s += chr(extra & 0xFF)
     return s
 
 
 def _cmp(a, b):
     return (a > b) - (a < b)
 
 
+_manifestflags = {b'', b'l', b't', b'x'}
+
+
 class _lazymanifest(object):
     """A pure python manifest backed by a byte string.  It is supplimented with
     internal lists as it is modified, until it is compacted back to a pure byte
     string.
 
     ``data`` is the initial manifest data.
 
     ``positions`` is a list of offsets, one per manifest entry.  Positive
     values are offsets into ``data``, negative values are offsets into the
     ``extradata`` list.  When an entry is removed, its entry is dropped from
     ``positions``.  The values are encoded such that when walking the list and
     indexing into ``data`` or ``extradata`` as appropriate, the entries are
     sorted by filename.
 
     ``extradata`` is a list of (key, hash, flags) for entries that were added or
     modified since the manifest was created or compacted.
     """
 
     def __init__(
         self,
         data,
         positions=None,
         extrainfo=None,
         extradata=None,
         hasremovals=False,
     ):
         if positions is None:
             self.positions = self.findlines(data)
             self.extrainfo = [0] * len(self.positions)
             self.data = data
             self.extradata = []
             self.hasremovals = False
         else:
             self.positions = positions[:]
             self.extrainfo = extrainfo[:]
             self.extradata = extradata[:]
             self.data = data
             self.hasremovals = hasremovals
 
     def findlines(self, data):
         if not data:
             return []
         pos = data.find(b"\n")
         if pos == -1 or data[-1:] != b'\n':
             raise ValueError(b"Manifest did not end in a newline.")
         positions = [0]
         prev = data[: data.find(b'\x00')]
         while pos < len(data) - 1 and pos != -1:
             positions.append(pos + 1)
             nexts = data[pos + 1 : data.find(b'\x00', pos + 1)]
             if nexts < prev:
                 raise ValueError(b"Manifest lines not in sorted order.")
             prev = nexts
             pos = data.find(b"\n", pos + 1)
         return positions
 
     def _get(self, index):
         # get the position encoded in pos:
         #   positive number is an index in 'data'
         #   negative number is in extrapieces
         pos = self.positions[index]
         if pos >= 0:
             return self.data, pos
         return self.extradata[-pos - 1], -1
 
     def _getkey(self, pos):
         if pos >= 0:
             return self.data[pos : self.data.find(b'\x00', pos + 1)]
         return self.extradata[-pos - 1][0]
 
     def bsearch(self, key):
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last) // 2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return midpoint
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return -1
 
     def bsearch2(self, key):
         # same as the above, but will always return the position
         # done for performance reasons
         first = 0
         last = len(self.positions) - 1
 
         while first <= last:
             midpoint = (first + last) // 2
             nextpos = self.positions[midpoint]
             candidate = self._getkey(nextpos)
             r = _cmp(key, candidate)
             if r == 0:
                 return (midpoint, True)
             else:
                 if r < 0:
                     last = midpoint - 1
                 else:
                     first = midpoint + 1
         return (first, False)
 
     def __contains__(self, key):
         return self.bsearch(key) != -1
 
-    def _getflags(self, data, needle, pos):
-        start = pos + 41
-        end = data.find(b"\n", start)
-        if end == -1:
-            end = len(data) - 1
-        if start == end:
-            return b''
-        return self.data[start:end]
-
     def __getitem__(self, key):
         if not isinstance(key, bytes):
             raise TypeError(b"getitem: manifest keys must be a bytes.")
         needle = self.bsearch(key)
         if needle == -1:
             raise KeyError
         data, pos = self._get(needle)
         if pos == -1:
             return (data[1], data[2])
         zeropos = data.find(b'\x00', pos)
         nlpos = data.find(b'\n', zeropos)
         assert 0 <= needle <= len(self.positions)
         assert len(self.extrainfo) == len(self.positions)
+        if zeropos == -1 or nlpos == -1 or nlpos < zeropos:
+            raise error.StorageError(b'Invalid manifest line')
         hlen = nlpos - zeropos - 1
-        # Hashes sometimes have an extra byte tucked on the end, so
-        # detect that.
-        if hlen % 2:
+        flags = data[nlpos - 1 : nlpos]
+        if flags in _manifestflags:
             hlen -= 1
+        else:
+            flags = b''
+        if hlen not in (40, 64):
+            raise error.StorageError(b'Invalid manifest line')
         hashval = unhexlify(data, self.extrainfo[needle], zeropos + 1, hlen)
-        flags = self._getflags(data, needle, zeropos)
         return (hashval, flags)
 
     def __delitem__(self, key):
         needle, found = self.bsearch2(key)
         if not found:
             raise KeyError
         cur = self.positions[needle]
         self.positions = self.positions[:needle] + self.positions[needle + 1 :]
         self.extrainfo = self.extrainfo[:needle] + self.extrainfo[needle + 1 :]
         if cur >= 0:
             # This does NOT unsort the list as far as the search functions are
             # concerned, as they only examine lines mapped by self.positions.
             self.data = self.data[:cur] + b'\x00' + self.data[cur + 1 :]
             self.hasremovals = True
 
     def __setitem__(self, key, value):
         if not isinstance(key, bytes):
             raise TypeError(b"setitem: manifest keys must be a byte string.")
         if not isinstance(value, tuple) or len(value) != 2:
             raise TypeError(
                 b"Manifest values must be a tuple of (node, flags)."
             )
         hashval = value[0]
         # hashes are either 20 or 32 bytes (sha1 or its replacement),
         # and allow one extra byte taht won't be persisted to disk but
         # is sometimes used in memory.
         if not isinstance(hashval, bytes) or not (
             20 <= len(hashval) <= 22 or 32 <= len(hashval) <= 34
         ):
             raise TypeError(b"node must be a 20-byte or 32-byte byte string")
         flags = value[1]
         if len(hashval) == 22:
             hashval = hashval[:-1]
         if not isinstance(flags, bytes) or len(flags) > 1:
             raise TypeError(b"flags must a 0 or 1 byte string, got %r", flags)
         needle, found = self.bsearch2(key)
         if found:
             # put the item
             pos = self.positions[needle]
             if pos < 0:
                 self.extradata[-pos - 1] = (key, hashval, value[1])
             else:
                 # just don't bother
                 self.extradata.append((key, hashval, value[1]))
                 self.positions[needle] = -len(self.extradata)
         else:
             # not found, put it in with extra positions
             self.extradata.append((key, hashval, value[1]))
             self.positions = (
                 self.positions[:needle]
                 + [-len(self.extradata)]
                 + self.positions[needle:]
             )
             self.extrainfo = (
                 self.extrainfo[:needle] + [0] + self.extrainfo[needle:]
             )
 
     def copy(self):
         # XXX call _compact like in C?
         return _lazymanifest(
             self.data,
             self.positions,
             self.extrainfo,
             self.extradata,
             self.hasremovals,
         )
 
     def _compact(self):
         # hopefully not called TOO often
         if len(self.extradata) == 0 and not self.hasremovals:
             return
         l = []
         i = 0
         offset = 0
         self.extrainfo = [0] * len(self.positions)
         while i < len(self.positions):
             if self.positions[i] >= 0:
                 cur = self.positions[i]
                 last_cut = cur
 
                 # Collect all contiguous entries in the buffer at the current
                 # offset, breaking out only for added/modified items held in
                 # extradata, or a deleted line prior to the next position.
                 while True:
                     self.positions[i] = offset
                     i += 1
                     if i == len(self.positions) or self.positions[i] < 0:
                         break
 
                     # A removed file has no positions[] entry, but does have an
                     # overwritten first byte.  Break out and find the end of the
                     # current good entry/entries if there is a removed file
                     # before the next position.
                     if (
                         self.hasremovals
                         and self.data.find(b'\n\x00', cur, self.positions[i])
                         != -1
                     ):
                         break
 
                     offset += self.positions[i] - cur
                     cur = self.positions[i]
                 end_cut = self.data.find(b'\n', cur)
                 if end_cut != -1:
                     end_cut += 1
                 offset += end_cut - cur
                 l.append(self.data[last_cut:end_cut])
             else:
                 while i < len(self.positions) and self.positions[i] < 0:
                     cur = self.positions[i]
                     t = self.extradata[-cur - 1]
                     l.append(self._pack(t))
                     self.positions[i] = offset
                     # Hashes are either 20 bytes (old sha1s) or 32
                     # bytes (new non-sha1).
                     hlen = 20
                     if len(t[1]) > 25:
                         hlen = 32
                     if len(t[1]) > hlen:
                         self.extrainfo[i] = ord(t[1][hlen + 1])
                     offset += len(l[-1])
                     i += 1
         self.data = b''.join(l)
         self.hasremovals = False
         self.extradata = []
 
     def _pack(self, d):
         n = d[1]
-        if len(n) == 21 or len(n) == 33:
-            n = n[:-1]
-        assert len(n) == 20 or len(n) == 32
+        assert len(n) in (20, 32)
         return d[0] + b'\x00' + hex(n) + d[2] + b'\n'
 
     def text(self):
         self._compact()
         return self.data
 
     def diff(self, m2, clean=False):
         '''Finds changes between the current manifest and m2.'''
         # XXX think whether efficiency matters here
         diff = {}
 
         for fn, e1, flags in self.iterentries():
             if fn not in m2:
                 diff[fn] = (e1, flags), (None, b'')
             else:
                 e2 = m2[fn]
                 if (e1, flags) != e2:
                     diff[fn] = (e1, flags), e2
                 elif clean:
                     diff[fn] = None
 
         for fn, e2, flags in m2.iterentries():
             if fn not in self:
                 diff[fn] = (None, b''), (e2, flags)
 
         return diff
 
     def iterentries(self):
         return lazymanifestiterentries(self)
 
     def iterkeys(self):
         return lazymanifestiter(self)
 
     def __iter__(self):
         return lazymanifestiter(self)
 
     def __len__(self):
         return len(self.positions)
 
     def filtercopy(self, filterfn):
         # XXX should be optimized
         c = _lazymanifest(b'')
         for f, n, fl in self.iterentries():
             if filterfn(f):
                 c[f] = n, fl
         return c
 
 
 try:
     _lazymanifest = parsers.lazymanifest
 except AttributeError:
     pass
 
 
 @interfaceutil.implementer(repository.imanifestdict)
 class manifestdict(object):
     def __init__(self, data=b''):
         self._lm = _lazymanifest(data)
 
     def __getitem__(self, key):
         return self._lm[key][0]
 
     def find(self, key):
         return self._lm[key]
 
     def __len__(self):
         return len(self._lm)
 
     def __nonzero__(self):
         # nonzero is covered by the __len__ function, but implementing it here
         # makes it easier for extensions to override.
         return len(self._lm) != 0
 
     __bool__ = __nonzero__
 
     def __setitem__(self, key, node):
         self._lm[key] = node, self.flags(key)
 
     def __contains__(self, key):
         if key is None:
             return False
         return key in self._lm
 
     def __delitem__(self, key):
         del self._lm[key]
 
     def __iter__(self):
         return self._lm.__iter__()
 
     def iterkeys(self):
         return self._lm.iterkeys()
 
     def keys(self):
         return list(self.iterkeys())
 
     def filesnotin(self, m2, match=None):
         '''Set of files in this manifest that are not in the other'''
         if match is not None:
             match = matchmod.badmatch(match, lambda path, msg: None)
             sm2 = set(m2.walk(match))
             return {f for f in self.walk(match) if f not in sm2}
         return {f for f in self if f not in m2}
 
     @propertycache
     def _dirs(self):
         return pathutil.dirs(self)
 
     def dirs(self):
         return self._dirs
 
     def hasdir(self, dir):
         return dir in self._dirs
 
     def _filesfastpath(self, match):
         '''Checks whether we can correctly and quickly iterate over matcher
         files instead of over manifest files.'''
         files = match.files()
         return len(files) < 100 and (
             match.isexact()
             or (match.prefix() and all(fn in self for fn in files))
         )
 
     def walk(self, match):
         '''Generates matching file names.
 
         Equivalent to manifest.matches(match).iterkeys(), but without creating
         an entirely new manifest.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         # avoid the entire walk if we're only looking for specific files
         if self._filesfastpath(match):
             for fn in sorted(fset):
                 if fn in self:
                     yield fn
             return
 
         for fn in self:
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             if match(fn):
                 yield fn
 
         # for dirstate.walk, files=[''] means "walk the whole tree".
         # follow that here, too
         fset.discard(b'')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _matches(self, match):
         '''generate a new manifest filtered by the match argument'''
         if match.always():
             return self.copy()
 
         if self._filesfastpath(match):
             m = manifestdict()
             lm = self._lm
             for fn in match.files():
                 if fn in lm:
                     m._lm[fn] = lm[fn]
             return m
 
         m = manifestdict()
         m._lm = self._lm.filtercopy(match)
         return m
 
     def diff(self, m2, match=None, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         if match:
             m1 = self._matches(match)
             m2 = m2._matches(match)
             return m1.diff(m2, clean=clean)
         return self._lm.diff(m2._lm, clean)
 
     def setflag(self, key, flag):
+        if flag not in _manifestflags:
+            raise TypeError(b"Invalid manifest flag set.")
         self._lm[key] = self[key], flag
 
     def get(self, key, default=None):
         try:
             return self._lm[key][0]
         except KeyError:
             return default
 
     def flags(self, key):
         try:
             return self._lm[key][1]
         except KeyError:
             return b''
 
     def copy(self):
         c = manifestdict()
         c._lm = self._lm.copy()
         return c
 
     def items(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iteritems(self):
         return (x[:2] for x in self._lm.iterentries())
 
     def iterentries(self):
         return self._lm.iterentries()
 
     def text(self):
         # most likely uses native version
         return self._lm.text()
 
     def fastdelta(self, base, changes):
         """Given a base manifest text as a bytearray and a list of changes
         relative to that text, compute a delta that can be used by revlog.
         """
         delta = []
         dstart = None
         dend = None
         dline = [b""]
         start = 0
         # zero copy representation of base as a buffer
         addbuf = util.buffer(base)
 
         changes = list(changes)
         if len(changes) < FASTDELTA_TEXTDIFF_THRESHOLD:
             # start with a readonly loop that finds the offset of
             # each line and creates the deltas
             for f, todelete in changes:
                 # bs will either be the index of the item or the insert point
                 start, end = _msearch(addbuf, f, start)
                 if not todelete:
                     h, fl = self._lm[f]
                     l = b"%s\0%s%s\n" % (f, hex(h), fl)
                 else:
                     if start == end:
                         # item we want to delete was not found, error out
                         raise AssertionError(
                             _(b"failed to remove %s from manifest") % f
                         )
                     l = b""
                 if dstart is not None and dstart <= start and dend >= start:
                     if dend < end:
                         dend = end
                     if l:
                         dline.append(l)
                 else:
                     if dstart is not None:
                         delta.append([dstart, dend, b"".join(dline)])
                     dstart = start
                     dend = end
                     dline = [l]
 
             if dstart is not None:
                 delta.append([dstart, dend, b"".join(dline)])
             # apply the delta to the base, and get a delta for addrevision
             deltatext, arraytext = _addlistdelta(base, delta)
         else:
             # For large changes, it's much cheaper to just build the text and
             # diff it.
             arraytext = bytearray(self.text())
             deltatext = mdiff.textdiff(
                 util.buffer(base), util.buffer(arraytext)
             )
 
         return arraytext, deltatext
 
 
 def _msearch(m, s, lo=0, hi=None):
     '''return a tuple (start, end) that says where to find s within m.
 
     If the string is found m[start:end] are the line containing
     that string.  If start == end the string was not found and
     they indicate the proper sorted insertion point.
 
     m should be a buffer, a memoryview or a byte string.
     s is a byte string'''
 
     def advance(i, c):
         while i < lenm and m[i : i + 1] != c:
             i += 1
         return i
 
     if not s:
         return (lo, lo)
     lenm = len(m)
     if not hi:
         hi = lenm
     while lo < hi:
         mid = (lo + hi) // 2
         start = mid
         while start > 0 and m[start - 1 : start] != b'\n':
             start -= 1
         end = advance(start, b'\0')
         if bytes(m[start:end]) < s:
             # we know that after the null there are 40 bytes of sha1
             # this translates to the bisect lo = mid + 1
             lo = advance(end + 40, b'\n') + 1
         else:
             # this translates to the bisect hi = mid
             hi = start
     end = advance(lo, b'\0')
     found = m[lo:end]
     if s == found:
         # we know that after the null there are 40 bytes of sha1
         end = advance(end + 40, b'\n')
         return (lo, end + 1)
     else:
         return (lo, lo)
 
 
 def _checkforbidden(l):
     """Check filenames for illegal characters."""
     for f in l:
         if b'\n' in f or b'\r' in f:
             raise error.StorageError(
                 _(b"'\\n' and '\\r' disallowed in filenames: %r")
                 % pycompat.bytestr(f)
             )
 
 
 # apply the changes collected during the bisect loop to our addlist
 # return a delta suitable for addrevision
 def _addlistdelta(addlist, x):
     # for large addlist arrays, building a new array is cheaper
     # than repeatedly modifying the existing one
     currentposition = 0
     newaddlist = bytearray()
 
     for start, end, content in x:
         newaddlist += addlist[currentposition:start]
         if content:
             newaddlist += bytearray(content)
 
         currentposition = end
 
     newaddlist += addlist[currentposition:]
 
     deltatext = b"".join(
         struct.pack(b">lll", start, end, len(content)) + content
         for start, end, content in x
     )
     return deltatext, newaddlist
 
 
 def _splittopdir(f):
     if b'/' in f:
         dir, subpath = f.split(b'/', 1)
         return dir + b'/', subpath
     else:
         return b'', f
 
 
 _noop = lambda s: None
 
 
 @interfaceutil.implementer(repository.imanifestdict)
 class treemanifest(object):
     def __init__(self, dir=b'', text=b''):
         self._dir = dir
         self._node = nullid
         self._loadfunc = _noop
         self._copyfunc = _noop
         self._dirty = False
         self._dirs = {}
         self._lazydirs = {}
         # Using _lazymanifest here is a little slower than plain old dicts
         self._files = {}
         self._flags = {}
         if text:
 
             def readsubtree(subdir, subm):
                 raise AssertionError(
                     b'treemanifest constructor only accepts flat manifests'
                 )
 
             self.parse(text, readsubtree)
             self._dirty = True  # Mark flat manifest dirty after parsing
 
     def _subpath(self, path):
         return self._dir + path
 
     def _loadalllazy(self):
         selfdirs = self._dirs
         for d, (path, node, readsubtree, docopy) in pycompat.iteritems(
             self._lazydirs
         ):
             if docopy:
                 selfdirs[d] = readsubtree(path, node).copy()
             else:
                 selfdirs[d] = readsubtree(path, node)
         self._lazydirs = {}
 
     def _loadlazy(self, d):
         v = self._lazydirs.get(d)
         if v:
             path, node, readsubtree, docopy = v
             if docopy:
                 self._dirs[d] = readsubtree(path, node).copy()
             else:
                 self._dirs[d] = readsubtree(path, node)
             del self._lazydirs[d]
 
     def _loadchildrensetlazy(self, visit):
         if not visit:
             return None
         if visit == b'all' or visit == b'this':
             self._loadalllazy()
             return None
 
         loadlazy = self._loadlazy
         for k in visit:
             loadlazy(k + b'/')
         return visit
 
     def _loaddifflazy(self, t1, t2):
         """load items in t1 and t2 if they're needed for diffing.
 
