dirstate-v2: Initial Python serializer

This adds code seralizing a `map` and `copy_map` dicts into dirstate-v2
file formate. This is not used yet.

Differential Revision: https://phab.mercurial-scm.org/D11519

mercurial/cext/parsers.c

@@ -1,1155 +1,1174
 /*
  parsers.c - efficient content parsing
 
  Copyright 2008 Olivia Mackall <[e-mail]> and others
 
  This software may be used and distributed according to the terms of
  the GNU General Public License, incorporated herein by reference.
 */
 
 #define PY_SSIZE_T_CLEAN
 #include <Python.h>
 #include <ctype.h>
 #include <stddef.h>
 #include <string.h>
 
 #include "bitmanipulation.h"
 #include "charencode.h"
 #include "util.h"
 
 #ifdef IS_PY3K
 /* The mapping of Python types is meant to be temporary to get Python
  * 3 to compile. We should remove this once Python 3 support is fully
  * supported and proper types are used in the extensions themselves. */
 #define PyInt_Check PyLong_Check
 #define PyInt_FromLong PyLong_FromLong
 #define PyInt_FromSsize_t PyLong_FromSsize_t
 #define PyInt_AsLong PyLong_AsLong
 #endif
 
 static const char *const versionerrortext = "Python minor version mismatch";
 
 static const int dirstate_v1_from_p2 = -2;
 static const int dirstate_v1_nonnormal = -1;
 static const int ambiguous_time = -1;
 
 static PyObject *dict_new_presized(PyObject *self, PyObject *args)
 {
 	Py_ssize_t expected_size;
 
 	if (!PyArg_ParseTuple(args, "n:make_presized_dict", &expected_size)) {
 		return NULL;
 	}
 
 	return _dict_new_presized(expected_size);
 }
 
 static PyObject *dirstate_item_new(PyTypeObject *subtype, PyObject *args,
                                    PyObject *kwds)
 {
 	/* We do all the initialization here and not a tp_init function because
 	 * dirstate_item is immutable. */
 	dirstateItemObject *t;
 	int wc_tracked;
 	int p1_tracked;
 	int p2_info;
 	int has_meaningful_data;
 	int has_meaningful_mtime;
 	int mode;
 	int size;
 	int mtime;
 	PyObject *parentfiledata;
 	static char *keywords_name[] = {
 	    "wc_tracked",
 	    "p1_tracked",
 	    "p2_info",
 	    "has_meaningful_data",
 	    "has_meaningful_mtime",
 	    "parentfiledata",
 	    NULL,
 	};
 	wc_tracked = 0;
 	p1_tracked = 0;
 	p2_info = 0;
 	has_meaningful_mtime = 1;
 	has_meaningful_data = 1;
 	parentfiledata = Py_None;
 	if (!PyArg_ParseTupleAndKeywords(
 	        args, kwds, "|iiiiiO", keywords_name, &wc_tracked, &p1_tracked,
 	        &p2_info, &has_meaningful_data, &has_meaningful_mtime,
 	        &parentfiledata)) {
 		return NULL;
 	}
 	t = (dirstateItemObject *)subtype->tp_alloc(subtype, 1);
 	if (!t) {
 		return NULL;
 	}
 
 	t->flags = 0;
 	if (wc_tracked) {
 		t->flags |= dirstate_flag_wc_tracked;
 	}
 	if (p1_tracked) {
 		t->flags |= dirstate_flag_p1_tracked;
 	}
 	if (p2_info) {
 		t->flags |= dirstate_flag_p2_info;
 	}
 
 	if (parentfiledata != Py_None) {
 		if (!PyTuple_CheckExact(parentfiledata)) {
 			PyErr_SetString(
 			    PyExc_TypeError,
 			    "parentfiledata should be a Tuple or None");
 			return NULL;
 		}
 		mode = (int)PyLong_AsLong(PyTuple_GetItem(parentfiledata, 0));
 		size = (int)PyLong_AsLong(PyTuple_GetItem(parentfiledata, 1));
 		mtime = (int)PyLong_AsLong(PyTuple_GetItem(parentfiledata, 2));
 	} else {
 		has_meaningful_data = 0;
 		has_meaningful_mtime = 0;
 	}
 	if (has_meaningful_data) {
 		t->flags |= dirstate_flag_has_meaningful_data;
 		t->mode = mode;
 		t->size = size;
 	} else {
 		t->mode = 0;
 		t->size = 0;
 	}
 	if (has_meaningful_mtime) {
 		t->flags |= dirstate_flag_has_meaningful_mtime;
 		t->mtime = mtime;
 	} else {
 		t->mtime = 0;
 	}
 	return (PyObject *)t;
 }
 
 static void dirstate_item_dealloc(PyObject *o)
 {
 	PyObject_Del(o);
 }
 
 static inline bool dirstate_item_c_tracked(dirstateItemObject *self)
 {
 	return (self->flags & dirstate_flag_wc_tracked);
 }
 
 static inline bool dirstate_item_c_any_tracked(dirstateItemObject *self)
 {
 	const unsigned char mask = dirstate_flag_wc_tracked |
 	                           dirstate_flag_p1_tracked |
 	                           dirstate_flag_p2_info;
 	return (self->flags & mask);
 }
 
 static inline bool dirstate_item_c_added(dirstateItemObject *self)
 {
 	const unsigned char mask =
 	    (dirstate_flag_wc_tracked | dirstate_flag_p1_tracked |
 	     dirstate_flag_p2_info);
 	const unsigned char target = dirstate_flag_wc_tracked;
 	return (self->flags & mask) == target;
 }
 
 static inline bool dirstate_item_c_removed(dirstateItemObject *self)
 {
 	if (self->flags & dirstate_flag_wc_tracked) {
 		return false;
 	}
 	return (self->flags &
 	        (dirstate_flag_p1_tracked | dirstate_flag_p2_info));
 }
 
 static inline bool dirstate_item_c_merged(dirstateItemObject *self)
 {
 	return ((self->flags & dirstate_flag_wc_tracked) &&
 	        (self->flags & dirstate_flag_p1_tracked) &&
 	        (self->flags & dirstate_flag_p2_info));
 }
 
 static inline bool dirstate_item_c_from_p2(dirstateItemObject *self)
 {
 	return ((self->flags & dirstate_flag_wc_tracked) &&
 	        !(self->flags & dirstate_flag_p1_tracked) &&
 	        (self->flags & dirstate_flag_p2_info));
 }
 
 static inline char dirstate_item_c_v1_state(dirstateItemObject *self)
 {
 	if (dirstate_item_c_removed(self)) {
 		return 'r';
 	} else if (dirstate_item_c_merged(self)) {
 		return 'm';
 	} else if (dirstate_item_c_added(self)) {
 		return 'a';
 	} else {
 		return 'n';
 	}
 }
 
 static inline int dirstate_item_c_v1_mode(dirstateItemObject *self)
 {
 	if (self->flags & dirstate_flag_has_meaningful_data) {
 		return self->mode;
 	} else {
 		return 0;
 	}
 }
 
 static inline int dirstate_item_c_v1_size(dirstateItemObject *self)
 {
 	if (!(self->flags & dirstate_flag_wc_tracked) &&
 	    (self->flags & dirstate_flag_p2_info)) {
 		if (self->flags & dirstate_flag_p1_tracked) {
 			return dirstate_v1_nonnormal;
 		} else {
 			return dirstate_v1_from_p2;
 		}
 	} else if (dirstate_item_c_removed(self)) {
 		return 0;
 	} else if (self->flags & dirstate_flag_p2_info) {
 		return dirstate_v1_from_p2;
 	} else if (dirstate_item_c_added(self)) {
 		return dirstate_v1_nonnormal;
 	} else if (self->flags & dirstate_flag_has_meaningful_data) {
 		return self->size;
 	} else {
 		return dirstate_v1_nonnormal;
 	}
 }
 
 static inline int dirstate_item_c_v1_mtime(dirstateItemObject *self)
 {
 	if (dirstate_item_c_removed(self)) {
 		return 0;
 	} else if (!(self->flags & dirstate_flag_has_meaningful_mtime) ||
 	           !(self->flags & dirstate_flag_p1_tracked) ||
 	           !(self->flags & dirstate_flag_wc_tracked) ||
 	           (self->flags & dirstate_flag_p2_info)) {
 		return ambiguous_time;
 	} else {
 		return self->mtime;
 	}
 }
 