         The criteria currently is:
         - if it's not present in _lazydirs in either t1 or t2, load it in the
           other (it may already be loaded or it may not exist, doesn't matter)
         - if it's present in _lazydirs in both, compare the nodeid; if it
           differs, load it in both
         """
         toloadlazy = []
         for d, v1 in pycompat.iteritems(t1._lazydirs):
             v2 = t2._lazydirs.get(d)
             if not v2 or v2[1] != v1[1]:
                 toloadlazy.append(d)
         for d, v1 in pycompat.iteritems(t2._lazydirs):
             if d not in t1._lazydirs:
                 toloadlazy.append(d)
 
         for d in toloadlazy:
             t1._loadlazy(d)
             t2._loadlazy(d)
 
     def __len__(self):
         self._load()
         size = len(self._files)
         self._loadalllazy()
         for m in self._dirs.values():
             size += m.__len__()
         return size
 
     def __nonzero__(self):
         # Faster than "__len() != 0" since it avoids loading sub-manifests
         return not self._isempty()
 
     __bool__ = __nonzero__
 
     def _isempty(self):
         self._load()  # for consistency; already loaded by all callers
         # See if we can skip loading everything.
         if self._files or (
             self._dirs and any(not m._isempty() for m in self._dirs.values())
         ):
             return False
         self._loadalllazy()
         return not self._dirs or all(m._isempty() for m in self._dirs.values())
 
     @encoding.strmethod
     def __repr__(self):
         return (
             b'<treemanifest dir=%s, node=%s, loaded=%r, dirty=%r at 0x%x>'
             % (
                 self._dir,
                 hex(self._node),
                 bool(self._loadfunc is _noop),
                 self._dirty,
                 id(self),
             )
         )
 
     def dir(self):
         '''The directory that this tree manifest represents, including a
         trailing '/'. Empty string for the repo root directory.'''
         return self._dir
 
     def node(self):
         '''This node of this instance. nullid for unsaved instances. Should
         be updated when the instance is read or written from a revlog.
         '''
         assert not self._dirty
         return self._node
 
     def setnode(self, node):
         self._node = node
         self._dirty = False
 
     def iterentries(self):
         self._load()
         self._loadalllazy()
         for p, n in sorted(
             itertools.chain(self._dirs.items(), self._files.items())
         ):
             if p in self._files:
                 yield self._subpath(p), n, self._flags.get(p, b'')
             else:
                 for x in n.iterentries():
                     yield x
 
     def items(self):
         self._load()
         self._loadalllazy()
         for p, n in sorted(
             itertools.chain(self._dirs.items(), self._files.items())
         ):
             if p in self._files:
                 yield self._subpath(p), n
             else:
                 for f, sn in pycompat.iteritems(n):
                     yield f, sn
 
     iteritems = items
 
     def iterkeys(self):
         self._load()
         self._loadalllazy()
         for p in sorted(itertools.chain(self._dirs, self._files)):
             if p in self._files:
                 yield self._subpath(p)
             else:
                 for f in self._dirs[p]:
                     yield f
 
     def keys(self):
         return list(self.iterkeys())
 
     def __iter__(self):
         return self.iterkeys()
 
     def __contains__(self, f):
         if f is None:
             return False
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return False
 
             return self._dirs[dir].__contains__(subpath)
         else:
             return f in self._files
 
     def get(self, f, default=None):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return default
             return self._dirs[dir].get(subpath, default)
         else:
             return self._files.get(f, default)
 
     def __getitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             return self._dirs[dir].__getitem__(subpath)
         else:
             return self._files[f]
 
     def flags(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             if dir not in self._dirs:
                 return b''
             return self._dirs[dir].flags(subpath)
         else:
             if f in self._lazydirs or f in self._dirs:
                 return b''
             return self._flags.get(f, b'')
 
     def find(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             return self._dirs[dir].find(subpath)
         else:
             return self._files[f], self._flags.get(f, b'')
 
     def __delitem__(self, f):
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
 
             self._dirs[dir].__delitem__(subpath)
             # If the directory is now empty, remove it
             if self._dirs[dir]._isempty():
                 del self._dirs[dir]
         else:
             del self._files[f]
             if f in self._flags:
                 del self._flags[f]
         self._dirty = True
 
     def __setitem__(self, f, n):
         assert n is not None
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].__setitem__(subpath, n)
         else:
             # manifest nodes are either 20 bytes or 32 bytes,
-            # depending on the hash in use. An extra byte is
-            # occasionally used by hg, but won't ever be
-            # persisted. Trim to 21 or 33 bytes as appropriate.
-            trim = 21 if len(n) < 25 else 33
-            self._files[f] = n[:trim]  # to match manifestdict's behavior
+            # depending on the hash in use. Assert this as historically
+            # sometimes extra bytes were added.
+            assert len(n) in (20, 32)
+            self._files[f] = n
         self._dirty = True
 
     def _load(self):
         if self._loadfunc is not _noop:
             lf, self._loadfunc = self._loadfunc, _noop
             lf(self)
         elif self._copyfunc is not _noop:
             cf, self._copyfunc = self._copyfunc, _noop
             cf(self)
 
     def setflag(self, f, flags):
         """Set the flags (symlink, executable) for path f."""
+        if flags not in _manifestflags:
+            raise TypeError(b"Invalid manifest flag set.")
         self._load()
         dir, subpath = _splittopdir(f)
         if dir:
             self._loadlazy(dir)
             if dir not in self._dirs:
                 self._dirs[dir] = treemanifest(self._subpath(dir))
             self._dirs[dir].setflag(subpath, flags)
         else:
             self._flags[f] = flags
         self._dirty = True
 
     def copy(self):
         copy = treemanifest(self._dir)
         copy._node = self._node
         copy._dirty = self._dirty
         if self._copyfunc is _noop:
 
             def _copyfunc(s):
                 self._load()
                 s._lazydirs = {
                     d: (p, n, r, True)
                     for d, (p, n, r, c) in pycompat.iteritems(self._lazydirs)
                 }
                 sdirs = s._dirs
                 for d, v in pycompat.iteritems(self._dirs):
                     sdirs[d] = v.copy()
                 s._files = dict.copy(self._files)
                 s._flags = dict.copy(self._flags)
 
             if self._loadfunc is _noop:
                 _copyfunc(copy)
             else:
                 copy._copyfunc = _copyfunc
         else:
             copy._copyfunc = self._copyfunc
         return copy
 
     def filesnotin(self, m2, match=None):
         '''Set of files in this manifest that are not in the other'''
         if match and not match.always():
             m1 = self._matches(match)
             m2 = m2._matches(match)
             return m1.filesnotin(m2)
 
         files = set()
 
         def _filesnotin(t1, t2):
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             self._loaddifflazy(t1, t2)
             for d, m1 in pycompat.iteritems(t1._dirs):
                 if d in t2._dirs:
                     m2 = t2._dirs[d]
                     _filesnotin(m1, m2)
                 else:
                     files.update(m1.iterkeys())
 
             for fn in t1._files:
                 if fn not in t2._files:
                     files.add(t1._subpath(fn))
 
         _filesnotin(self, m2)
         return files
 
     @propertycache
     def _alldirs(self):
         return pathutil.dirs(self)
 
     def dirs(self):
         return self._alldirs
 
     def hasdir(self, dir):
         self._load()
         topdir, subdir = _splittopdir(dir)
         if topdir:
             self._loadlazy(topdir)
             if topdir in self._dirs:
                 return self._dirs[topdir].hasdir(subdir)
             return False
         dirslash = dir + b'/'
         return dirslash in self._dirs or dirslash in self._lazydirs
 
     def walk(self, match):
         '''Generates matching file names.
 
         It also reports nonexistent files by marking them bad with match.bad().
         '''
         if match.always():
             for f in iter(self):
                 yield f
             return
 
         fset = set(match.files())
 
         for fn in self._walk(match):
             if fn in fset:
                 # specified pattern is the exact name
                 fset.remove(fn)
             yield fn
 
         # for dirstate.walk, files=[''] means "walk the whole tree".
         # follow that here, too
         fset.discard(b'')
 
         for fn in sorted(fset):
             if not self.hasdir(fn):
                 match.bad(fn, None)
 
     def _walk(self, match):
         '''Recursively generates matching file names for walk().'''
         visit = match.visitchildrenset(self._dir[:-1])
         if not visit:
             return
 
         # yield this dir's files and walk its submanifests
         self._load()
         visit = self._loadchildrensetlazy(visit)
         for p in sorted(list(self._dirs) + list(self._files)):
             if p in self._files:
                 fullp = self._subpath(p)
                 if match(fullp):
                     yield fullp
             else:
                 if not visit or p[:-1] in visit:
                     for f in self._dirs[p]._walk(match):
                         yield f
 
     def _matches(self, match):
         '''recursively generate a new manifest filtered by the match argument.
         '''
         if match.always():
             return self.copy()
         return self._matches_inner(match)
 
     def _matches_inner(self, match):
         if match.always():
             return self.copy()
 
         visit = match.visitchildrenset(self._dir[:-1])
         if visit == b'all':
             return self.copy()
         ret = treemanifest(self._dir)
         if not visit:
             return ret
 
         self._load()
         for fn in self._files:
             # While visitchildrenset *usually* lists only subdirs, this is
             # actually up to the matcher and may have some files in the set().
             # If visit == 'this', we should obviously look at the files in this
             # directory; if visit is a set, and fn is in it, we should inspect
             # fn (but no need to inspect things not in the set).
             if visit != b'this' and fn not in visit:
                 continue
             fullp = self._subpath(fn)
             # visitchildrenset isn't perfect, we still need to call the regular
             # matcher code to further filter results.
             if not match(fullp):
                 continue
             ret._files[fn] = self._files[fn]
             if fn in self._flags:
                 ret._flags[fn] = self._flags[fn]
 
         visit = self._loadchildrensetlazy(visit)
         for dir, subm in pycompat.iteritems(self._dirs):
             if visit and dir[:-1] not in visit:
                 continue
             m = subm._matches_inner(match)
             if not m._isempty():
                 ret._dirs[dir] = m
 
         if not ret._isempty():
             ret._dirty = True
         return ret
 
     def fastdelta(self, base, changes):
         raise FastdeltaUnavailable()
 
     def diff(self, m2, match=None, clean=False):
         '''Finds changes between the current manifest and m2.
 
         Args:
           m2: the manifest to which this manifest should be compared.
           clean: if true, include files unchanged between these manifests
                  with a None value in the returned dictionary.
 
         The result is returned as a dict with filename as key and
         values of the form ((n1,fl1),(n2,fl2)), where n1/n2 is the
         nodeid in the current/other manifest and fl1/fl2 is the flag
         in the current/other manifest. Where the file does not exist,
         the nodeid will be None and the flags will be the empty
         string.
         '''
         if match and not match.always():
             m1 = self._matches(match)
             m2 = m2._matches(match)
             return m1.diff(m2, clean=clean)
         result = {}
         emptytree = treemanifest()
 
         def _iterativediff(t1, t2, stack):
             """compares two tree manifests and append new tree-manifests which
             needs to be compared to stack"""
             if t1._node == t2._node and not t1._dirty and not t2._dirty:
                 return
             t1._load()
             t2._load()
             self._loaddifflazy(t1, t2)
 
             for d, m1 in pycompat.iteritems(t1._dirs):
                 m2 = t2._dirs.get(d, emptytree)
                 stack.append((m1, m2))
 
             for d, m2 in pycompat.iteritems(t2._dirs):
                 if d not in t1._dirs:
                     stack.append((emptytree, m2))
 
             for fn, n1 in pycompat.iteritems(t1._files):
                 fl1 = t1._flags.get(fn, b'')
                 n2 = t2._files.get(fn, None)
                 fl2 = t2._flags.get(fn, b'')
                 if n1 != n2 or fl1 != fl2:
                     result[t1._subpath(fn)] = ((n1, fl1), (n2, fl2))
                 elif clean:
                     result[t1._subpath(fn)] = None
 
             for fn, n2 in pycompat.iteritems(t2._files):
                 if fn not in t1._files:
                     fl2 = t2._flags.get(fn, b'')
                     result[t2._subpath(fn)] = ((None, b''), (n2, fl2))
 
         stackls = []
         _iterativediff(self, m2, stackls)
         while stackls:
             t1, t2 = stackls.pop()
             # stackls is populated in the function call
             _iterativediff(t1, t2, stackls)
         return result
 
     def unmodifiedsince(self, m2):
         return not self._dirty and not m2._dirty and self._node == m2._node
 
     def parse(self, text, readsubtree):
         selflazy = self._lazydirs
         subpath = self._subpath
         for f, n, fl in _parse(text):
             if fl == b't':
                 f = f + b'/'
                 # False below means "doesn't need to be copied" and can use the
                 # cached value from readsubtree directly.
                 selflazy[f] = (subpath(f), n, readsubtree, False)
             elif b'/' in f:
                 # This is a flat manifest, so use __setitem__ and setflag rather
                 # than assigning directly to _files and _flags, so we can
                 # assign a path in a subdirectory, and to mark dirty (compared
                 # to nullid).
                 self[f] = n
                 if fl:
                     self.setflag(f, fl)
             else:
                 # Assigning to _files and _flags avoids marking as dirty,
                 # and should be a little faster.
                 self._files[f] = n
                 if fl:
                     self._flags[f] = fl
 
     def text(self):
         """Get the full data of this manifest as a bytestring."""
         self._load()
         return _text(self.iterentries())
 
     def dirtext(self):
         """Get the full data of this directory as a bytestring. Make sure that
         any submanifests have been written first, so their nodeids are correct.
         """
         self._load()
         flags = self.flags
         lazydirs = [
             (d[:-1], v[1], b't') for d, v in pycompat.iteritems(self._lazydirs)
         ]
         dirs = [(d[:-1], self._dirs[d]._node, b't') for d in self._dirs]
         files = [(f, self._files[f], flags(f)) for f in self._files]
         return _text(sorted(dirs + files + lazydirs))
 
     def read(self, gettext, readsubtree):
         def _load_for_read(s):
             s.parse(gettext(), readsubtree)
             s._dirty = False
 
         self._loadfunc = _load_for_read
 
     def writesubtrees(self, m1, m2, writesubtree, match):
         self._load()  # for consistency; should never have any effect here
         m1._load()
         m2._load()
         emptytree = treemanifest()
 
         def getnode(m, d):
             ld = m._lazydirs.get(d)
             if ld:
                 return ld[1]
             return m._dirs.get(d, emptytree)._node
 
         # let's skip investigating things that `match` says we do not need.
         visit = match.visitchildrenset(self._dir[:-1])
         visit = self._loadchildrensetlazy(visit)
         if visit == b'this' or visit == b'all':
             visit = None
         for d, subm in pycompat.iteritems(self._dirs):
             if visit and d[:-1] not in visit:
                 continue
             subp1 = getnode(m1, d)
             subp2 = getnode(m2, d)
             if subp1 == nullid:
                 subp1, subp2 = subp2, subp1
             writesubtree(subm, subp1, subp2, match)
 
     def walksubtrees(self, matcher=None):
         """Returns an iterator of the subtrees of this manifest, including this
         manifest itself.
 
         If `matcher` is provided, it only returns subtrees that match.
         """
         if matcher and not matcher.visitdir(self._dir[:-1]):
             return
         if not matcher or matcher(self._dir[:-1]):
             yield self
 
         self._load()
         # OPT: use visitchildrenset to avoid loading everything.
         self._loadalllazy()
         for d, subm in pycompat.iteritems(self._dirs):
             for subtree in subm.walksubtrees(matcher=matcher):
                 yield subtree
 
 
 class manifestfulltextcache(util.lrucachedict):
     """File-backed LRU cache for the manifest cache
 
     File consists of entries, up to EOF:
 
     - 20 bytes node, 4 bytes length, <length> manifest data
 
     These are written in reverse cache order (oldest to newest).
 
     """
 
     _file = b'manifestfulltextcache'
 
     def __init__(self, max):
         super(manifestfulltextcache, self).__init__(max)
         self._dirty = False
         self._read = False
         self._opener = None
 
     def read(self):
         if self._read or self._opener is None:
             return
 
         try:
             with self._opener(self._file) as fp:
                 set = super(manifestfulltextcache, self).__setitem__
                 # ignore trailing data, this is a cache, corruption is skipped
                 while True:
                     # TODO do we need to do work here for sha1 portability?
                     node = fp.read(20)
                     if len(node) < 20:
                         break
                     try:
                         size = struct.unpack(b'>L', fp.read(4))[0]
                     except struct.error:
                         break
                     value = bytearray(fp.read(size))
                     if len(value) != size:
                         break
                     set(node, value)
         except IOError:
             # the file is allowed to be missing
             pass
 
         self._read = True
         self._dirty = False
 
     def write(self):
         if not self._dirty or self._opener is None:
             return
         # rotate backwards to the first used node
         try:
             with self._opener(
                 self._file, b'w', atomictemp=True, checkambig=True
             ) as fp:
                 node = self._head.prev
                 while True:
                     if node.key in self._cache:
                         fp.write(node.key)
                         fp.write(struct.pack(b'>L', len(node.value)))
                         fp.write(node.value)
                     if node is self._head:
                         break
                     node = node.prev
         except IOError:
             # We could not write the cache (eg: permission error)
             # the content can be missing.
             #
             # We could try harder and see if we could recreate a wcache
             # directory were we coudl write too.
             #
             # XXX the error pass silently, having some way to issue an error
             # log `ui.log` would be nice.
             pass
 
     def __len__(self):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__len__()
 
     def __contains__(self, k):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__contains__(k)
 
     def __iter__(self):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).__iter__()
 
     def __getitem__(self, k):
         if not self._read:
             self.read()
         # the cache lru order can change on read
         setdirty = self._cache.get(k) is not self._head
         value = super(manifestfulltextcache, self).__getitem__(k)
         if setdirty:
             self._dirty = True
         return value
 
     def __setitem__(self, k, v):
         if not self._read:
             self.read()
         super(manifestfulltextcache, self).__setitem__(k, v)
         self._dirty = True
 
     def __delitem__(self, k):
         if not self._read:
             self.read()
         super(manifestfulltextcache, self).__delitem__(k)
         self._dirty = True
 
     def get(self, k, default=None):
         if not self._read:
             self.read()
         return super(manifestfulltextcache, self).get(k, default=default)
 
     def clear(self, clear_persisted_data=False):
         super(manifestfulltextcache, self).clear()
         if clear_persisted_data:
             self._dirty = True
             self.write()
         self._read = False
 
 
 # and upper bound of what we expect from compression
 # (real live value seems to be "3")
 MAXCOMPRESSION = 3
 
 
 class FastdeltaUnavailable(Exception):
     """Exception raised when fastdelta isn't usable on a manifest."""
 