+static PyObject *dirstate_item_v2_data(dirstateItemObject *self)
+{
+	unsigned char flags = self->flags;
+	int mode = dirstate_item_c_v1_mode(self);
+	if ((mode & S_IXUSR) != 0) {
+		flags |= dirstate_flag_mode_exec_perm;
+	} else {
+		flags &= ~dirstate_flag_mode_exec_perm;
+	}
+	if (S_ISLNK(mode)) {
+		flags |= dirstate_flag_mode_is_symlink;
+	} else {
+		flags &= ~dirstate_flag_mode_is_symlink;
+	}
+	return Py_BuildValue("Bii", flags, self->size, self->mtime);
+};
+
 static PyObject *dirstate_item_v1_state(dirstateItemObject *self)
 {
 	char state = dirstate_item_c_v1_state(self);
 	return PyBytes_FromStringAndSize(&state, 1);
 };
 
 static PyObject *dirstate_item_v1_mode(dirstateItemObject *self)
 {
 	return PyInt_FromLong(dirstate_item_c_v1_mode(self));
 };
 
 static PyObject *dirstate_item_v1_size(dirstateItemObject *self)
 {
 	return PyInt_FromLong(dirstate_item_c_v1_size(self));
 };
 
 static PyObject *dirstate_item_v1_mtime(dirstateItemObject *self)
 {
 	return PyInt_FromLong(dirstate_item_c_v1_mtime(self));
 };
 
 static PyObject *dirstate_item_need_delay(dirstateItemObject *self,
                                           PyObject *value)
 {
 	long now;
 	if (!pylong_to_long(value, &now)) {
 		return NULL;
 	}
 	if (dirstate_item_c_v1_state(self) == 'n' &&
 	    dirstate_item_c_v1_mtime(self) == now) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 /* This will never change since it's bound to V1
  */
 static inline dirstateItemObject *
 dirstate_item_from_v1_data(char state, int mode, int size, int mtime)
 {
 	dirstateItemObject *t =
 	    PyObject_New(dirstateItemObject, &dirstateItemType);
 	if (!t) {
 		return NULL;
 	}
 	t->flags = 0;
 	t->mode = 0;
 	t->size = 0;
 	t->mtime = 0;
 
 	if (state == 'm') {
 		t->flags = (dirstate_flag_wc_tracked |
 		            dirstate_flag_p1_tracked | dirstate_flag_p2_info);
 	} else if (state == 'a') {
 		t->flags = dirstate_flag_wc_tracked;
 	} else if (state == 'r') {
 		if (size == dirstate_v1_nonnormal) {
 			t->flags =
 			    dirstate_flag_p1_tracked | dirstate_flag_p2_info;
 		} else if (size == dirstate_v1_from_p2) {
 			t->flags = dirstate_flag_p2_info;
 		} else {
 			t->flags = dirstate_flag_p1_tracked;
 		}
 	} else if (state == 'n') {
 		if (size == dirstate_v1_from_p2) {
 			t->flags =
 			    dirstate_flag_wc_tracked | dirstate_flag_p2_info;
 		} else if (size == dirstate_v1_nonnormal) {
 			t->flags =
 			    dirstate_flag_wc_tracked | dirstate_flag_p1_tracked;
 		} else if (mtime == ambiguous_time) {
 			t->flags = (dirstate_flag_wc_tracked |
 			            dirstate_flag_p1_tracked |
 			            dirstate_flag_has_meaningful_data);
 			t->mode = mode;
 			t->size = size;
 		} else {
 			t->flags = (dirstate_flag_wc_tracked |
 			            dirstate_flag_p1_tracked |
 			            dirstate_flag_has_meaningful_data |
 			            dirstate_flag_has_meaningful_mtime);
 			t->mode = mode;
 			t->size = size;
 			t->mtime = mtime;
 		}
 	} else {
 		PyErr_Format(PyExc_RuntimeError,
 		             "unknown state: `%c` (%d, %d, %d)", state, mode,
 		             size, mtime, NULL);
 		Py_DECREF(t);
 		return NULL;
 	}
 
 	return t;
 }
 
 /* This will never change since it's bound to V1, unlike `dirstate_item_new` */
 static PyObject *dirstate_item_from_v1_meth(PyTypeObject *subtype,
                                             PyObject *args)
 {
 	/* We do all the initialization here and not a tp_init function because
 	 * dirstate_item is immutable. */
 	char state;
 	int size, mode, mtime;
 	if (!PyArg_ParseTuple(args, "ciii", &state, &mode, &size, &mtime)) {
 		return NULL;
 	}
 	return (PyObject *)dirstate_item_from_v1_data(state, mode, size, mtime);
 };
 
 static PyObject *dirstate_item_from_v2_meth(PyTypeObject *subtype,
                                             PyObject *args)
 {
 	dirstateItemObject *t =
 	    PyObject_New(dirstateItemObject, &dirstateItemType);
 	if (!t) {
 		return NULL;
 	}
 	if (!PyArg_ParseTuple(args, "bii", &t->flags, &t->size, &t->mtime)) {
 		return NULL;
 	}
 	t->mode = 0;
 	if (t->flags & dirstate_flag_has_meaningful_data) {
 		if (t->flags & dirstate_flag_mode_exec_perm) {
 			t->mode = 0755;
 		} else {
 			t->mode = 0644;
 		}
 		if (t->flags & dirstate_flag_mode_is_symlink) {
 			t->mode |= S_IFLNK;
 		} else {
 			t->mode |= S_IFREG;
 		}
 	}
 	return (PyObject *)t;
 };
 
 /* This means the next status call will have to actually check its content
    to make sure it is correct. */
 static PyObject *dirstate_item_set_possibly_dirty(dirstateItemObject *self)
 {
 	self->flags &= ~dirstate_flag_has_meaningful_mtime;
 	Py_RETURN_NONE;
 }
 
 /* See docstring of the python implementation for details */
 static PyObject *dirstate_item_set_clean(dirstateItemObject *self,
                                          PyObject *args)
 {
 	int size, mode, mtime;
 	if (!PyArg_ParseTuple(args, "iii", &mode, &size, &mtime)) {
 		return NULL;
 	}
 	self->flags = dirstate_flag_wc_tracked | dirstate_flag_p1_tracked |
 	              dirstate_flag_has_meaningful_data |
 	              dirstate_flag_has_meaningful_mtime;
 	self->mode = mode;
 	self->size = size;
 	self->mtime = mtime;
 	Py_RETURN_NONE;
 }
 
 static PyObject *dirstate_item_set_tracked(dirstateItemObject *self)
 {
 	self->flags |= dirstate_flag_wc_tracked;
 	self->flags &= ~dirstate_flag_has_meaningful_mtime;
 	Py_RETURN_NONE;
 }
 
 static PyObject *dirstate_item_set_untracked(dirstateItemObject *self)
 {
 	self->flags &= ~dirstate_flag_wc_tracked;
 	self->mode = 0;
 	self->mtime = 0;
 	self->size = 0;
 	Py_RETURN_NONE;
 }
 
 static PyObject *dirstate_item_drop_merge_data(dirstateItemObject *self)
 {
 	if (self->flags & dirstate_flag_p2_info) {
 		self->flags &= ~(dirstate_flag_p2_info |
 		                 dirstate_flag_has_meaningful_data |
 		                 dirstate_flag_has_meaningful_mtime);
 		self->mode = 0;
 		self->mtime = 0;
 		self->size = 0;
 	}
 	Py_RETURN_NONE;
 }
 static PyMethodDef dirstate_item_methods[] = {
+    {"v2_data", (PyCFunction)dirstate_item_v2_data, METH_NOARGS,
+     "return data suitable for v2 serialization"},
     {"v1_state", (PyCFunction)dirstate_item_v1_state, METH_NOARGS,
      "return a \"state\" suitable for v1 serialization"},
     {"v1_mode", (PyCFunction)dirstate_item_v1_mode, METH_NOARGS,
      "return a \"mode\" suitable for v1 serialization"},
     {"v1_size", (PyCFunction)dirstate_item_v1_size, METH_NOARGS,
      "return a \"size\" suitable for v1 serialization"},
     {"v1_mtime", (PyCFunction)dirstate_item_v1_mtime, METH_NOARGS,
      "return a \"mtime\" suitable for v1 serialization"},
     {"need_delay", (PyCFunction)dirstate_item_need_delay, METH_O,
      "True if the stored mtime would be ambiguous with the current time"},
     {"from_v1_data", (PyCFunction)dirstate_item_from_v1_meth,
      METH_VARARGS | METH_CLASS, "build a new DirstateItem object from V1 data"},
     {"from_v2_data", (PyCFunction)dirstate_item_from_v2_meth,
      METH_VARARGS | METH_CLASS, "build a new DirstateItem object from V2 data"},
     {"set_possibly_dirty", (PyCFunction)dirstate_item_set_possibly_dirty,
      METH_NOARGS, "mark a file as \"possibly dirty\""},
     {"set_clean", (PyCFunction)dirstate_item_set_clean, METH_VARARGS,
      "mark a file as \"clean\""},
     {"set_tracked", (PyCFunction)dirstate_item_set_tracked, METH_NOARGS,
      "mark a file as \"tracked\""},
     {"set_untracked", (PyCFunction)dirstate_item_set_untracked, METH_NOARGS,
      "mark a file as \"untracked\""},
     {"drop_merge_data", (PyCFunction)dirstate_item_drop_merge_data, METH_NOARGS,
      "remove all \"merge-only\" from a DirstateItem"},
     {NULL} /* Sentinel */
 };
 