 
 @interfaceutil.implementer(repository.imanifeststorage)
 class manifestrevlog(object):
     '''A revlog that stores manifest texts. This is responsible for caching the
     full-text manifest contents.
     '''
 
     def __init__(
         self,
         opener,
         tree=b'',
         dirlogcache=None,
         indexfile=None,
         treemanifest=False,
     ):
         """Constructs a new manifest revlog
 
         `indexfile` - used by extensions to have two manifests at once, like
         when transitioning between flatmanifeset and treemanifests.
 
         `treemanifest` - used to indicate this is a tree manifest revlog. Opener
         options can also be used to make this a tree manifest revlog. The opener
         option takes precedence, so if it is set to True, we ignore whatever
         value is passed in to the constructor.
         """
         # During normal operations, we expect to deal with not more than four
         # revs at a time (such as during commit --amend). When rebasing large
         # stacks of commits, the number can go up, hence the config knob below.
         cachesize = 4
         optiontreemanifest = False
         opts = getattr(opener, 'options', None)
         if opts is not None:
             cachesize = opts.get(b'manifestcachesize', cachesize)
             optiontreemanifest = opts.get(b'treemanifest', False)
 
         self._treeondisk = optiontreemanifest or treemanifest
 
         self._fulltextcache = manifestfulltextcache(cachesize)
 
         if tree:
             assert self._treeondisk, b'opts is %r' % opts
 
         if indexfile is None:
             indexfile = b'00manifest.i'
             if tree:
                 indexfile = b"meta/" + tree + indexfile
 
         self.tree = tree
 
         # The dirlogcache is kept on the root manifest log
         if tree:
             self._dirlogcache = dirlogcache
         else:
             self._dirlogcache = {b'': self}
 
         self._revlog = revlog.revlog(
             opener,
             indexfile,
             # only root indexfile is cached
             checkambig=not bool(tree),
             mmaplargeindex=True,
             upperboundcomp=MAXCOMPRESSION,
             persistentnodemap=opener.options.get(b'persistent-nodemap', False),
         )
 
         self.index = self._revlog.index
         self.version = self._revlog.version
         self._generaldelta = self._revlog._generaldelta
 
     def _setupmanifestcachehooks(self, repo):
         """Persist the manifestfulltextcache on lock release"""
         if not util.safehasattr(repo, b'_wlockref'):
             return
 
         self._fulltextcache._opener = repo.wcachevfs
         if repo._currentlock(repo._wlockref) is None:
             return
 
         reporef = weakref.ref(repo)
         manifestrevlogref = weakref.ref(self)
 
         def persistmanifestcache(success):
             # Repo is in an unknown state, do not persist.
             if not success:
                 return
 
             repo = reporef()
             self = manifestrevlogref()
             if repo is None or self is None:
                 return
             if repo.manifestlog.getstorage(b'') is not self:
                 # there's a different manifest in play now, abort
                 return
             self._fulltextcache.write()
 
         repo._afterlock(persistmanifestcache)
 
     @property
     def fulltextcache(self):
         return self._fulltextcache
 
     def clearcaches(self, clear_persisted_data=False):
         self._revlog.clearcaches()
         self._fulltextcache.clear(clear_persisted_data=clear_persisted_data)
         self._dirlogcache = {self.tree: self}
 
     def dirlog(self, d):
         if d:
             assert self._treeondisk
         if d not in self._dirlogcache:
             mfrevlog = manifestrevlog(
                 self.opener, d, self._dirlogcache, treemanifest=self._treeondisk
             )
             self._dirlogcache[d] = mfrevlog
         return self._dirlogcache[d]
 
     def add(
         self,
         m,
         transaction,
         link,
         p1,
         p2,
         added,
         removed,
         readtree=None,
         match=None,
     ):
         """add some manifest entry in to the manifest log
 
         input:
 
           m:           the manifest dict we want to store
           transaction: the open transaction
           p1:          manifest-node of p1
           p2:          manifest-node of p2
           added:       file added/changed compared to parent
           removed:     file removed compared to parent
 
         tree manifest input:
 
           readtree:    a function to read a subtree
           match:       a filematcher for the subpart of the tree manifest
         """
         try:
             if p1 not in self.fulltextcache:
                 raise FastdeltaUnavailable()
             # If our first parent is in the manifest cache, we can
             # compute a delta here using properties we know about the
             # manifest up-front, which may save time later for the
             # revlog layer.
 
             _checkforbidden(added)
             # combine the changed lists into one sorted iterator
             work = heapq.merge(
                 [(x, False) for x in sorted(added)],
                 [(x, True) for x in sorted(removed)],
             )
 
             arraytext, deltatext = m.fastdelta(self.fulltextcache[p1], work)
             cachedelta = self._revlog.rev(p1), deltatext
             text = util.buffer(arraytext)
             n = self._revlog.addrevision(
                 text, transaction, link, p1, p2, cachedelta
             )
         except FastdeltaUnavailable:
             # The first parent manifest isn't already loaded or the
             # manifest implementation doesn't support fastdelta, so
             # we'll just encode a fulltext of the manifest and pass
             # that through to the revlog layer, and let it handle the
             # delta process.
             if self._treeondisk:
                 assert readtree, b"readtree must be set for treemanifest writes"
                 assert match, b"match must be specified for treemanifest writes"
                 m1 = readtree(self.tree, p1)
                 m2 = readtree(self.tree, p2)
                 n = self._addtree(
                     m, transaction, link, m1, m2, readtree, match=match
                 )
                 arraytext = None
             else:
                 text = m.text()
                 n = self._revlog.addrevision(text, transaction, link, p1, p2)
                 arraytext = bytearray(text)
 
         if arraytext is not None:
             self.fulltextcache[n] = arraytext
 
         return n
 
     def _addtree(self, m, transaction, link, m1, m2, readtree, match):
         # If the manifest is unchanged compared to one parent,
         # don't write a new revision
         if self.tree != b'' and (
             m.unmodifiedsince(m1) or m.unmodifiedsince(m2)
         ):
             return m.node()
 
         def writesubtree(subm, subp1, subp2, match):
             sublog = self.dirlog(subm.dir())
             sublog.add(
                 subm,
                 transaction,
                 link,
                 subp1,
                 subp2,
                 None,
                 None,
                 readtree=readtree,
                 match=match,
             )
 
         m.writesubtrees(m1, m2, writesubtree, match)
         text = m.dirtext()
         n = None
         if self.tree != b'':
             # Double-check whether contents are unchanged to one parent
             if text == m1.dirtext():
                 n = m1.node()
             elif text == m2.dirtext():
                 n = m2.node()
 
         if not n:
             n = self._revlog.addrevision(
                 text, transaction, link, m1.node(), m2.node()
             )
 
         # Save nodeid so parent manifest can calculate its nodeid
         m.setnode(n)
         return n
 
     def __len__(self):
         return len(self._revlog)
 
     def __iter__(self):
         return self._revlog.__iter__()
 
     def rev(self, node):
         return self._revlog.rev(node)
 
     def node(self, rev):
         return self._revlog.node(rev)
 
     def lookup(self, value):
         return self._revlog.lookup(value)
 
     def parentrevs(self, rev):
         return self._revlog.parentrevs(rev)
 
     def parents(self, node):
         return self._revlog.parents(node)
 
     def linkrev(self, rev):
         return self._revlog.linkrev(rev)
 
     def checksize(self):
         return self._revlog.checksize()
 
     def revision(self, node, _df=None, raw=False):
         return self._revlog.revision(node, _df=_df, raw=raw)
 
     def rawdata(self, node, _df=None):
         return self._revlog.rawdata(node, _df=_df)
 
     def revdiff(self, rev1, rev2):
         return self._revlog.revdiff(rev1, rev2)
 
     def cmp(self, node, text):
         return self._revlog.cmp(node, text)
 
     def deltaparent(self, rev):
         return self._revlog.deltaparent(rev)
 
     def emitrevisions(
         self,
         nodes,
         nodesorder=None,
         revisiondata=False,
         assumehaveparentrevisions=False,
         deltamode=repository.CG_DELTAMODE_STD,
     ):
         return self._revlog.emitrevisions(
             nodes,
             nodesorder=nodesorder,
             revisiondata=revisiondata,
             assumehaveparentrevisions=assumehaveparentrevisions,
             deltamode=deltamode,
         )
 
     def addgroup(self, deltas, linkmapper, transaction, addrevisioncb=None):
         return self._revlog.addgroup(
             deltas, linkmapper, transaction, addrevisioncb=addrevisioncb
         )
 
     def rawsize(self, rev):
         return self._revlog.rawsize(rev)
 
     def getstrippoint(self, minlink):
         return self._revlog.getstrippoint(minlink)
 
     def strip(self, minlink, transaction):
         return self._revlog.strip(minlink, transaction)
 
     def files(self):
         return self._revlog.files()
 
     def clone(self, tr, destrevlog, **kwargs):
         if not isinstance(destrevlog, manifestrevlog):
             raise error.ProgrammingError(b'expected manifestrevlog to clone()')
 
         return self._revlog.clone(tr, destrevlog._revlog, **kwargs)
 
     def storageinfo(
         self,
         exclusivefiles=False,
         sharedfiles=False,
         revisionscount=False,
         trackedsize=False,
         storedsize=False,
     ):
         return self._revlog.storageinfo(
             exclusivefiles=exclusivefiles,
             sharedfiles=sharedfiles,
             revisionscount=revisionscount,
             trackedsize=trackedsize,
             storedsize=storedsize,
         )
 
     @property
     def indexfile(self):
         return self._revlog.indexfile
 
     @indexfile.setter
     def indexfile(self, value):
         self._revlog.indexfile = value
 
     @property
     def opener(self):
         return self._revlog.opener
 
     @opener.setter
     def opener(self, value):
         self._revlog.opener = value
 
 
 @interfaceutil.implementer(repository.imanifestlog)
 class manifestlog(object):
     """A collection class representing the collection of manifest snapshots
     referenced by commits in the repository.
 
     In this situation, 'manifest' refers to the abstract concept of a snapshot
     of the list of files in the given commit. Consumers of the output of this
     class do not care about the implementation details of the actual manifests
     they receive (i.e. tree or flat or lazily loaded, etc)."""
 
     def __init__(self, opener, repo, rootstore, narrowmatch):
         usetreemanifest = False
         cachesize = 4
 
         opts = getattr(opener, 'options', None)
         if opts is not None:
             usetreemanifest = opts.get(b'treemanifest', usetreemanifest)
             cachesize = opts.get(b'manifestcachesize', cachesize)
 
         self._treemanifests = usetreemanifest
 
         self._rootstore = rootstore
         self._rootstore._setupmanifestcachehooks(repo)
         self._narrowmatch = narrowmatch
 
         # A cache of the manifestctx or treemanifestctx for each directory
         self._dirmancache = {}
         self._dirmancache[b''] = util.lrucachedict(cachesize)
 
         self._cachesize = cachesize
 
     def __getitem__(self, node):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
         """
         return self.get(b'', node)
 
     def get(self, tree, node, verify=True):
         """Retrieves the manifest instance for the given node. Throws a
         LookupError if not found.
 
         `verify` - if True an exception will be thrown if the node is not in
                    the revlog
         """
         if node in self._dirmancache.get(tree, ()):
             return self._dirmancache[tree][node]
 
         if not self._narrowmatch.always():
             if not self._narrowmatch.visitdir(tree[:-1]):
                 return excludeddirmanifestctx(tree, node)
         if tree:
             if self._rootstore._treeondisk:
                 if verify:
                     # Side-effect is LookupError is raised if node doesn't
                     # exist.
                     self.getstorage(tree).rev(node)
 
                 m = treemanifestctx(self, tree, node)
             else:
                 raise error.Abort(
                     _(
                         b"cannot ask for manifest directory '%s' in a flat "
                         b"manifest"
                     )
                     % tree
                 )
         else:
             if verify:
                 # Side-effect is LookupError is raised if node doesn't exist.
                 self._rootstore.rev(node)
 
             if self._treemanifests:
                 m = treemanifestctx(self, b'', node)
             else:
                 m = manifestctx(self, node)
 
         if node != nullid:
             mancache = self._dirmancache.get(tree)
             if not mancache:
                 mancache = util.lrucachedict(self._cachesize)
                 self._dirmancache[tree] = mancache
             mancache[node] = m
         return m
 
     def getstorage(self, tree):
         return self._rootstore.dirlog(tree)
 
     def clearcaches(self, clear_persisted_data=False):
         self._dirmancache.clear()
         self._rootstore.clearcaches(clear_persisted_data=clear_persisted_data)
 
     def rev(self, node):
         return self._rootstore.rev(node)
 
     def update_caches(self, transaction):
         return self._rootstore._revlog.update_caches(transaction=transaction)
 
 
 @interfaceutil.implementer(repository.imanifestrevisionwritable)
 class memmanifestctx(object):
     def __init__(self, manifestlog):
         self._manifestlog = manifestlog
         self._manifestdict = manifestdict()
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def copy(self):
         memmf = memmanifestctx(self._manifestlog)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     def read(self):
         return self._manifestdict
 
     def write(self, transaction, link, p1, p2, added, removed, match=None):
         return self._storage().add(
             self._manifestdict,
             transaction,
             link,
             p1,
             p2,
             added,
             removed,
             match=match,
         )
 
 
 @interfaceutil.implementer(repository.imanifestrevisionstored)
 class manifestctx(object):
     """A class representing a single revision of a manifest, including its
     contents, its parent revs, and its linkrev.
     """
 
     def __init__(self, manifestlog, node):
         self._manifestlog = manifestlog
         self._data = None
 
         self._node = node
 
         # TODO: We eventually want p1, p2, and linkrev exposed on this class,
         # but let's add it later when something needs it and we can load it
         # lazily.
         # self.p1, self.p2 = store.parents(node)
         # rev = store.rev(node)
         # self.linkrev = store.linkrev(rev)
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def node(self):
         return self._node
 
     def copy(self):
         memmf = memmanifestctx(self._manifestlog)
         memmf._manifestdict = self.read().copy()
         return memmf
 
     @propertycache
     def parents(self):
         return self._storage().parents(self._node)
 
     def read(self):
         if self._data is None:
             if self._node == nullid:
                 self._data = manifestdict()
             else:
                 store = self._storage()
                 if self._node in store.fulltextcache:
                     text = pycompat.bytestr(store.fulltextcache[self._node])
                 else:
                     text = store.revision(self._node)
                     arraytext = bytearray(text)
                     store.fulltextcache[self._node] = arraytext
                 self._data = manifestdict(text)
         return self._data
 
     def readfast(self, shallow=False):
         '''Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, nothing changes since this is a flat manifest.
         '''
         store = self._storage()
         r = store.rev(self._node)
         deltaparent = store.deltaparent(r)
         if deltaparent != nullrev and deltaparent in store.parentrevs(r):
             return self.readdelta()
         return self.read()
 
     def readdelta(self, shallow=False):
         '''Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         Changing the value of `shallow` has no effect on flat manifests.
         '''
         store = self._storage()
         r = store.rev(self._node)
         d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r))
         return manifestdict(d)
 
     def find(self, key):
         return self.read().find(key)
 
 
 @interfaceutil.implementer(repository.imanifestrevisionwritable)
 class memtreemanifestctx(object):
     def __init__(self, manifestlog, dir=b''):
         self._manifestlog = manifestlog
         self._dir = dir
         self._treemanifest = treemanifest()
 
     def _storage(self):
         return self._manifestlog.getstorage(b'')
 
     def copy(self):
         memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
         memmf._treemanifest = self._treemanifest.copy()
         return memmf
 
     def read(self):
         return self._treemanifest
 
     def write(self, transaction, link, p1, p2, added, removed, match=None):
         def readtree(dir, node):
             return self._manifestlog.get(dir, node).read()
 
         return self._storage().add(
             self._treemanifest,
             transaction,
             link,
             p1,
             p2,
             added,
             removed,
             readtree=readtree,
             match=match,
         )
 
 
 @interfaceutil.implementer(repository.imanifestrevisionstored)
 class treemanifestctx(object):
     def __init__(self, manifestlog, dir, node):
         self._manifestlog = manifestlog
         self._dir = dir
         self._data = None
 
         self._node = node
 
         # TODO: Load p1/p2/linkrev lazily. They need to be lazily loaded so that
         # we can instantiate treemanifestctx objects for directories we don't
         # have on disk.
         # self.p1, self.p2 = store.parents(node)
         # rev = store.rev(node)
         # self.linkrev = store.linkrev(rev)
 
     def _storage(self):
         narrowmatch = self._manifestlog._narrowmatch
         if not narrowmatch.always():
             if not narrowmatch.visitdir(self._dir[:-1]):
                 return excludedmanifestrevlog(self._dir)
         return self._manifestlog.getstorage(self._dir)
 
     def read(self):
         if self._data is None:
             store = self._storage()
             if self._node == nullid:
                 self._data = treemanifest()
             # TODO accessing non-public API
             elif store._treeondisk:
                 m = treemanifest(dir=self._dir)
 
                 def gettext():
                     return store.revision(self._node)
 
                 def readsubtree(dir, subm):
                     # Set verify to False since we need to be able to create
                     # subtrees for trees that don't exist on disk.
                     return self._manifestlog.get(dir, subm, verify=False).read()
 
                 m.read(gettext, readsubtree)
                 m.setnode(self._node)
                 self._data = m
             else:
                 if self._node in store.fulltextcache:
                     text = pycompat.bytestr(store.fulltextcache[self._node])
                 else:
                     text = store.revision(self._node)
                     arraytext = bytearray(text)
                     store.fulltextcache[self._node] = arraytext
                 self._data = treemanifest(dir=self._dir, text=text)
 
         return self._data
 
     def node(self):
         return self._node
 
     def copy(self):
         memmf = memtreemanifestctx(self._manifestlog, dir=self._dir)
         memmf._treemanifest = self.read().copy()
         return memmf
 
     @propertycache
     def parents(self):
         return self._storage().parents(self._node)
 
     def readdelta(self, shallow=False):
         '''Returns a manifest containing just the entries that are present
         in this manifest, but not in its p1 manifest. This is efficient to read
         if the revlog delta is already p1.
 
         If `shallow` is True, this will read the delta for this directory,
         without recursively reading subdirectory manifests. Instead, any
         subdirectory entry will be reported as it appears in the manifest, i.e.
         the subdirectory will be reported among files and distinguished only by
         its 't' flag.
         '''
         store = self._storage()
         if shallow:
             r = store.rev(self._node)
             d = mdiff.patchtext(store.revdiff(store.deltaparent(r), r))
             return manifestdict(d)
         else:
             # Need to perform a slow delta
             r0 = store.deltaparent(store.rev(self._node))
             m0 = self._manifestlog.get(self._dir, store.node(r0)).read()
             m1 = self.read()
             md = treemanifest(dir=self._dir)
             for f, ((n0, fl0), (n1, fl1)) in pycompat.iteritems(m0.diff(m1)):
                 if n1:
                     md[f] = n1
                     if fl1:
                         md.setflag(f, fl1)
             return md
 
     def readfast(self, shallow=False):
         '''Calls either readdelta or read, based on which would be less work.
         readdelta is called if the delta is against the p1, and therefore can be
         read quickly.
 
         If `shallow` is True, it only returns the entries from this manifest,
         and not any submanifests.
         '''
         store = self._storage()
         r = store.rev(self._node)
         deltaparent = store.deltaparent(r)
         if deltaparent != nullrev and deltaparent in store.parentrevs(r):
             return self.readdelta(shallow=shallow)
 
         if shallow:
             return manifestdict(store.revision(self._node))
         else:
             return self.read()
 
     def find(self, key):
         return self.read().find(key)
 
 
 class excludeddir(treemanifest):
     """Stand-in for a directory that is excluded from the repository.
 
     With narrowing active on a repository that uses treemanifests,
     some of the directory revlogs will be excluded from the resulting
     clone. This is a huge storage win for clients, but means we need
     some sort of pseudo-manifest to surface to internals so we can
     detect a merge conflict outside the narrowspec. That's what this
     class is: it stands in for a directory whose node is known, but
     whose contents are unknown.
     """
 
     def __init__(self, dir, node):
         super(excludeddir, self).__init__(dir)
         self._node = node
         # Add an empty file, which will be included by iterators and such,
         # appearing as the directory itself (i.e. something like "dir/")
         self._files[b''] = node
         self._flags[b''] = b't'
 
     # Manifests outside the narrowspec should never be modified, so avoid
     # copying. This makes a noticeable difference when there are very many
     # directories outside the narrowspec. Also, it makes sense for the copy to
     # be of the same type as the original, which would not happen with the
     # super type's copy().
     def copy(self):
         return self
 
 
 class excludeddirmanifestctx(treemanifestctx):
     """context wrapper for excludeddir - see that docstring for rationale"""
 
     def __init__(self, dir, node):
         self._dir = dir
         self._node = node
 
     def read(self):
         return excludeddir(self._dir, self._node)
 
     def write(self, *args):
         raise error.ProgrammingError(
             b'attempt to write manifest from excluded dir %s' % self._dir
         )
 
 
 class excludedmanifestrevlog(manifestrevlog):
     """Stand-in for excluded treemanifest revlogs.
 
     When narrowing is active on a treemanifest repository, we'll have
     references to directories we can't see due to the revlog being
     skipped. This class exists to conform to the manifestrevlog
     interface for those directories and proactively prevent writes to
     outside the narrowspec.
     """
 
     def __init__(self, dir):
         self._dir = dir
 
     def __len__(self):
         raise error.ProgrammingError(
             b'attempt to get length of excluded dir %s' % self._dir
         )
 
     def rev(self, node):
         raise error.ProgrammingError(
             b'attempt to get rev from excluded dir %s' % self._dir
         )
 
     def linkrev(self, node):
         raise error.ProgrammingError(
             b'attempt to get linkrev from excluded dir %s' % self._dir
         )
 
     def node(self, rev):
         raise error.ProgrammingError(
             b'attempt to get node from excluded dir %s' % self._dir
         )
 
     def add(self, *args, **kwargs):
         # We should never write entries in dirlogs outside the narrow clone.
         # However, the method still gets called from writesubtree() in
         # _addtree(), so we need to handle it. We should possibly make that
         # avoid calling add() with a clean manifest (_dirty is always False
         # in excludeddir instances).
         pass

mercurial/merge.py

@@ -1,2083 +1,2082 b''
 # merge.py - directory-level update/merge handling for Mercurial
 #
 # Copyright 2006, 2007 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import errno
 import stat
 import struct
 
 from .i18n import _
 from .node import (
     addednodeid,
     modifiednodeid,
     nullid,
     nullrev,
 )
 from .thirdparty import attr
 from . import (
     copies,
     encoding,
     error,
     filemerge,
     match as matchmod,
     mergestate as mergestatemod,
     obsutil,
     pathutil,
     pycompat,
     scmutil,
     subrepoutil,
     util,
     worker,
 )
 
 _pack = struct.pack
 _unpack = struct.unpack
 
 
 def _getcheckunknownconfig(repo, section, name):
     config = repo.ui.config(section, name)
     valid = [b'abort', b'ignore', b'warn']
     if config not in valid:
         validstr = b', '.join([b"'" + v + b"'" for v in valid])
         raise error.ConfigError(
             _(b"%s.%s not valid ('%s' is none of %s)")
             % (section, name, config, validstr)
         )
     return config
 
 
 def _checkunknownfile(repo, wctx, mctx, f, f2=None):
     if wctx.isinmemory():
         # Nothing to do in IMM because nothing in the "working copy" can be an
         # unknown file.
         #
         # Note that we should bail out here, not in ``_checkunknownfiles()``,
         # because that function does other useful work.
         return False
 
     if f2 is None:
         f2 = f
     return (
         repo.wvfs.audit.check(f)
         and repo.wvfs.isfileorlink(f)
         and repo.dirstate.normalize(f) not in repo.dirstate
         and mctx[f2].cmp(wctx[f])
     )
 
 
 class _unknowndirschecker(object):
     """
     Look for any unknown files or directories that may have a path conflict
     with a file.  If any path prefix of the file exists as a file or link,
     then it conflicts.  If the file itself is a directory that contains any
     file that is not tracked, then it conflicts.
 
     Returns the shortest path at which a conflict occurs, or None if there is
     no conflict.
     """
 
     def __init__(self):
         # A set of paths known to be good.  This prevents repeated checking of
         # dirs.  It will be updated with any new dirs that are checked and found
         # to be safe.
         self._unknowndircache = set()
 
         # A set of paths that are known to be absent.  This prevents repeated
         # checking of subdirectories that are known not to exist. It will be
         # updated with any new dirs that are checked and found to be absent.
         self._missingdircache = set()
 
     def __call__(self, repo, wctx, f):
         if wctx.isinmemory():
             # Nothing to do in IMM for the same reason as ``_checkunknownfile``.
             return False
 
         # Check for path prefixes that exist as unknown files.
         for p in reversed(list(pathutil.finddirs(f))):
             if p in self._missingdircache:
                 return
             if p in self._unknowndircache:
                 continue
             if repo.wvfs.audit.check(p):
                 if (
                     repo.wvfs.isfileorlink(p)
                     and repo.dirstate.normalize(p) not in repo.dirstate
                 ):
                     return p
                 if not repo.wvfs.lexists(p):
                     self._missingdircache.add(p)
                     return
                 self._unknowndircache.add(p)
 
         # Check if the file conflicts with a directory containing unknown files.
         if repo.wvfs.audit.check(f) and repo.wvfs.isdir(f):
             # Does the directory contain any files that are not in the dirstate?
             for p, dirs, files in repo.wvfs.walk(f):
                 for fn in files:
                     relf = util.pconvert(repo.wvfs.reljoin(p, fn))
                     relf = repo.dirstate.normalize(relf, isknown=True)
                     if relf not in repo.dirstate:
                         return f
         return None
 
 
 def _checkunknownfiles(repo, wctx, mctx, force, actions, mergeforce):
     """
     Considers any actions that care about the presence of conflicting unknown
     files. For some actions, the result is to abort; for others, it is to
     choose a different action.
     """
     fileconflicts = set()
     pathconflicts = set()
     warnconflicts = set()
     abortconflicts = set()
     unknownconfig = _getcheckunknownconfig(repo, b'merge', b'checkunknown')
     ignoredconfig = _getcheckunknownconfig(repo, b'merge', b'checkignored')
     pathconfig = repo.ui.configbool(
         b'experimental', b'merge.checkpathconflicts'
     )
     if not force:
 
         def collectconflicts(conflicts, config):
             if config == b'abort':
                 abortconflicts.update(conflicts)
             elif config == b'warn':
                 warnconflicts.update(conflicts)
 
         checkunknowndirs = _unknowndirschecker()
         for f, (m, args, msg) in pycompat.iteritems(actions):
             if m in (
                 mergestatemod.ACTION_CREATED,
                 mergestatemod.ACTION_DELETED_CHANGED,
             ):
                 if _checkunknownfile(repo, wctx, mctx, f):
                     fileconflicts.add(f)
                 elif pathconfig and f not in wctx:
                     path = checkunknowndirs(repo, wctx, f)
                     if path is not None:
                         pathconflicts.add(path)
             elif m == mergestatemod.ACTION_LOCAL_DIR_RENAME_GET:
                 if _checkunknownfile(repo, wctx, mctx, f, args[0]):
                     fileconflicts.add(f)
 
         allconflicts = fileconflicts | pathconflicts
         ignoredconflicts = {c for c in allconflicts if repo.dirstate._ignore(c)}
         unknownconflicts = allconflicts - ignoredconflicts
         collectconflicts(ignoredconflicts, ignoredconfig)
         collectconflicts(unknownconflicts, unknownconfig)
     else:
         for f, (m, args, msg) in pycompat.iteritems(actions):
             if m == mergestatemod.ACTION_CREATED_MERGE:
                 fl2, anc = args
                 different = _checkunknownfile(repo, wctx, mctx, f)
                 if repo.dirstate._ignore(f):
                     config = ignoredconfig
                 else:
                     config = unknownconfig
 
                 # The behavior when force is True is described by this table:
                 #  config  different  mergeforce  |    action    backup
                 #    *         n          *       |      get        n
                 #    *         y          y       |     merge       -
                 #   abort      y          n       |     merge       -   (1)
                 #   warn       y          n       |  warn + get     y
                 #  ignore      y          n       |      get        y
                 #
                 # (1) this is probably the wrong behavior here -- we should
                 #     probably abort, but some actions like rebases currently
                 #     don't like an abort happening in the middle of
                 #     merge.update.
                 if not different:
                     actions[f] = (
                         mergestatemod.ACTION_GET,
                         (fl2, False),
                         b'remote created',
                     )
                 elif mergeforce or config == b'abort':
                     actions[f] = (
                         mergestatemod.ACTION_MERGE,
                         (f, f, None, False, anc),
                         b'remote differs from untracked local',
                     )
                 elif config == b'abort':
                     abortconflicts.add(f)
                 else:
                     if config == b'warn':
                         warnconflicts.add(f)
                     actions[f] = (
                         mergestatemod.ACTION_GET,
                         (fl2, True),
                         b'remote created',
                     )
 
     for f in sorted(abortconflicts):
         warn = repo.ui.warn
         if f in pathconflicts:
             if repo.wvfs.isfileorlink(f):
                 warn(_(b"%s: untracked file conflicts with directory\n") % f)
             else:
                 warn(_(b"%s: untracked directory conflicts with file\n") % f)
         else:
             warn(_(b"%s: untracked file differs\n") % f)
     if abortconflicts:
         raise error.Abort(
             _(
                 b"untracked files in working directory "
                 b"differ from files in requested revision"
             )
         )
 
     for f in sorted(warnconflicts):
         if repo.wvfs.isfileorlink(f):
             repo.ui.warn(_(b"%s: replacing untracked file\n") % f)
         else:
             repo.ui.warn(_(b"%s: replacing untracked files in directory\n") % f)
 
     for f, (m, args, msg) in pycompat.iteritems(actions):
         if m == mergestatemod.ACTION_CREATED:
             backup = (
                 f in fileconflicts
                 or f in pathconflicts
                 or any(p in pathconflicts for p in pathutil.finddirs(f))
             )
             (flags,) = args
             actions[f] = (mergestatemod.ACTION_GET, (flags, backup), msg)
 
 
 def _forgetremoved(wctx, mctx, branchmerge):
     """
     Forget removed files
 
     If we're jumping between revisions (as opposed to merging), and if
     neither the working directory nor the target rev has the file,
     then we need to remove it from the dirstate, to prevent the
     dirstate from listing the file when it is no longer in the
     manifest.
 
     If we're merging, and the other revision has removed a file
     that is not present in the working directory, we need to mark it
     as removed.
     """
 
     actions = {}
     m = mergestatemod.ACTION_FORGET
     if branchmerge:
         m = mergestatemod.ACTION_REMOVE
     for f in wctx.deleted():
         if f not in mctx:
             actions[f] = m, None, b"forget deleted"
 
     if not branchmerge:
         for f in wctx.removed():
             if f not in mctx:
                 actions[f] = (
                     mergestatemod.ACTION_FORGET,
                     None,
                     b"forget removed",
                 )
 
     return actions
 
 
 def _checkcollision(repo, wmf, actions):
     """
     Check for case-folding collisions.
     """
     # If the repo is narrowed, filter out files outside the narrowspec.
     narrowmatch = repo.narrowmatch()
     if not narrowmatch.always():
         pmmf = set(wmf.walk(narrowmatch))
         if actions:
             narrowactions = {}
             for m, actionsfortype in pycompat.iteritems(actions):
                 narrowactions[m] = []
                 for (f, args, msg) in actionsfortype:
                     if narrowmatch(f):
                         narrowactions[m].append((f, args, msg))
             actions = narrowactions
     else:
         # build provisional merged manifest up
         pmmf = set(wmf)
 
     if actions:
         # KEEP and EXEC are no-op
         for m in (
             mergestatemod.ACTION_ADD,
             mergestatemod.ACTION_ADD_MODIFIED,
             mergestatemod.ACTION_FORGET,
             mergestatemod.ACTION_GET,
             mergestatemod.ACTION_CHANGED_DELETED,
             mergestatemod.ACTION_DELETED_CHANGED,
         ):
             for f, args, msg in actions[m]:
                 pmmf.add(f)
         for f, args, msg in actions[mergestatemod.ACTION_REMOVE]:
             pmmf.discard(f)
         for f, args, msg in actions[mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL]:
             f2, flags = args
             pmmf.discard(f2)
             pmmf.add(f)
         for f, args, msg in actions[mergestatemod.ACTION_LOCAL_DIR_RENAME_GET]:
             pmmf.add(f)
         for f, args, msg in actions[mergestatemod.ACTION_MERGE]:
             f1, f2, fa, move, anc = args
             if move:
                 pmmf.discard(f1)
             pmmf.add(f)
 
     # check case-folding collision in provisional merged manifest
     foldmap = {}
     for f in pmmf:
         fold = util.normcase(f)
         if fold in foldmap:
             raise error.Abort(
                 _(b"case-folding collision between %s and %s")
                 % (f, foldmap[fold])
             )
         foldmap[fold] = f
 
     # check case-folding of directories
     foldprefix = unfoldprefix = lastfull = b''
     for fold, f in sorted(foldmap.items()):
         if fold.startswith(foldprefix) and not f.startswith(unfoldprefix):
             # the folded prefix matches but actual casing is different
             raise error.Abort(
                 _(b"case-folding collision between %s and directory of %s")
                 % (lastfull, f)
             )
         foldprefix = fold + b'/'
         unfoldprefix = f + b'/'
         lastfull = f
 
 
 def driverpreprocess(repo, ms, wctx, labels=None):
     """run the preprocess step of the merge driver, if any
 
     This is currently not implemented -- it's an extension point."""
     return True
 
 
 def driverconclude(repo, ms, wctx, labels=None):
     """run the conclude step of the merge driver, if any
 
     This is currently not implemented -- it's an extension point."""
     return True
 
 
 def _filesindirs(repo, manifest, dirs):
     """
     Generator that yields pairs of all the files in the manifest that are found
     inside the directories listed in dirs, and which directory they are found
     in.
     """
     for f in manifest:
         for p in pathutil.finddirs(f):
             if p in dirs:
                 yield f, p
                 break
 
 
 def checkpathconflicts(repo, wctx, mctx, actions):
     """
     Check if any actions introduce path conflicts in the repository, updating
     actions to record or handle the path conflict accordingly.
     """
     mf = wctx.manifest()
 
     # The set of local files that conflict with a remote directory.
     localconflicts = set()
 
     # The set of directories that conflict with a remote file, and so may cause
     # conflicts if they still contain any files after the merge.
     remoteconflicts = set()
 
     # The set of directories that appear as both a file and a directory in the
     # remote manifest.  These indicate an invalid remote manifest, which
     # can't be updated to cleanly.
     invalidconflicts = set()
 
     # The set of directories that contain files that are being created.
     createdfiledirs = set()
 
     # The set of files deleted by all the actions.
     deletedfiles = set()
 
     for f, (m, args, msg) in actions.items():
         if m in (
             mergestatemod.ACTION_CREATED,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_MERGE,
             mergestatemod.ACTION_CREATED_MERGE,
         ):
             # This action may create a new local file.
             createdfiledirs.update(pathutil.finddirs(f))
             if mf.hasdir(f):
                 # The file aliases a local directory.  This might be ok if all
                 # the files in the local directory are being deleted.  This
                 # will be checked once we know what all the deleted files are.
                 remoteconflicts.add(f)
         # Track the names of all deleted files.
         if m == mergestatemod.ACTION_REMOVE:
             deletedfiles.add(f)
         if m == mergestatemod.ACTION_MERGE:
             f1, f2, fa, move, anc = args
             if move:
                 deletedfiles.add(f1)
         if m == mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL:
             f2, flags = args
             deletedfiles.add(f2)
 