 static PyObject *dirstate_item_get_mode(dirstateItemObject *self)
 {
 	return PyInt_FromLong(dirstate_item_c_v1_mode(self));
 };
 
 static PyObject *dirstate_item_get_size(dirstateItemObject *self)
 {
 	return PyInt_FromLong(dirstate_item_c_v1_size(self));
 };
 
 static PyObject *dirstate_item_get_mtime(dirstateItemObject *self)
 {
 	return PyInt_FromLong(dirstate_item_c_v1_mtime(self));
 };
 
 static PyObject *dirstate_item_get_state(dirstateItemObject *self)
 {
 	char state = dirstate_item_c_v1_state(self);
 	return PyBytes_FromStringAndSize(&state, 1);
 };
 
 static PyObject *dirstate_item_get_tracked(dirstateItemObject *self)
 {
 	if (dirstate_item_c_tracked(self)) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 static PyObject *dirstate_item_get_p1_tracked(dirstateItemObject *self)
 {
 	if (self->flags & dirstate_flag_p1_tracked) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyObject *dirstate_item_get_added(dirstateItemObject *self)
 {
 	if (dirstate_item_c_added(self)) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyObject *dirstate_item_get_p2_info(dirstateItemObject *self)
 {
 	if (self->flags & dirstate_flag_wc_tracked &&
 	    self->flags & dirstate_flag_p2_info) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyObject *dirstate_item_get_merged(dirstateItemObject *self)
 {
 	if (dirstate_item_c_merged(self)) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyObject *dirstate_item_get_from_p2(dirstateItemObject *self)
 {
 	if (dirstate_item_c_from_p2(self)) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyObject *dirstate_item_get_maybe_clean(dirstateItemObject *self)
 {
 	if (!(self->flags & dirstate_flag_wc_tracked)) {
 		Py_RETURN_FALSE;
 	} else if (!(self->flags & dirstate_flag_p1_tracked)) {
 		Py_RETURN_FALSE;
 	} else if (self->flags & dirstate_flag_p2_info) {
 		Py_RETURN_FALSE;
 	} else {
 		Py_RETURN_TRUE;
 	}
 };
 
 static PyObject *dirstate_item_get_any_tracked(dirstateItemObject *self)
 {
 	if (dirstate_item_c_any_tracked(self)) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyObject *dirstate_item_get_removed(dirstateItemObject *self)
 {
 	if (dirstate_item_c_removed(self)) {
 		Py_RETURN_TRUE;
 	} else {
 		Py_RETURN_FALSE;
 	}
 };
 
 static PyGetSetDef dirstate_item_getset[] = {
     {"mode", (getter)dirstate_item_get_mode, NULL, "mode", NULL},
     {"size", (getter)dirstate_item_get_size, NULL, "size", NULL},
     {"mtime", (getter)dirstate_item_get_mtime, NULL, "mtime", NULL},
     {"state", (getter)dirstate_item_get_state, NULL, "state", NULL},
     {"tracked", (getter)dirstate_item_get_tracked, NULL, "tracked", NULL},
     {"p1_tracked", (getter)dirstate_item_get_p1_tracked, NULL, "p1_tracked",
      NULL},
     {"added", (getter)dirstate_item_get_added, NULL, "added", NULL},
     {"p2_info", (getter)dirstate_item_get_p2_info, NULL, "p2_info", NULL},
     {"merged", (getter)dirstate_item_get_merged, NULL, "merged", NULL},
     {"from_p2", (getter)dirstate_item_get_from_p2, NULL, "from_p2", NULL},
     {"maybe_clean", (getter)dirstate_item_get_maybe_clean, NULL, "maybe_clean",
      NULL},
     {"any_tracked", (getter)dirstate_item_get_any_tracked, NULL, "any_tracked",
      NULL},
     {"removed", (getter)dirstate_item_get_removed, NULL, "removed", NULL},
     {NULL} /* Sentinel */
 };
 
 PyTypeObject dirstateItemType = {
     PyVarObject_HEAD_INIT(NULL, 0)     /* header */
     "dirstate_tuple",                  /* tp_name */
     sizeof(dirstateItemObject),        /* tp_basicsize */
     0,                                 /* tp_itemsize */
     (destructor)dirstate_item_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 */
     Py_TPFLAGS_DEFAULT,                /* tp_flags */
     "dirstate tuple",                  /* tp_doc */
     0,                                 /* tp_traverse */
     0,                                 /* tp_clear */
     0,                                 /* tp_richcompare */
     0,                                 /* tp_weaklistoffset */
     0,                                 /* tp_iter */
     0,                                 /* tp_iternext */
     dirstate_item_methods,             /* tp_methods */
     0,                                 /* tp_members */
     dirstate_item_getset,              /* tp_getset */
     0,                                 /* tp_base */
     0,                                 /* tp_dict */
     0,                                 /* tp_descr_get */
     0,                                 /* tp_descr_set */
     0,                                 /* tp_dictoffset */
     0,                                 /* tp_init */
     0,                                 /* tp_alloc */
     dirstate_item_new,                 /* tp_new */
 };
 
 static PyObject *parse_dirstate(PyObject *self, PyObject *args)
 {
 	PyObject *dmap, *cmap, *parents = NULL, *ret = NULL;
 	PyObject *fname = NULL, *cname = NULL, *entry = NULL;
 	char state, *cur, *str, *cpos;
 	int mode, size, mtime;
 	unsigned int flen, pos = 40;
 	Py_ssize_t len = 40;
 	Py_ssize_t readlen;
 
 	if (!PyArg_ParseTuple(
 	        args, PY23("O!O!s#:parse_dirstate", "O!O!y#:parse_dirstate"),
 	        &PyDict_Type, &dmap, &PyDict_Type, &cmap, &str, &readlen)) {
 		goto quit;
 	}
 
 	len = readlen;
 
 	/* read parents */
 	if (len < 40) {
 		PyErr_SetString(PyExc_ValueError,
 		                "too little data for parents");
 		goto quit;
 	}
 
 	parents = Py_BuildValue(PY23("s#s#", "y#y#"), str, (Py_ssize_t)20,
 	                        str + 20, (Py_ssize_t)20);
 	if (!parents) {
 		goto quit;
 	}
 
 	/* read filenames */
 	while (pos >= 40 && pos < len) {
 		if (pos + 17 > len) {
 			PyErr_SetString(PyExc_ValueError,
 			                "overflow in dirstate");
 			goto quit;
 		}
 		cur = str + pos;
 		/* unpack header */
 		state = *cur;
 		mode = getbe32(cur + 1);
 		size = getbe32(cur + 5);
 		mtime = getbe32(cur + 9);
 		flen = getbe32(cur + 13);
 		pos += 17;
 		cur += 17;
 		if (flen > len - pos) {
 			PyErr_SetString(PyExc_ValueError,
 			                "overflow in dirstate");
 			goto quit;
 		}
 
 		entry = (PyObject *)dirstate_item_from_v1_data(state, mode,
 		                                               size, mtime);
 		if (!entry)
 			goto quit;
 		cpos = memchr(cur, 0, flen);
 		if (cpos) {
 			fname = PyBytes_FromStringAndSize(cur, cpos - cur);
 			cname = PyBytes_FromStringAndSize(
 			    cpos + 1, flen - (cpos - cur) - 1);
 			if (!fname || !cname ||
 			    PyDict_SetItem(cmap, fname, cname) == -1 ||
 			    PyDict_SetItem(dmap, fname, entry) == -1) {
 				goto quit;
 			}
 			Py_DECREF(cname);
 		} else {
 			fname = PyBytes_FromStringAndSize(cur, flen);
 			if (!fname ||
 			    PyDict_SetItem(dmap, fname, entry) == -1) {
 				goto quit;
 			}
 		}
 		Py_DECREF(fname);
 		Py_DECREF(entry);
 		fname = cname = entry = NULL;
 		pos += flen;
 	}
 
 	ret = parents;
 	Py_INCREF(ret);
 quit:
 	Py_XDECREF(fname);
 	Py_XDECREF(cname);
 	Py_XDECREF(entry);
 	Py_XDECREF(parents);
 	return ret;
 }
 