     # Check all directories that contain created files for path conflicts.
     for p in createdfiledirs:
         if p in mf:
             if p in mctx:
                 # A file is in a directory which aliases both a local
                 # and a remote file.  This is an internal inconsistency
                 # within the remote manifest.
                 invalidconflicts.add(p)
             else:
                 # A file is in a directory which aliases a local file.
                 # We will need to rename the local file.
                 localconflicts.add(p)
         if p in actions and actions[p][0] in (
             mergestatemod.ACTION_CREATED,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_MERGE,
             mergestatemod.ACTION_CREATED_MERGE,
         ):
             # The file is in a directory which aliases a remote file.
             # This is an internal inconsistency within the remote
             # manifest.
             invalidconflicts.add(p)
 
     # Rename all local conflicting files that have not been deleted.
     for p in localconflicts:
         if p not in deletedfiles:
             ctxname = bytes(wctx).rstrip(b'+')
             pnew = util.safename(p, ctxname, wctx, set(actions.keys()))
             porig = wctx[p].copysource() or p
             actions[pnew] = (
                 mergestatemod.ACTION_PATH_CONFLICT_RESOLVE,
                 (p, porig),
                 b'local path conflict',
             )
             actions[p] = (
                 mergestatemod.ACTION_PATH_CONFLICT,
                 (pnew, b'l'),
                 b'path conflict',
             )
 
     if remoteconflicts:
         # Check if all files in the conflicting directories have been removed.
         ctxname = bytes(mctx).rstrip(b'+')
         for f, p in _filesindirs(repo, mf, remoteconflicts):
             if f not in deletedfiles:
                 m, args, msg = actions[p]
                 pnew = util.safename(p, ctxname, wctx, set(actions.keys()))
                 if m in (
                     mergestatemod.ACTION_DELETED_CHANGED,
                     mergestatemod.ACTION_MERGE,
                 ):
                     # Action was merge, just update target.
                     actions[pnew] = (m, args, msg)
                 else:
                     # Action was create, change to renamed get action.
                     fl = args[0]
                     actions[pnew] = (
                         mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
                         (p, fl),
                         b'remote path conflict',
                     )
                 actions[p] = (
                     mergestatemod.ACTION_PATH_CONFLICT,
                     (pnew, mergestatemod.ACTION_REMOVE),
                     b'path conflict',
                 )
                 remoteconflicts.remove(p)
                 break
 
     if invalidconflicts:
         for p in invalidconflicts:
             repo.ui.warn(_(b"%s: is both a file and a directory\n") % p)
         raise error.Abort(_(b"destination manifest contains path conflicts"))
 
 
 def _filternarrowactions(narrowmatch, branchmerge, actions):
     """
     Filters out actions that can ignored because the repo is narrowed.
 
     Raise an exception if the merge cannot be completed because the repo is
     narrowed.
     """
     nooptypes = {b'k'}  # TODO: handle with nonconflicttypes
     nonconflicttypes = set(b'a am c cm f g gs r e'.split())
     # We mutate the items in the dict during iteration, so iterate
     # over a copy.
     for f, action in list(actions.items()):
         if narrowmatch(f):
             pass
         elif not branchmerge:
             del actions[f]  # just updating, ignore changes outside clone
         elif action[0] in nooptypes:
             del actions[f]  # merge does not affect file
         elif action[0] in nonconflicttypes:
             raise error.Abort(
                 _(
                     b'merge affects file \'%s\' outside narrow, '
                     b'which is not yet supported'
                 )
                 % f,
                 hint=_(b'merging in the other direction may work'),
             )
         else:
             raise error.Abort(
                 _(b'conflict in file \'%s\' is outside narrow clone') % f
             )
 
 
 def manifestmerge(
     repo,
     wctx,
     p2,
     pa,
     branchmerge,
     force,
     matcher,
     acceptremote,
     followcopies,
     forcefulldiff=False,
 ):
     """
     Merge wctx and p2 with ancestor pa and generate merge action list
 
     branchmerge and force are as passed in to update
     matcher = matcher to filter file lists
     acceptremote = accept the incoming changes without prompting
     """
     if matcher is not None and matcher.always():
         matcher = None
 
     # manifests fetched in order are going to be faster, so prime the caches
     [
         x.manifest()
         for x in sorted(wctx.parents() + [p2, pa], key=scmutil.intrev)
     ]
 
     branch_copies1 = copies.branch_copies()
     branch_copies2 = copies.branch_copies()
     diverge = {}
     if followcopies:
         branch_copies1, branch_copies2, diverge = copies.mergecopies(
             repo, wctx, p2, pa
         )
 
     boolbm = pycompat.bytestr(bool(branchmerge))
     boolf = pycompat.bytestr(bool(force))
     boolm = pycompat.bytestr(bool(matcher))
     repo.ui.note(_(b"resolving manifests\n"))
     repo.ui.debug(
         b" branchmerge: %s, force: %s, partial: %s\n" % (boolbm, boolf, boolm)
     )
     repo.ui.debug(b" ancestor: %s, local: %s, remote: %s\n" % (pa, wctx, p2))
 
     m1, m2, ma = wctx.manifest(), p2.manifest(), pa.manifest()
     copied1 = set(branch_copies1.copy.values())
     copied1.update(branch_copies1.movewithdir.values())
     copied2 = set(branch_copies2.copy.values())
     copied2.update(branch_copies2.movewithdir.values())
 
     if b'.hgsubstate' in m1 and wctx.rev() is None:
         # Check whether sub state is modified, and overwrite the manifest
         # to flag the change. If wctx is a committed revision, we shouldn't
         # care for the dirty state of the working directory.
         if any(wctx.sub(s).dirty() for s in wctx.substate):
             m1[b'.hgsubstate'] = modifiednodeid
 
     # Don't use m2-vs-ma optimization if:
     # - ma is the same as m1 or m2, which we're just going to diff again later
     # - The caller specifically asks for a full diff, which is useful during bid
     #   merge.
     if pa not in ([wctx, p2] + wctx.parents()) and not forcefulldiff:
         # Identify which files are relevant to the merge, so we can limit the
         # total m1-vs-m2 diff to just those files. This has significant
         # performance benefits in large repositories.
         relevantfiles = set(ma.diff(m2).keys())
 
         # For copied and moved files, we need to add the source file too.
         for copykey, copyvalue in pycompat.iteritems(branch_copies1.copy):
             if copyvalue in relevantfiles:
                 relevantfiles.add(copykey)
         for movedirkey in branch_copies1.movewithdir:
             relevantfiles.add(movedirkey)
         filesmatcher = scmutil.matchfiles(repo, relevantfiles)
         matcher = matchmod.intersectmatchers(matcher, filesmatcher)
 
     diff = m1.diff(m2, match=matcher)
 
     actions = {}
     for f, ((n1, fl1), (n2, fl2)) in pycompat.iteritems(diff):
         if n1 and n2:  # file exists on both local and remote side
             if f not in ma:
                 # TODO: what if they're renamed from different sources?
                 fa = branch_copies1.copy.get(
                     f, None
                 ) or branch_copies2.copy.get(f, None)
                 if fa is not None:
                     actions[f] = (
                         mergestatemod.ACTION_MERGE,
                         (f, f, fa, False, pa.node()),
                         b'both renamed from %s' % fa,
                     )
                 else:
                     actions[f] = (
                         mergestatemod.ACTION_MERGE,
                         (f, f, None, False, pa.node()),
                         b'both created',
                     )
             else:
                 a = ma[f]
                 fla = ma.flags(f)
                 nol = b'l' not in fl1 + fl2 + fla
                 if n2 == a and fl2 == fla:
                     actions[f] = (
                         mergestatemod.ACTION_KEEP,
                         (),
                         b'remote unchanged',
                     )
                 elif n1 == a and fl1 == fla:  # local unchanged - use remote
                     if n1 == n2:  # optimization: keep local content
                         actions[f] = (
                             mergestatemod.ACTION_EXEC,
                             (fl2,),
                             b'update permissions',
                         )
                     else:
                         actions[f] = (
                             mergestatemod.ACTION_GET_OTHER_AND_STORE
                             if branchmerge
                             else mergestatemod.ACTION_GET,
                             (fl2, False),
                             b'remote is newer',
                         )
                 elif nol and n2 == a:  # remote only changed 'x'
                     actions[f] = (
                         mergestatemod.ACTION_EXEC,
                         (fl2,),
                         b'update permissions',
                     )
                 elif nol and n1 == a:  # local only changed 'x'
                     actions[f] = (
                         mergestatemod.ACTION_GET_OTHER_AND_STORE
                         if branchmerge
                         else mergestatemod.ACTION_GET,
                         (fl1, False),
                         b'remote is newer',
                     )
                 else:  # both changed something
                     actions[f] = (
                         mergestatemod.ACTION_MERGE,
                         (f, f, f, False, pa.node()),
                         b'versions differ',
                     )
         elif n1:  # file exists only on local side
             if f in copied2:
                 pass  # we'll deal with it on m2 side
             elif (
                 f in branch_copies1.movewithdir
             ):  # directory rename, move local
                 f2 = branch_copies1.movewithdir[f]
                 if f2 in m2:
                     actions[f2] = (
                         mergestatemod.ACTION_MERGE,
                         (f, f2, None, True, pa.node()),
                         b'remote directory rename, both created',
                     )
                 else:
                     actions[f2] = (
                         mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL,
                         (f, fl1),
                         b'remote directory rename - move from %s' % f,
                     )
             elif f in branch_copies1.copy:
                 f2 = branch_copies1.copy[f]
                 actions[f] = (
                     mergestatemod.ACTION_MERGE,
                     (f, f2, f2, False, pa.node()),
                     b'local copied/moved from %s' % f2,
                 )
             elif f in ma:  # clean, a different, no remote
                 if n1 != ma[f]:
                     if acceptremote:
                         actions[f] = (
                             mergestatemod.ACTION_REMOVE,
                             None,
                             b'remote delete',
                         )
                     else:
                         actions[f] = (
                             mergestatemod.ACTION_CHANGED_DELETED,
                             (f, None, f, False, pa.node()),
                             b'prompt changed/deleted',
                         )
                 elif n1 == addednodeid:
-                    # This extra 'a' is added by working copy manifest to mark
-                    # the file as locally added. We should forget it instead of
+                    # This file was locally added. We should forget it instead of
                     # deleting it.
                     actions[f] = (
                         mergestatemod.ACTION_FORGET,
                         None,
                         b'remote deleted',
                     )
                 else:
                     actions[f] = (
                         mergestatemod.ACTION_REMOVE,
                         None,
                         b'other deleted',
                     )
         elif n2:  # file exists only on remote side
             if f in copied1:
                 pass  # we'll deal with it on m1 side
             elif f in branch_copies2.movewithdir:
                 f2 = branch_copies2.movewithdir[f]
                 if f2 in m1:
                     actions[f2] = (
                         mergestatemod.ACTION_MERGE,
                         (f2, f, None, False, pa.node()),
                         b'local directory rename, both created',
                     )
                 else:
                     actions[f2] = (
                         mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
                         (f, fl2),
                         b'local directory rename - get from %s' % f,
                     )
             elif f in branch_copies2.copy:
                 f2 = branch_copies2.copy[f]
                 if f2 in m2:
                     actions[f] = (
                         mergestatemod.ACTION_MERGE,
                         (f2, f, f2, False, pa.node()),
                         b'remote copied from %s' % f2,
                     )
                 else:
                     actions[f] = (
                         mergestatemod.ACTION_MERGE,
                         (f2, f, f2, True, pa.node()),
                         b'remote moved from %s' % f2,
                     )
             elif f not in ma:
                 # local unknown, remote created: the logic is described by the
                 # following table:
                 #
                 # force  branchmerge  different  |  action
                 #   n         *           *      |   create
                 #   y         n           *      |   create
                 #   y         y           n      |   create
                 #   y         y           y      |   merge
                 #
                 # Checking whether the files are different is expensive, so we
                 # don't do that when we can avoid it.
                 if not force:
                     actions[f] = (
                         mergestatemod.ACTION_CREATED,
                         (fl2,),
                         b'remote created',
                     )
                 elif not branchmerge:
                     actions[f] = (
                         mergestatemod.ACTION_CREATED,
                         (fl2,),
                         b'remote created',
                     )
                 else:
                     actions[f] = (
                         mergestatemod.ACTION_CREATED_MERGE,
                         (fl2, pa.node()),
                         b'remote created, get or merge',
                     )
             elif n2 != ma[f]:
                 df = None
                 for d in branch_copies1.dirmove:
                     if f.startswith(d):
                         # new file added in a directory that was moved
                         df = branch_copies1.dirmove[d] + f[len(d) :]
                         break
                 if df is not None and df in m1:
                     actions[df] = (
                         mergestatemod.ACTION_MERGE,
                         (df, f, f, False, pa.node()),
                         b'local directory rename - respect move '
                         b'from %s' % f,
                     )
                 elif acceptremote:
                     actions[f] = (
                         mergestatemod.ACTION_CREATED,
                         (fl2,),
                         b'remote recreating',
                     )
                 else:
                     actions[f] = (
                         mergestatemod.ACTION_DELETED_CHANGED,
                         (None, f, f, False, pa.node()),
                         b'prompt deleted/changed',
                     )
 
     if repo.ui.configbool(b'experimental', b'merge.checkpathconflicts'):
         # If we are merging, look for path conflicts.
         checkpathconflicts(repo, wctx, p2, actions)
 
     narrowmatch = repo.narrowmatch()
     if not narrowmatch.always():
         # Updates "actions" in place
         _filternarrowactions(narrowmatch, branchmerge, actions)
 
     renamedelete = branch_copies1.renamedelete
     renamedelete.update(branch_copies2.renamedelete)
 
     return actions, diverge, renamedelete
 
 
 def _resolvetrivial(repo, wctx, mctx, ancestor, actions):
     """Resolves false conflicts where the nodeid changed but the content
        remained the same."""
     # We force a copy of actions.items() because we're going to mutate
     # actions as we resolve trivial conflicts.
     for f, (m, args, msg) in list(actions.items()):
         if (
             m == mergestatemod.ACTION_CHANGED_DELETED
             and f in ancestor
             and not wctx[f].cmp(ancestor[f])
         ):
             # local did change but ended up with same content
             actions[f] = mergestatemod.ACTION_REMOVE, None, b'prompt same'
         elif (
             m == mergestatemod.ACTION_DELETED_CHANGED
             and f in ancestor
             and not mctx[f].cmp(ancestor[f])
         ):
             # remote did change but ended up with same content
             del actions[f]  # don't get = keep local deleted
 
 
 def calculateupdates(
     repo,
     wctx,
     mctx,
     ancestors,
     branchmerge,
     force,
     acceptremote,
     followcopies,
     matcher=None,
     mergeforce=False,
 ):
     """Calculate the actions needed to merge mctx into wctx using ancestors"""
     # Avoid cycle.
     from . import sparse
 
     if len(ancestors) == 1:  # default
         actions, diverge, renamedelete = manifestmerge(
             repo,
             wctx,
             mctx,
             ancestors[0],
             branchmerge,
             force,
             matcher,
             acceptremote,
             followcopies,
         )
         _checkunknownfiles(repo, wctx, mctx, force, actions, mergeforce)
 
     else:  # only when merge.preferancestor=* - the default
         repo.ui.note(
             _(b"note: merging %s and %s using bids from ancestors %s\n")
             % (
                 wctx,
                 mctx,
                 _(b' and ').join(pycompat.bytestr(anc) for anc in ancestors),
             )
         )
 
         # Call for bids
         fbids = (
             {}
         )  # mapping filename to bids (action method to list af actions)
         diverge, renamedelete = None, None
         for ancestor in ancestors:
             repo.ui.note(_(b'\ncalculating bids for ancestor %s\n') % ancestor)
             actions, diverge1, renamedelete1 = manifestmerge(
                 repo,
                 wctx,
                 mctx,
                 ancestor,
                 branchmerge,
                 force,
                 matcher,
                 acceptremote,
                 followcopies,
                 forcefulldiff=True,
             )
             _checkunknownfiles(repo, wctx, mctx, force, actions, mergeforce)
 
             # Track the shortest set of warning on the theory that bid
             # merge will correctly incorporate more information
             if diverge is None or len(diverge1) < len(diverge):
                 diverge = diverge1
             if renamedelete is None or len(renamedelete) < len(renamedelete1):
                 renamedelete = renamedelete1
 
             for f, a in sorted(pycompat.iteritems(actions)):
                 m, args, msg = a
                 if m == mergestatemod.ACTION_GET_OTHER_AND_STORE:
                     m = mergestatemod.ACTION_GET
                 repo.ui.debug(b' %s: %s -> %s\n' % (f, msg, m))
                 if f in fbids:
                     d = fbids[f]
                     if m in d:
                         d[m].append(a)
                     else:
                         d[m] = [a]
                 else:
                     fbids[f] = {m: [a]}
 
         # Pick the best bid for each file
         repo.ui.note(_(b'\nauction for merging merge bids\n'))
         actions = {}
         for f, bids in sorted(fbids.items()):
             # bids is a mapping from action method to list af actions
             # Consensus?
             if len(bids) == 1:  # all bids are the same kind of method
                 m, l = list(bids.items())[0]
                 if all(a == l[0] for a in l[1:]):  # len(bids) is > 1
                     repo.ui.note(_(b" %s: consensus for %s\n") % (f, m))
                     actions[f] = l[0]
                     continue
             # If keep is an option, just do it.
             if mergestatemod.ACTION_KEEP in bids:
                 repo.ui.note(_(b" %s: picking 'keep' action\n") % f)
                 actions[f] = bids[mergestatemod.ACTION_KEEP][0]
                 continue
             # If there are gets and they all agree [how could they not?], do it.
             if mergestatemod.ACTION_GET in bids:
                 ga0 = bids[mergestatemod.ACTION_GET][0]
                 if all(a == ga0 for a in bids[mergestatemod.ACTION_GET][1:]):
                     repo.ui.note(_(b" %s: picking 'get' action\n") % f)
                     actions[f] = ga0
                     continue
             # TODO: Consider other simple actions such as mode changes
             # Handle inefficient democrazy.
             repo.ui.note(_(b' %s: multiple bids for merge action:\n') % f)
             for m, l in sorted(bids.items()):
                 for _f, args, msg in l:
                     repo.ui.note(b'  %s -> %s\n' % (msg, m))
             # Pick random action. TODO: Instead, prompt user when resolving
             m, l = list(bids.items())[0]
             repo.ui.warn(
                 _(b' %s: ambiguous merge - picked %s action\n') % (f, m)
             )
             actions[f] = l[0]
             continue
         repo.ui.note(_(b'end of auction\n\n'))
 
     if wctx.rev() is None:
         fractions = _forgetremoved(wctx, mctx, branchmerge)
         actions.update(fractions)
 
     prunedactions = sparse.filterupdatesactions(
         repo, wctx, mctx, branchmerge, actions
     )
     _resolvetrivial(repo, wctx, mctx, ancestors[0], actions)
 
     return prunedactions, diverge, renamedelete
 
 
 def _getcwd():
     try:
         return encoding.getcwd()
     except OSError as err:
         if err.errno == errno.ENOENT:
             return None
         raise
 
 
 def batchremove(repo, wctx, actions):
     """apply removes to the working directory
 
     yields tuples for progress updates
     """
     verbose = repo.ui.verbose
     cwd = _getcwd()
     i = 0
     for f, args, msg in actions:
         repo.ui.debug(b" %s: %s -> r\n" % (f, msg))
         if verbose:
             repo.ui.note(_(b"removing %s\n") % f)
         wctx[f].audit()
         try:
             wctx[f].remove(ignoremissing=True)
         except OSError as inst:
             repo.ui.warn(
                 _(b"update failed to remove %s: %s!\n") % (f, inst.strerror)
             )
         if i == 100:
             yield i, f
             i = 0
         i += 1
     if i > 0:
         yield i, f
 
     if cwd and not _getcwd():
         # cwd was removed in the course of removing files; print a helpful
         # warning.
         repo.ui.warn(
             _(
                 b"current directory was removed\n"
                 b"(consider changing to repo root: %s)\n"
             )
             % repo.root
         )
 
 
 def batchget(repo, mctx, wctx, wantfiledata, actions):
     """apply gets to the working directory
 
     mctx is the context to get from
 
     Yields arbitrarily many (False, tuple) for progress updates, followed by
     exactly one (True, filedata). When wantfiledata is false, filedata is an
     empty dict. When wantfiledata is true, filedata[f] is a triple (mode, size,
     mtime) of the file f written for each action.
     """
     filedata = {}
     verbose = repo.ui.verbose
     fctx = mctx.filectx
     ui = repo.ui
     i = 0
     with repo.wvfs.backgroundclosing(ui, expectedcount=len(actions)):
         for f, (flags, backup), msg in actions:
             repo.ui.debug(b" %s: %s -> g\n" % (f, msg))
             if verbose:
                 repo.ui.note(_(b"getting %s\n") % f)
 
             if backup:
                 # If a file or directory exists with the same name, back that
                 # up.  Otherwise, look to see if there is a file that conflicts
                 # with a directory this file is in, and if so, back that up.
                 conflicting = f
                 if not repo.wvfs.lexists(f):
                     for p in pathutil.finddirs(f):
                         if repo.wvfs.isfileorlink(p):
                             conflicting = p
                             break
                 if repo.wvfs.lexists(conflicting):
                     orig = scmutil.backuppath(ui, repo, conflicting)
                     util.rename(repo.wjoin(conflicting), orig)
             wfctx = wctx[f]
             wfctx.clearunknown()
             atomictemp = ui.configbool(b"experimental", b"update.atomic-file")
             size = wfctx.write(
                 fctx(f).data(),
                 flags,
                 backgroundclose=True,
                 atomictemp=atomictemp,
             )
             if wantfiledata:
                 s = wfctx.lstat()
                 mode = s.st_mode
                 mtime = s[stat.ST_MTIME]
                 filedata[f] = (mode, size, mtime)  # for dirstate.normal
             if i == 100:
                 yield False, (i, f)
                 i = 0
             i += 1
     if i > 0:
         yield False, (i, f)
     yield True, filedata
 
 
 def _prefetchfiles(repo, ctx, actions):
     """Invoke ``scmutil.prefetchfiles()`` for the files relevant to the dict
     of merge actions.  ``ctx`` is the context being merged in."""
 
     # Skipping 'a', 'am', 'f', 'r', 'dm', 'e', 'k', 'p' and 'pr', because they
     # don't touch the context to be merged in.  'cd' is skipped, because
     # changed/deleted never resolves to something from the remote side.
     oplist = [
         actions[a]
         for a in (
             mergestatemod.ACTION_GET,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
             mergestatemod.ACTION_MERGE,
         )
     ]
     prefetch = scmutil.prefetchfiles
     matchfiles = scmutil.matchfiles
     prefetch(
         repo,
         [
             (
                 ctx.rev(),
                 matchfiles(
                     repo, [f for sublist in oplist for f, args, msg in sublist]
                 ),
             )
         ],
     )
 
 
 @attr.s(frozen=True)
 class updateresult(object):
     updatedcount = attr.ib()
     mergedcount = attr.ib()
     removedcount = attr.ib()
     unresolvedcount = attr.ib()
 
     def isempty(self):
         return not (
             self.updatedcount
             or self.mergedcount
             or self.removedcount
             or self.unresolvedcount
         )
 
 
 def emptyactions():
     """create an actions dict, to be populated and passed to applyupdates()"""
     return {
         m: []
         for m in (
             mergestatemod.ACTION_ADD,
             mergestatemod.ACTION_ADD_MODIFIED,
             mergestatemod.ACTION_FORGET,
             mergestatemod.ACTION_GET,
             mergestatemod.ACTION_CHANGED_DELETED,
             mergestatemod.ACTION_DELETED_CHANGED,
             mergestatemod.ACTION_REMOVE,
             mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL,
             mergestatemod.ACTION_LOCAL_DIR_RENAME_GET,
             mergestatemod.ACTION_MERGE,
             mergestatemod.ACTION_EXEC,
             mergestatemod.ACTION_KEEP,
             mergestatemod.ACTION_PATH_CONFLICT,
             mergestatemod.ACTION_PATH_CONFLICT_RESOLVE,
             mergestatemod.ACTION_GET_OTHER_AND_STORE,
         )
     }
 
 
 def applyupdates(
     repo, actions, wctx, mctx, overwrite, wantfiledata, labels=None
 ):
     """apply the merge action list to the working directory
 
     wctx is the working copy context
     mctx is the context to be merged into the working copy
 
     Return a tuple of (counts, filedata), where counts is a tuple
     (updated, merged, removed, unresolved) that describes how many
     files were affected by the update, and filedata is as described in
     batchget.
     """
 
     _prefetchfiles(repo, mctx, actions)
 
     updated, merged, removed = 0, 0, 0
     ms = mergestatemod.mergestate.clean(
         repo, wctx.p1().node(), mctx.node(), labels
     )
 
     # add ACTION_GET_OTHER_AND_STORE to mergestate
     for e in actions[mergestatemod.ACTION_GET_OTHER_AND_STORE]:
         ms.addmergedother(e[0])
 
     moves = []
     for m, l in actions.items():
         l.sort()
 
     # 'cd' and 'dc' actions are treated like other merge conflicts
     mergeactions = sorted(actions[mergestatemod.ACTION_CHANGED_DELETED])
     mergeactions.extend(sorted(actions[mergestatemod.ACTION_DELETED_CHANGED]))
     mergeactions.extend(actions[mergestatemod.ACTION_MERGE])
     for f, args, msg in mergeactions:
         f1, f2, fa, move, anc = args
         if f == b'.hgsubstate':  # merged internally
             continue
         if f1 is None:
             fcl = filemerge.absentfilectx(wctx, fa)
         else:
             repo.ui.debug(b" preserving %s for resolve of %s\n" % (f1, f))
             fcl = wctx[f1]
         if f2 is None:
             fco = filemerge.absentfilectx(mctx, fa)
         else:
             fco = mctx[f2]
         actx = repo[anc]
         if fa in actx:
             fca = actx[fa]
         else:
             # TODO: move to absentfilectx
             fca = repo.filectx(f1, fileid=nullrev)
         ms.add(fcl, fco, fca, f)
         if f1 != f and move:
             moves.append(f1)
 
     # remove renamed files after safely stored
     for f in moves:
         if wctx[f].lexists():
             repo.ui.debug(b"removing %s\n" % f)
             wctx[f].audit()
             wctx[f].remove()
 
     numupdates = sum(
         len(l) for m, l in actions.items() if m != mergestatemod.ACTION_KEEP
     )
     progress = repo.ui.makeprogress(
         _(b'updating'), unit=_(b'files'), total=numupdates
     )
 
     if [
         a
         for a in actions[mergestatemod.ACTION_REMOVE]
         if a[0] == b'.hgsubstate'
     ]:
         subrepoutil.submerge(repo, wctx, mctx, wctx, overwrite, labels)
 
     # record path conflicts
     for f, args, msg in actions[mergestatemod.ACTION_PATH_CONFLICT]:
         f1, fo = args
         s = repo.ui.status
         s(
             _(
                 b"%s: path conflict - a file or link has the same name as a "
                 b"directory\n"
             )
             % f
         )
         if fo == b'l':
             s(_(b"the local file has been renamed to %s\n") % f1)
         else:
             s(_(b"the remote file has been renamed to %s\n") % f1)
         s(_(b"resolve manually then use 'hg resolve --mark %s'\n") % f)
         ms.addpath(f, f1, fo)
         progress.increment(item=f)
 
     # When merging in-memory, we can't support worker processes, so set the
     # per-item cost at 0 in that case.
     cost = 0 if wctx.isinmemory() else 0.001
 
     # remove in parallel (must come before resolving path conflicts and getting)
     prog = worker.worker(
         repo.ui,
         cost,
         batchremove,
         (repo, wctx),
         actions[mergestatemod.ACTION_REMOVE],
     )
     for i, item in prog:
         progress.increment(step=i, item=item)
     removed = len(actions[mergestatemod.ACTION_REMOVE])
 
     # resolve path conflicts (must come before getting)
     for f, args, msg in actions[mergestatemod.ACTION_PATH_CONFLICT_RESOLVE]:
         repo.ui.debug(b" %s: %s -> pr\n" % (f, msg))
         (f0, origf0) = args
         if wctx[f0].lexists():
             repo.ui.note(_(b"moving %s to %s\n") % (f0, f))
             wctx[f].audit()
             wctx[f].write(wctx.filectx(f0).data(), wctx.filectx(f0).flags())
             wctx[f0].remove()
         progress.increment(item=f)
 
     # get in parallel.
     threadsafe = repo.ui.configbool(
         b'experimental', b'worker.wdir-get-thread-safe'
     )
     prog = worker.worker(
         repo.ui,
         cost,
         batchget,
         (repo, mctx, wctx, wantfiledata),
         actions[mergestatemod.ACTION_GET],
         threadsafe=threadsafe,
         hasretval=True,
     )
     getfiledata = {}
     for final, res in prog:
         if final:
             getfiledata = res
         else:
             i, item = res
             progress.increment(step=i, item=item)
     updated = len(actions[mergestatemod.ACTION_GET])
 
     if [a for a in actions[mergestatemod.ACTION_GET] if a[0] == b'.hgsubstate']:
         subrepoutil.submerge(repo, wctx, mctx, wctx, overwrite, labels)
 
     # forget (manifest only, just log it) (must come first)
     for f, args, msg in actions[mergestatemod.ACTION_FORGET]:
         repo.ui.debug(b" %s: %s -> f\n" % (f, msg))
         progress.increment(item=f)
 
     # re-add (manifest only, just log it)
     for f, args, msg in actions[mergestatemod.ACTION_ADD]:
         repo.ui.debug(b" %s: %s -> a\n" % (f, msg))
         progress.increment(item=f)
 
     # re-add/mark as modified (manifest only, just log it)
     for f, args, msg in actions[mergestatemod.ACTION_ADD_MODIFIED]:
         repo.ui.debug(b" %s: %s -> am\n" % (f, msg))
         progress.increment(item=f)
 
     # keep (noop, just log it)
     for f, args, msg in actions[mergestatemod.ACTION_KEEP]:
         repo.ui.debug(b" %s: %s -> k\n" % (f, msg))
         # no progress
 
     # directory rename, move local
     for f, args, msg in actions[mergestatemod.ACTION_DIR_RENAME_MOVE_LOCAL]:
         repo.ui.debug(b" %s: %s -> dm\n" % (f, msg))
         progress.increment(item=f)
         f0, flags = args
         repo.ui.note(_(b"moving %s to %s\n") % (f0, f))
         wctx[f].audit()
         wctx[f].write(wctx.filectx(f0).data(), flags)
         wctx[f0].remove()
         updated += 1
 
     # local directory rename, get
     for f, args, msg in actions[mergestatemod.ACTION_LOCAL_DIR_RENAME_GET]:
         repo.ui.debug(b" %s: %s -> dg\n" % (f, msg))
         progress.increment(item=f)
         f0, flags = args
         repo.ui.note(_(b"getting %s to %s\n") % (f0, f))
         wctx[f].write(mctx.filectx(f0).data(), flags)
         updated += 1
 
     # exec
     for f, args, msg in actions[mergestatemod.ACTION_EXEC]:
         repo.ui.debug(b" %s: %s -> e\n" % (f, msg))
         progress.increment(item=f)
         (flags,) = args
         wctx[f].audit()
         wctx[f].setflags(b'l' in flags, b'x' in flags)
         updated += 1
 
     # the ordering is important here -- ms.mergedriver will raise if the merge
     # driver has changed, and we want to be able to bypass it when overwrite is
     # True
     usemergedriver = not overwrite and mergeactions and ms.mergedriver
 
     if usemergedriver:
         if wctx.isinmemory():
             raise error.InMemoryMergeConflictsError(
                 b"in-memory merge does not support mergedriver"
             )
         ms.commit()
         proceed = driverpreprocess(repo, ms, wctx, labels=labels)
         # the driver might leave some files unresolved
         unresolvedf = set(ms.unresolved())
         if not proceed:
             # XXX setting unresolved to at least 1 is a hack to make sure we
             # error out
             return updateresult(
                 updated, merged, removed, max(len(unresolvedf), 1)
             )
         newactions = []
         for f, args, msg in mergeactions:
             if f in unresolvedf:
                 newactions.append((f, args, msg))
         mergeactions = newactions
 
     try:
         # premerge
         tocomplete = []
         for f, args, msg in mergeactions:
             repo.ui.debug(b" %s: %s -> m (premerge)\n" % (f, msg))
             progress.increment(item=f)
             if f == b'.hgsubstate':  # subrepo states need updating
                 subrepoutil.submerge(
                     repo, wctx, mctx, wctx.ancestor(mctx), overwrite, labels
                 )
                 continue
             wctx[f].audit()
             complete, r = ms.preresolve(f, wctx)
             if not complete:
                 numupdates += 1
                 tocomplete.append((f, args, msg))
 
         # merge
         for f, args, msg in tocomplete:
             repo.ui.debug(b" %s: %s -> m (merge)\n" % (f, msg))
             progress.increment(item=f, total=numupdates)
             ms.resolve(f, wctx)
 
     finally:
         ms.commit()
 
     unresolved = ms.unresolvedcount()
 
     if (
         usemergedriver
         and not unresolved
         and ms.mdstate() != mergestatemod.MERGE_DRIVER_STATE_SUCCESS
     ):
         if not driverconclude(repo, ms, wctx, labels=labels):
             # XXX setting unresolved to at least 1 is a hack to make sure we
             # error out
             unresolved = max(unresolved, 1)
 
         ms.commit()
 
     msupdated, msmerged, msremoved = ms.counts()
     updated += msupdated
     merged += msmerged
     removed += msremoved
 
     extraactions = ms.actions()
     if extraactions:
         mfiles = {a[0] for a in actions[mergestatemod.ACTION_MERGE]}
         for k, acts in pycompat.iteritems(extraactions):
             actions[k].extend(acts)
             if k == mergestatemod.ACTION_GET and wantfiledata:
                 # no filedata until mergestate is updated to provide it
                 for a in acts:
                     getfiledata[a[0]] = None
             # Remove these files from actions[ACTION_MERGE] as well. This is
             # important because in recordupdates, files in actions[ACTION_MERGE]
             # are processed after files in other actions, and the merge driver
             # might add files to those actions via extraactions above. This can
             # lead to a file being recorded twice, with poor results. This is
             # especially problematic for actions[ACTION_REMOVE] (currently only
             # possible with the merge driver in the initial merge process;
             # interrupted merges don't go through this flow).
             #
             # The real fix here is to have indexes by both file and action so
             # that when the action for a file is changed it is automatically
             # reflected in the other action lists. But that involves a more
             # complex data structure, so this will do for now.
             #
             # We don't need to do the same operation for 'dc' and 'cd' because
             # those lists aren't consulted again.
             mfiles.difference_update(a[0] for a in acts)
 
         actions[mergestatemod.ACTION_MERGE] = [
             a for a in actions[mergestatemod.ACTION_MERGE] if a[0] in mfiles
         ]
 
     progress.complete()
     assert len(getfiledata) == (
         len(actions[mergestatemod.ACTION_GET]) if wantfiledata else 0
     )
     return updateresult(updated, merged, removed, unresolved), getfiledata
 
 
 UPDATECHECK_ABORT = b'abort'  # handled at higher layers
 UPDATECHECK_NONE = b'none'
 UPDATECHECK_LINEAR = b'linear'
 UPDATECHECK_NO_CONFLICT = b'noconflict'
 
 
 def update(
     repo,
     node,
     branchmerge,
     force,
     ancestor=None,
     mergeancestor=False,
     labels=None,
     matcher=None,
     mergeforce=False,
     updatedirstate=True,
     updatecheck=None,
     wc=None,
 ):
     """
     Perform a merge between the working directory and the given node
 
     node = the node to update to
     branchmerge = whether to merge between branches
     force = whether to force branch merging or file overwriting
     matcher = a matcher to filter file lists (dirstate not updated)
     mergeancestor = whether it is merging with an ancestor. If true,
       we should accept the incoming changes for any prompts that occur.
       If false, merging with an ancestor (fast-forward) is only allowed
       between different named branches. This flag is used by rebase extension
       as a temporary fix and should be avoided in general.
     labels = labels to use for base, local and other
     mergeforce = whether the merge was run with 'merge --force' (deprecated): if
       this is True, then 'force' should be True as well.
 
     The table below shows all the behaviors of the update command given the
     -c/--check and -C/--clean or no options, whether the working directory is
     dirty, whether a revision is specified, and the relationship of the parent
     rev to the target rev (linear or not). Match from top first. The -n
     option doesn't exist on the command line, but represents the
     experimental.updatecheck=noconflict option.
 
     This logic is tested by test-update-branches.t.
 
     -c  -C  -n  -m  dirty  rev  linear  |  result
      y   y   *   *    *     *     *     |    (1)
      y   *   y   *    *     *     *     |    (1)
      y   *   *   y    *     *     *     |    (1)
      *   y   y   *    *     *     *     |    (1)
      *   y   *   y    *     *     *     |    (1)
      *   *   y   y    *     *     *     |    (1)
      *   *   *   *    *     n     n     |     x
      *   *   *   *    n     *     *     |    ok
      n   n   n   n    y     *     y     |   merge
      n   n   n   n    y     y     n     |    (2)
      n   n   n   y    y     *     *     |   merge
      n   n   y   n    y     *     *     |  merge if no conflict
      n   y   n   n    y     *     *     |  discard
      y   n   n   n    y     *     *     |    (3)
 
     x = can't happen
     * = don't-care
     1 = incompatible options (checked in commands.py)
     2 = abort: uncommitted changes (commit or update --clean to discard changes)
     3 = abort: uncommitted changes (checked in commands.py)
 
     The merge is performed inside ``wc``, a workingctx-like objects. It defaults
     to repo[None] if None is passed.
 
     Return the same tuple as applyupdates().
     """
     # Avoid cycle.
     from . import sparse
 
     # This function used to find the default destination if node was None, but
     # that's now in destutil.py.
     assert node is not None
     if not branchmerge and not force:
         # TODO: remove the default once all callers that pass branchmerge=False
         # and force=False pass a value for updatecheck. We may want to allow
         # updatecheck='abort' to better suppport some of these callers.
         if updatecheck is None:
             updatecheck = UPDATECHECK_LINEAR
         if updatecheck not in (
             UPDATECHECK_NONE,
             UPDATECHECK_LINEAR,
             UPDATECHECK_NO_CONFLICT,
         ):
             raise ValueError(
                 r'Invalid updatecheck %r (can accept %r)'
                 % (
                     updatecheck,
                     (
                         UPDATECHECK_NONE,
                         UPDATECHECK_LINEAR,
                         UPDATECHECK_NO_CONFLICT,
                     ),
                 )
             )
     if wc is not None and wc.isinmemory():
         maybe_wlock = util.nullcontextmanager()
     else:
         maybe_wlock = repo.wlock()
     with maybe_wlock:
         if wc is None:
             wc = repo[None]
         pl = wc.parents()
         p1 = pl[0]
         p2 = repo[node]
         if ancestor is not None:
             pas = [repo[ancestor]]
         else:
             if repo.ui.configlist(b'merge', b'preferancestor') == [b'*']:
                 cahs = repo.changelog.commonancestorsheads(p1.node(), p2.node())
                 pas = [repo[anc] for anc in (sorted(cahs) or [nullid])]
             else:
                 pas = [p1.ancestor(p2, warn=branchmerge)]
 
         fp1, fp2, xp1, xp2 = p1.node(), p2.node(), bytes(p1), bytes(p2)
 
         overwrite = force and not branchmerge
         ### check phase
         if not overwrite:
             if len(pl) > 1:
                 raise error.Abort(_(b"outstanding uncommitted merge"))
             ms = mergestatemod.mergestate.read(repo)
             if list(ms.unresolved()):
                 raise error.Abort(
                     _(b"outstanding merge conflicts"),
                     hint=_(b"use 'hg resolve' to resolve"),
                 )
         if branchmerge:
             if pas == [p2]:
                 raise error.Abort(
                     _(
                         b"merging with a working directory ancestor"
                         b" has no effect"
                     )
                 )
             elif pas == [p1]:
                 if not mergeancestor and wc.branch() == p2.branch():
                     raise error.Abort(
                         _(b"nothing to merge"),
                         hint=_(b"use 'hg update' or check 'hg heads'"),
                     )
             if not force and (wc.files() or wc.deleted()):
                 raise error.Abort(
                     _(b"uncommitted changes"),
                     hint=_(b"use 'hg status' to list changes"),
                 )
             if not wc.isinmemory():
                 for s in sorted(wc.substate):
                     wc.sub(s).bailifchanged()
 
         elif not overwrite:
             if p1 == p2:  # no-op update
                 # call the hooks and exit early
                 repo.hook(b'preupdate', throw=True, parent1=xp2, parent2=b'')
                 repo.hook(b'update', parent1=xp2, parent2=b'', error=0)
                 return updateresult(0, 0, 0, 0)
 
             if updatecheck == UPDATECHECK_LINEAR and pas not in (
                 [p1],
                 [p2],
             ):  # nonlinear
                 dirty = wc.dirty(missing=True)
                 if dirty:
                     # Branching is a bit strange to ensure we do the minimal
                     # amount of call to obsutil.foreground.
                     foreground = obsutil.foreground(repo, [p1.node()])
                     # note: the <node> variable contains a random identifier
                     if repo[node].node() in foreground:
                         pass  # allow updating to successors
                     else:
                         msg = _(b"uncommitted changes")
                         hint = _(b"commit or update --clean to discard changes")
                         raise error.UpdateAbort(msg, hint=hint)
                 else:
                     # Allow jumping branches if clean and specific rev given
                     pass
 
         if overwrite:
             pas = [wc]
         elif not branchmerge:
             pas = [p1]
 
         # deprecated config: merge.followcopies
         followcopies = repo.ui.configbool(b'merge', b'followcopies')
         if overwrite:
             followcopies = False
         elif not pas[0]:
             followcopies = False
         if not branchmerge and not wc.dirty(missing=True):
             followcopies = False
 
         ### calculate phase
         actionbyfile, diverge, renamedelete = calculateupdates(
             repo,
             wc,
             p2,
             pas,
             branchmerge,
             force,
             mergeancestor,
             followcopies,
             matcher=matcher,
             mergeforce=mergeforce,
         )
 
         if updatecheck == UPDATECHECK_NO_CONFLICT:
             for f, (m, args, msg) in pycompat.iteritems(actionbyfile):
                 if m not in (
                     mergestatemod.ACTION_GET,
                     mergestatemod.ACTION_KEEP,
                     mergestatemod.ACTION_EXEC,
                     mergestatemod.ACTION_REMOVE,
                     mergestatemod.ACTION_PATH_CONFLICT_RESOLVE,
                     mergestatemod.ACTION_GET_OTHER_AND_STORE,
                 ):
                     msg = _(b"conflicting changes")
                     hint = _(b"commit or update --clean to discard changes")
                     raise error.Abort(msg, hint=hint)
 
         # Prompt and create actions. Most of this is in the resolve phase
         # already, but we can't handle .hgsubstate in filemerge or
         # subrepoutil.submerge yet so we have to keep prompting for it.
         if b'.hgsubstate' in actionbyfile:
             f = b'.hgsubstate'
             m, args, msg = actionbyfile[f]
             prompts = filemerge.partextras(labels)
             prompts[b'f'] = f
             if m == mergestatemod.ACTION_CHANGED_DELETED:
                 if repo.ui.promptchoice(
                     _(
                         b"local%(l)s changed %(f)s which other%(o)s deleted\n"
                         b"use (c)hanged version or (d)elete?"
                         b"$$ &Changed $$ &Delete"
                     )
                     % prompts,
                     0,
                 ):
                     actionbyfile[f] = (
                         mergestatemod.ACTION_REMOVE,
                         None,
                         b'prompt delete',
                     )
                 elif f in p1:
                     actionbyfile[f] = (
                         mergestatemod.ACTION_ADD_MODIFIED,
                         None,
                         b'prompt keep',
                     )
                 else:
                     actionbyfile[f] = (
                         mergestatemod.ACTION_ADD,
                         None,
                         b'prompt keep',
                     )
             elif m == mergestatemod.ACTION_DELETED_CHANGED:
                 f1, f2, fa, move, anc = args
                 flags = p2[f2].flags()
                 if (
                     repo.ui.promptchoice(
                         _(
                             b"other%(o)s changed %(f)s which local%(l)s deleted\n"
                             b"use (c)hanged version or leave (d)eleted?"
                             b"$$ &Changed $$ &Deleted"
                         )
                         % prompts,
                         0,
                     )
                     == 0
                 ):
                     actionbyfile[f] = (
                         mergestatemod.ACTION_GET,
                         (flags, False),
                         b'prompt recreating',
                     )
                 else:
                     del actionbyfile[f]
 
         # Convert to dictionary-of-lists format
         actions = emptyactions()
         for f, (m, args, msg) in pycompat.iteritems(actionbyfile):
             if m not in actions:
                 actions[m] = []
             actions[m].append((f, args, msg))
 
         # ACTION_GET_OTHER_AND_STORE is a mergestatemod.ACTION_GET + store in mergestate
         for e in actions[mergestatemod.ACTION_GET_OTHER_AND_STORE]:
             actions[mergestatemod.ACTION_GET].append(e)
 
         if not util.fscasesensitive(repo.path):
             # check collision between files only in p2 for clean update
             if not branchmerge and (
                 force or not wc.dirty(missing=True, branch=False)
             ):
                 _checkcollision(repo, p2.manifest(), None)
             else:
                 _checkcollision(repo, wc.manifest(), actions)
 
         # divergent renames
         for f, fl in sorted(pycompat.iteritems(diverge)):
             repo.ui.warn(
                 _(
                     b"note: possible conflict - %s was renamed "
                     b"multiple times to:\n"
                 )
                 % f
             )
             for nf in sorted(fl):
                 repo.ui.warn(b" %s\n" % nf)
 
         # rename and delete
         for f, fl in sorted(pycompat.iteritems(renamedelete)):
             repo.ui.warn(
                 _(
                     b"note: possible conflict - %s was deleted "
                     b"and renamed to:\n"
                 )
                 % f
             )
             for nf in sorted(fl):
                 repo.ui.warn(b" %s\n" % nf)
 
         ### apply phase
         if not branchmerge:  # just jump to the new rev
             fp1, fp2, xp1, xp2 = fp2, nullid, xp2, b''
         # If we're doing a partial update, we need to skip updating
         # the dirstate.
         always = matcher is None or matcher.always()
         updatedirstate = updatedirstate and always and not wc.isinmemory()
         if updatedirstate:
             repo.hook(b'preupdate', throw=True, parent1=xp1, parent2=xp2)
             # note that we're in the middle of an update
             repo.vfs.write(b'updatestate', p2.hex())
 
         # Advertise fsmonitor when its presence could be useful.
         #
         # We only advertise when performing an update from an empty working
         # directory. This typically only occurs during initial clone.
         #
         # We give users a mechanism to disable the warning in case it is
         # annoying.
         #
         # We only allow on Linux and MacOS because that's where fsmonitor is
         # considered stable.
         fsmonitorwarning = repo.ui.configbool(b'fsmonitor', b'warn_when_unused')
         fsmonitorthreshold = repo.ui.configint(
             b'fsmonitor', b'warn_update_file_count'
         )
         try:
             # avoid cycle: extensions -> cmdutil -> merge
             from . import extensions
 
             extensions.find(b'fsmonitor')
             fsmonitorenabled = repo.ui.config(b'fsmonitor', b'mode') != b'off'
             # We intentionally don't look at whether fsmonitor has disabled
             # itself because a) fsmonitor may have already printed a warning
             # b) we only care about the config state here.
         except KeyError:
             fsmonitorenabled = False
 
         if (
             fsmonitorwarning
             and not fsmonitorenabled
             and p1.node() == nullid
             and len(actions[mergestatemod.ACTION_GET]) >= fsmonitorthreshold
             and pycompat.sysplatform.startswith((b'linux', b'darwin'))
         ):
             repo.ui.warn(
                 _(
                     b'(warning: large working directory being used without '
                     b'fsmonitor enabled; enable fsmonitor to improve performance; '
                     b'see "hg help -e fsmonitor")\n'
                 )
             )
 
         wantfiledata = updatedirstate and not branchmerge
         stats, getfiledata = applyupdates(
             repo, actions, wc, p2, overwrite, wantfiledata, labels=labels
         )
 
         if updatedirstate:
             with repo.dirstate.parentchange():
                 repo.setparents(fp1, fp2)
                 mergestatemod.recordupdates(
                     repo, actions, branchmerge, getfiledata
                 )
                 # update completed, clear state
                 util.unlink(repo.vfs.join(b'updatestate'))
 
                 if not branchmerge:
                     repo.dirstate.setbranch(p2.branch())
 
     # If we're updating to a location, clean up any stale temporary includes
     # (ex: this happens during hg rebase --abort).
     if not branchmerge:
         sparse.prunetemporaryincludes(repo)
 
     if updatedirstate:
         repo.hook(
             b'update', parent1=xp1, parent2=xp2, error=stats.unresolvedcount
         )
     return stats
 
 
 def merge(ctx, labels=None, force=False, wc=None):
     """Merge another topological branch into the working copy.
 
     force = whether the merge was run with 'merge --force' (deprecated)
     """
 
     return update(
         ctx.repo(),
         ctx.rev(),
         labels=labels,
         branchmerge=True,
         force=force,
         mergeforce=force,
         wc=wc,
     )
 
 
 def clean_update(ctx, wc=None):
     """Do a clean update to the given commit.
 
     This involves updating to the commit and discarding any changes in the
     working copy.
     """
     return update(ctx.repo(), ctx.rev(), branchmerge=False, force=True, wc=wc)
 
 
 def revert_to(ctx, matcher=None, wc=None):
     """Revert the working copy to the given commit.
 
     The working copy will keep its current parent(s) but its content will
     be the same as in the given commit.
     """
 
     return update(
         ctx.repo(),
         ctx.rev(),
         branchmerge=False,
         force=True,
         updatedirstate=False,
         matcher=matcher,
         wc=wc,
     )
 
 
 def graft(
     repo,
     ctx,
     base=None,
     labels=None,
     keepparent=False,
     keepconflictparent=False,
     wctx=None,
 ):
     """Do a graft-like merge.
 
     This is a merge where the merge ancestor is chosen such that one
     or more changesets are grafted onto the current changeset. In
     addition to the merge, this fixes up the dirstate to include only
     a single parent (if keepparent is False) and tries to duplicate any
     renames/copies appropriately.
 
     ctx - changeset to rebase
     base - merge base, or ctx.p1() if not specified
     labels - merge labels eg ['local', 'graft']
     keepparent - keep second parent if any
     keepconflictparent - if unresolved, keep parent used for the merge
 
     """
     # If we're grafting a descendant onto an ancestor, be sure to pass
     # mergeancestor=True to update. This does two things: 1) allows the merge if
     # the destination is the same as the parent of the ctx (so we can use graft
     # to copy commits), and 2) informs update that the incoming changes are
     # newer than the destination so it doesn't prompt about "remote changed foo
     # which local deleted".
     # We also pass mergeancestor=True when base is the same revision as p1. 2)
     # doesn't matter as there can't possibly be conflicts, but 1) is necessary.
     wctx = wctx or repo[None]
     pctx = wctx.p1()
     base = base or ctx.p1()
     mergeancestor = (
         repo.changelog.isancestor(pctx.node(), ctx.node())
         or pctx.rev() == base.rev()
     )
 
     stats = update(
         repo,
         ctx.node(),
         True,
         True,
         base.node(),
         mergeancestor=mergeancestor,
         labels=labels,
         wc=wctx,
     )
 
     if keepconflictparent and stats.unresolvedcount:
         pother = ctx.node()
     else:
         pother = nullid
         parents = ctx.parents()
         if keepparent and len(parents) == 2 and base in parents:
             parents.remove(base)
             pother = parents[0].node()
     # Never set both parents equal to each other
     if pother == pctx.node():
         pother = nullid
 
     if wctx.isinmemory():
         wctx.setparents(pctx.node(), pother)
         # fix up dirstate for copies and renames
         copies.graftcopies(wctx, ctx, base)
     else:
         with repo.dirstate.parentchange():
             repo.setparents(pctx.node(), pother)
             repo.dirstate.write(repo.currenttransaction())
             # fix up dirstate for copies and renames
             copies.graftcopies(wctx, ctx, base)
     return stats
 
 
 def purge(
     repo,
     matcher,
     unknown=True,
     ignored=False,
     removeemptydirs=True,
     removefiles=True,
     abortonerror=False,
     noop=False,
 ):
     """Purge the working directory of untracked files.
 
     ``matcher`` is a matcher configured to scan the working directory -
     potentially a subset.
 
     ``unknown`` controls whether unknown files should be purged.
 
     ``ignored`` controls whether ignored files should be purged.
 
     ``removeemptydirs`` controls whether empty directories should be removed.
 
     ``removefiles`` controls whether files are removed.
 
     ``abortonerror`` causes an exception to be raised if an error occurs
     deleting a file or directory.
 
     ``noop`` controls whether to actually remove files. If not defined, actions
     will be taken.
 
     Returns an iterable of relative paths in the working directory that were
     or would be removed.
     """
 
     def remove(removefn, path):
         try:
             removefn(path)
         except OSError:
             m = _(b'%s cannot be removed') % path
             if abortonerror:
                 raise error.Abort(m)
             else:
                 repo.ui.warn(_(b'warning: %s\n') % m)
 
     # There's no API to copy a matcher. So mutate the passed matcher and
     # restore it when we're done.
     oldtraversedir = matcher.traversedir
 
     res = []
 
     try:
         if removeemptydirs:
             directories = []
             matcher.traversedir = directories.append
 
         status = repo.status(match=matcher, ignored=ignored, unknown=unknown)
 
         if removefiles:
             for f in sorted(status.unknown + status.ignored):
                 if not noop:
                     repo.ui.note(_(b'removing file %s\n') % f)
                     remove(repo.wvfs.unlink, f)
                 res.append(f)
 
         if removeemptydirs:
             for f in sorted(directories, reverse=True):
                 if matcher(f) and not repo.wvfs.listdir(f):
                     if not noop:
                         repo.ui.note(_(b'removing directory %s\n') % f)
                         remove(repo.wvfs.rmdir, f)
                     res.append(f)
 
         return res
 
     finally:
         matcher.traversedir = oldtraversedir

tests/test-manifest.py

@@ -1,487 +1,454 b''
 from __future__ import absolute_import
 
 import binascii
 import itertools
 import silenttestrunner
 import unittest
 import zlib
 
 from mercurial import (
     manifest as manifestmod,
     match as matchmod,
     util,
 )
 
 EMTPY_MANIFEST = b''
 
 HASH_1 = b'1' * 40
 BIN_HASH_1 = binascii.unhexlify(HASH_1)
 HASH_2 = b'f' * 40
 BIN_HASH_2 = binascii.unhexlify(HASH_2)
 HASH_3 = b'1234567890abcdef0987654321deadbeef0fcafe'
 BIN_HASH_3 = binascii.unhexlify(HASH_3)
 A_SHORT_MANIFEST = (
     b'bar/baz/qux.py\0%(hash2)s%(flag2)s\n' b'foo\0%(hash1)s%(flag1)s\n'
 ) % {b'hash1': HASH_1, b'flag1': b'', b'hash2': HASH_2, b'flag2': b'l',}
 
 A_DEEPER_MANIFEST = (
     b'a/b/c/bar.py\0%(hash3)s%(flag1)s\n'
     b'a/b/c/bar.txt\0%(hash1)s%(flag1)s\n'
     b'a/b/c/foo.py\0%(hash3)s%(flag1)s\n'
     b'a/b/c/foo.txt\0%(hash2)s%(flag2)s\n'
     b'a/b/d/baz.py\0%(hash3)s%(flag1)s\n'
     b'a/b/d/qux.py\0%(hash1)s%(flag2)s\n'
     b'a/b/d/ten.txt\0%(hash3)s%(flag2)s\n'
     b'a/b/dog.py\0%(hash3)s%(flag1)s\n'
     b'a/b/fish.py\0%(hash2)s%(flag1)s\n'
     b'a/c/london.py\0%(hash3)s%(flag2)s\n'
     b'a/c/paper.txt\0%(hash2)s%(flag2)s\n'
     b'a/c/paris.py\0%(hash2)s%(flag1)s\n'
     b'a/d/apple.py\0%(hash3)s%(flag1)s\n'
     b'a/d/pizza.py\0%(hash3)s%(flag2)s\n'
     b'a/green.py\0%(hash1)s%(flag2)s\n'
     b'a/purple.py\0%(hash2)s%(flag1)s\n'
     b'app.py\0%(hash3)s%(flag1)s\n'
     b'readme.txt\0%(hash2)s%(flag1)s\n'
 ) % {
     b'hash1': HASH_1,
     b'flag1': b'',
     b'hash2': HASH_2,
     b'flag2': b'l',
     b'hash3': HASH_3,
 }
 
 HUGE_MANIFEST_ENTRIES = 200001
 
 izip = getattr(itertools, 'izip', zip)
 if 'xrange' not in globals():
     xrange = range
 
 A_HUGE_MANIFEST = b''.join(
     sorted(
         b'file%d\0%s%s\n' % (i, h, f)
         for i, h, f in izip(
             xrange(200001),
             itertools.cycle((HASH_1, HASH_2)),
             itertools.cycle((b'', b'x', b'l')),
         )
     )
 )
 
 
 class basemanifesttests(object):
     def parsemanifest(self, text):
         raise NotImplementedError('parsemanifest not implemented by test case')
 
     def testEmptyManifest(self):
         m = self.parsemanifest(EMTPY_MANIFEST)
         self.assertEqual(0, len(m))
         self.assertEqual([], list(m))
 
     def testManifest(self):
         m = self.parsemanifest(A_SHORT_MANIFEST)
         self.assertEqual([b'bar/baz/qux.py', b'foo'], list(m))
         self.assertEqual(BIN_HASH_2, m[b'bar/baz/qux.py'])
         self.assertEqual(b'l', m.flags(b'bar/baz/qux.py'))
         self.assertEqual(BIN_HASH_1, m[b'foo'])
         self.assertEqual(b'', m.flags(b'foo'))
         with self.assertRaises(KeyError):
             m[b'wat']
 
     def testManifestLongHashes(self):
         m = self.parsemanifest(b'a\0' + b'f' * 64 + b'\n')
         self.assertEqual(binascii.unhexlify(b'f' * 64), m[b'a'])
 
     def testSetItem(self):
         want = BIN_HASH_1
 
         m = self.parsemanifest(EMTPY_MANIFEST)
         m[b'a'] = want
         self.assertIn(b'a', m)
         self.assertEqual(want, m[b'a'])
         self.assertEqual(b'a\0' + HASH_1 + b'\n', m.text())
 
         m = self.parsemanifest(A_SHORT_MANIFEST)
         m[b'a'] = want
         self.assertEqual(want, m[b'a'])
         self.assertEqual(b'a\0' + HASH_1 + b'\n' + A_SHORT_MANIFEST, m.text())
 
     def testSetFlag(self):
         want = b'x'
 
         m = self.parsemanifest(EMTPY_MANIFEST)
         # first add a file; a file-less flag makes no sense
         m[b'a'] = BIN_HASH_1
         m.setflag(b'a', want)
         self.assertEqual(want, m.flags(b'a'))
         self.assertEqual(b'a\0' + HASH_1 + want + b'\n', m.text())
 
         m = self.parsemanifest(A_SHORT_MANIFEST)
         # first add a file; a file-less flag makes no sense
         m[b'a'] = BIN_HASH_1
         m.setflag(b'a', want)
         self.assertEqual(want, m.flags(b'a'))
         self.assertEqual(
             b'a\0' + HASH_1 + want + b'\n' + A_SHORT_MANIFEST, m.text()
         )
 
     def testCopy(self):
         m = self.parsemanifest(A_SHORT_MANIFEST)
         m[b'a'] = BIN_HASH_1
         m2 = m.copy()
         del m
         del m2  # make sure we don't double free() anything
 
     def testCompaction(self):
         unhex = binascii.unhexlify
         h1, h2 = unhex(HASH_1), unhex(HASH_2)
         m = self.parsemanifest(A_SHORT_MANIFEST)
         m[b'alpha'] = h1
         m[b'beta'] = h2
         del m[b'foo']
         want = b'alpha\0%s\nbar/baz/qux.py\0%sl\nbeta\0%s\n' % (
             HASH_1,
             HASH_2,
             HASH_2,
         )
         self.assertEqual(want, m.text())
         self.assertEqual(3, len(m))
         self.assertEqual([b'alpha', b'bar/baz/qux.py', b'beta'], list(m))
         self.assertEqual(h1, m[b'alpha'])
         self.assertEqual(h2, m[b'bar/baz/qux.py'])
         self.assertEqual(h2, m[b'beta'])
         self.assertEqual(b'', m.flags(b'alpha'))
         self.assertEqual(b'l', m.flags(b'bar/baz/qux.py'))
         self.assertEqual(b'', m.flags(b'beta'))
         with self.assertRaises(KeyError):
             m[b'foo']
 
-    def testSetGetNodeSuffix(self):
-        clean = self.parsemanifest(A_SHORT_MANIFEST)
-        m = self.parsemanifest(A_SHORT_MANIFEST)
-        h = m[b'foo']
-        f = m.flags(b'foo')
-        want = h + b'a'
-        # Merge code wants to set 21-byte fake hashes at times
-        m[b'foo'] = want
-        self.assertEqual(want, m[b'foo'])
-        self.assertEqual(
-            [(b'bar/baz/qux.py', BIN_HASH_2), (b'foo', BIN_HASH_1 + b'a')],
-            list(m.items()),
-        )
-        # Sometimes it even tries a 22-byte fake hash, but we can
-        # return 21 and it'll work out
-        m[b'foo'] = want + b'+'
-        self.assertEqual(want, m[b'foo'])
-        # make sure the suffix survives a copy
-        match = matchmod.match(util.localpath(b'/repo'), b'', [b're:foo'])
-        m2 = m._matches(match)
-        self.assertEqual(want, m2[b'foo'])
-        self.assertEqual(1, len(m2))
-        m2 = m.copy()
-        self.assertEqual(want, m2[b'foo'])
-        # suffix with iteration
-        self.assertEqual(
-            [(b'bar/baz/qux.py', BIN_HASH_2), (b'foo', want)], list(m.items())
-        )
-
-        # shows up in diff
-        self.assertEqual({b'foo': ((want, f), (h, b''))}, m.diff(clean))
-        self.assertEqual({b'foo': ((h, b''), (want, f))}, clean.diff(m))
-
     def testMatchException(self):
         m = self.parsemanifest(A_SHORT_MANIFEST)
         match = matchmod.match(util.localpath(b'/repo'), b'', [b're:.*'])
 
         def filt(path):
             if path == b'foo':
                 assert False
             return True
 
         match.matchfn = filt
         with self.assertRaises(AssertionError):
             m._matches(match)
 
     def testRemoveItem(self):
         m = self.parsemanifest(A_SHORT_MANIFEST)
         del m[b'foo']
         with self.assertRaises(KeyError):
             m[b'foo']
         self.assertEqual(1, len(m))
         self.assertEqual(1, len(list(m)))
         # now restore and make sure everything works right
         m[b'foo'] = b'a' * 20
         self.assertEqual(2, len(m))
         self.assertEqual(2, len(list(m)))
 
     def testManifestDiff(self):
         MISSING = (None, b'')
         addl = b'z-only-in-left\0' + HASH_1 + b'\n'
         addr = b'z-only-in-right\0' + HASH_2 + b'x\n'
         left = self.parsemanifest(
             A_SHORT_MANIFEST.replace(HASH_1, HASH_3 + b'x') + addl
         )
         right = self.parsemanifest(A_SHORT_MANIFEST + addr)
         want = {
             b'foo': ((BIN_HASH_3, b'x'), (BIN_HASH_1, b'')),
             b'z-only-in-left': ((BIN_HASH_1, b''), MISSING),
             b'z-only-in-right': (MISSING, (BIN_HASH_2, b'x')),
         }
         self.assertEqual(want, left.diff(right))
 
         want = {
             b'bar/baz/qux.py': (MISSING, (BIN_HASH_2, b'l')),
             b'foo': (MISSING, (BIN_HASH_3, b'x')),
             b'z-only-in-left': (MISSING, (BIN_HASH_1, b'')),
         }
         self.assertEqual(want, self.parsemanifest(EMTPY_MANIFEST).diff(left))
 
         want = {
             b'bar/baz/qux.py': ((BIN_HASH_2, b'l'), MISSING),
             b'foo': ((BIN_HASH_3, b'x'), MISSING),
             b'z-only-in-left': ((BIN_HASH_1, b''), MISSING),
         }
         self.assertEqual(want, left.diff(self.parsemanifest(EMTPY_MANIFEST)))
         copy = right.copy()
         del copy[b'z-only-in-right']
         del right[b'foo']
         want = {
             b'foo': (MISSING, (BIN_HASH_1, b'')),
             b'z-only-in-right': ((BIN_HASH_2, b'x'), MISSING),
         }
         self.assertEqual(want, right.diff(copy))
 
         short = self.parsemanifest(A_SHORT_MANIFEST)
         pruned = short.copy()
         del pruned[b'foo']
         want = {
             b'foo': ((BIN_HASH_1, b''), MISSING),
         }
         self.assertEqual(want, short.diff(pruned))
         want = {
             b'foo': (MISSING, (BIN_HASH_1, b'')),
         }
         self.assertEqual(want, pruned.diff(short))
         want = {
             b'bar/baz/qux.py': None,
             b'foo': (MISSING, (BIN_HASH_1, b'')),
         }
         self.assertEqual(want, pruned.diff(short, clean=True))
 
     def testReversedLines(self):
         backwards = b''.join(
             l + b'\n' for l in reversed(A_SHORT_MANIFEST.split(b'\n')) if l
         )
         try:
             self.parsemanifest(backwards)
             self.fail('Should have raised ValueError')
         except ValueError as v:
             self.assertIn('Manifest lines not in sorted order.', str(v))
 
     def testNoTerminalNewline(self):
         try:
             self.parsemanifest(A_SHORT_MANIFEST + b'wat')
             self.fail('Should have raised ValueError')
         except ValueError as v:
             self.assertIn('Manifest did not end in a newline.', str(v))
 
     def testNoNewLineAtAll(self):
         try:
             self.parsemanifest(b'wat')
             self.fail('Should have raised ValueError')
         except ValueError as v:
             self.assertIn('Manifest did not end in a newline.', str(v))
 
     def testHugeManifest(self):
         m = self.parsemanifest(A_HUGE_MANIFEST)
         self.assertEqual(HUGE_MANIFEST_ENTRIES, len(m))
         self.assertEqual(len(m), len(list(m)))
 
     def testMatchesMetadata(self):
         '''Tests matches() for a few specific files to make sure that both
         the set of files as well as their flags and nodeids are correct in
         the resulting manifest.'''
         m = self.parsemanifest(A_HUGE_MANIFEST)
 
         match = matchmod.exact([b'file1', b'file200', b'file300'])
         m2 = m._matches(match)
 
         w = (b'file1\0%sx\n' b'file200\0%sl\n' b'file300\0%s\n') % (
             HASH_2,
             HASH_1,
             HASH_1,
         )
         self.assertEqual(w, m2.text())
 
     def testMatchesNonexistentFile(self):
         '''Tests matches() for a small set of specific files, including one
         nonexistent file to make sure in only matches against existing files.
         '''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.exact(
             [b'a/b/c/bar.txt', b'a/b/d/qux.py', b'readme.txt', b'nonexistent']
         )
         m2 = m._matches(match)
 
         self.assertEqual(
             [b'a/b/c/bar.txt', b'a/b/d/qux.py', b'readme.txt'], m2.keys()
         )
 
     def testMatchesNonexistentDirectory(self):
         '''Tests matches() for a relpath match on a directory that doesn't
         actually exist.'''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.match(
             util.localpath(b'/repo'), b'', [b'a/f'], default=b'relpath'
         )
         m2 = m._matches(match)
 
         self.assertEqual([], m2.keys())
 
     def testMatchesExactLarge(self):
         '''Tests matches() for files matching a large list of exact files.
         '''
         m = self.parsemanifest(A_HUGE_MANIFEST)
 
         flist = m.keys()[80:300]
         match = matchmod.exact(flist)
         m2 = m._matches(match)
 
         self.assertEqual(flist, m2.keys())
 
     def testMatchesFull(self):
         '''Tests matches() for what should be a full match.'''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.match(util.localpath(b'/repo'), b'', [b''])
         m2 = m._matches(match)
 
         self.assertEqual(m.keys(), m2.keys())
 
     def testMatchesDirectory(self):
         '''Tests matches() on a relpath match on a directory, which should
         match against all files within said directory.'''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.match(
             util.localpath(b'/repo'), b'', [b'a/b'], default=b'relpath'
         )
         m2 = m._matches(match)
 
         self.assertEqual(
             [
                 b'a/b/c/bar.py',
                 b'a/b/c/bar.txt',
                 b'a/b/c/foo.py',
                 b'a/b/c/foo.txt',
                 b'a/b/d/baz.py',
                 b'a/b/d/qux.py',
                 b'a/b/d/ten.txt',
                 b'a/b/dog.py',
                 b'a/b/fish.py',
             ],
             m2.keys(),
         )
 
     def testMatchesExactPath(self):
         '''Tests matches() on an exact match on a directory, which should
         result in an empty manifest because you can't perform an exact match
         against a directory.'''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.exact([b'a/b'])
         m2 = m._matches(match)
 
         self.assertEqual([], m2.keys())
 
     def testMatchesCwd(self):
         '''Tests matches() on a relpath match with the current directory ('.')
         when not in the root directory.'''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.match(
             util.localpath(b'/repo'), b'a/b', [b'.'], default=b'relpath'
         )
         m2 = m._matches(match)
 
         self.assertEqual(
             [
                 b'a/b/c/bar.py',
                 b'a/b/c/bar.txt',
                 b'a/b/c/foo.py',
                 b'a/b/c/foo.txt',
                 b'a/b/d/baz.py',
                 b'a/b/d/qux.py',
                 b'a/b/d/ten.txt',
                 b'a/b/dog.py',
                 b'a/b/fish.py',
             ],
             m2.keys(),
         )
 
     def testMatchesWithPattern(self):
         '''Tests matches() for files matching a pattern that reside
         deeper than the specified directory.'''
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         match = matchmod.match(util.localpath(b'/repo'), b'', [b'a/b/*/*.txt'])
         m2 = m._matches(match)
 
         self.assertEqual(
             [b'a/b/c/bar.txt', b'a/b/c/foo.txt', b'a/b/d/ten.txt'], m2.keys()
         )
 
 
 class testmanifestdict(unittest.TestCase, basemanifesttests):
     def parsemanifest(self, text):
         return manifestmod.manifestdict(text)
 
     def testObviouslyBogusManifest(self):
         # This is a 163k manifest that came from oss-fuzz. It was a
         # timeout there, but when run normally it doesn't seem to
         # present any particular slowness.
         data = zlib.decompress(
             b'x\x9c\xed\xce;\n\x83\x00\x10\x04\xd0\x8deNa\x93~\xf1\x03\xc9q\xf4'
             b'\x14\xeaU\xbdB\xda\xd4\xe6Cj\xc1FA\xde+\x86\xe9f\xa2\xfci\xbb\xfb'
             b'\xa3\xef\xea\xba\xca\x7fk\x86q\x9a\xc6\xc8\xcc&\xb3\xcf\xf8\xb8|#'
             b'\x8a9\x00\xd8\xe6v\xf4\x01N\xe1\n\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00'
             b'\x00\x00\xc0\x8aey\x1d}\x01\xd8\xe0\xb9\xf3\xde\x1b\xcf\x17'
             b'\xac\xbe'
         )
         with self.assertRaises(ValueError):
             self.parsemanifest(data)
 
 
 class testtreemanifest(unittest.TestCase, basemanifesttests):
     def parsemanifest(self, text):
         return manifestmod.treemanifest(b'', text)
 
     def testWalkSubtrees(self):
         m = self.parsemanifest(A_DEEPER_MANIFEST)
 
         dirs = [s._dir for s in m.walksubtrees()]
         self.assertEqual(
             sorted(
                 [b'', b'a/', b'a/c/', b'a/d/', b'a/b/', b'a/b/c/', b'a/b/d/']
             ),
             sorted(dirs),
         )
 
         match = matchmod.match(util.localpath(b'/repo'), b'', [b'path:a/b/'])
         dirs = [s._dir for s in m.walksubtrees(matcher=match)]
         self.assertEqual(sorted([b'a/b/', b'a/b/c/', b'a/b/d/']), sorted(dirs))
 
 
 if __name__ == '__main__':
     silenttestrunner.main(__name__)