 /*
  * Efficiently pack a dirstate object into its on-disk format.
  */
 static PyObject *pack_dirstate(PyObject *self, PyObject *args)
 {
 	PyObject *packobj = NULL;
 	PyObject *map, *copymap, *pl, *mtime_unset = NULL;
 	Py_ssize_t nbytes, pos, l;
 	PyObject *k, *v = NULL, *pn;
 	char *p, *s;
 	int now;
 
 	if (!PyArg_ParseTuple(args, "O!O!O!i:pack_dirstate", &PyDict_Type, &map,
 	                      &PyDict_Type, &copymap, &PyTuple_Type, &pl,
 	                      &now)) {
 		return NULL;
 	}
 
 	if (PyTuple_Size(pl) != 2) {
 		PyErr_SetString(PyExc_TypeError, "expected 2-element tuple");
 		return NULL;
 	}
 
 	/* Figure out how much we need to allocate. */
 	for (nbytes = 40, pos = 0; PyDict_Next(map, &pos, &k, &v);) {
 		PyObject *c;
 		if (!PyBytes_Check(k)) {
 			PyErr_SetString(PyExc_TypeError, "expected string key");
 			goto bail;
 		}
 		nbytes += PyBytes_GET_SIZE(k) + 17;
 		c = PyDict_GetItem(copymap, k);
 		if (c) {
 			if (!PyBytes_Check(c)) {
 				PyErr_SetString(PyExc_TypeError,
 				                "expected string key");
 				goto bail;
 			}
 			nbytes += PyBytes_GET_SIZE(c) + 1;
 		}
 	}
 
 	packobj = PyBytes_FromStringAndSize(NULL, nbytes);
 	if (packobj == NULL) {
 		goto bail;
 	}
 
 	p = PyBytes_AS_STRING(packobj);
 
 	pn = PyTuple_GET_ITEM(pl, 0);
 	if (PyBytes_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
 		PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
 		goto bail;
 	}
 	memcpy(p, s, l);
 	p += 20;
 	pn = PyTuple_GET_ITEM(pl, 1);
 	if (PyBytes_AsStringAndSize(pn, &s, &l) == -1 || l != 20) {
 		PyErr_SetString(PyExc_TypeError, "expected a 20-byte hash");
 		goto bail;
 	}
 	memcpy(p, s, l);
 	p += 20;
 
 	for (pos = 0; PyDict_Next(map, &pos, &k, &v);) {
 		dirstateItemObject *tuple;
 		char state;
 		int mode, size, mtime;
 		Py_ssize_t len, l;
 		PyObject *o;
 		char *t;
 
 		if (!dirstate_tuple_check(v)) {
 			PyErr_SetString(PyExc_TypeError,
 			                "expected a dirstate tuple");
 			goto bail;
 		}
 		tuple = (dirstateItemObject *)v;
 
 		state = dirstate_item_c_v1_state(tuple);
 		mode = dirstate_item_c_v1_mode(tuple);
 		size = dirstate_item_c_v1_size(tuple);
 		mtime = dirstate_item_c_v1_mtime(tuple);
 		if (state == 'n' && mtime == now) {
 			/* See pure/parsers.py:pack_dirstate for why we do
 			 * this. */
 			mtime = -1;
 			mtime_unset = (PyObject *)dirstate_item_from_v1_data(
 			    state, mode, size, mtime);
 			if (!mtime_unset) {
 				goto bail;
 			}
 			if (PyDict_SetItem(map, k, mtime_unset) == -1) {
 				goto bail;
 			}
 			Py_DECREF(mtime_unset);
 			mtime_unset = NULL;
 		}
 		*p++ = state;
 		putbe32((uint32_t)mode, p);
 		putbe32((uint32_t)size, p + 4);
 		putbe32((uint32_t)mtime, p + 8);
 		t = p + 12;
 		p += 16;
 		len = PyBytes_GET_SIZE(k);
 		memcpy(p, PyBytes_AS_STRING(k), len);
 		p += len;
 		o = PyDict_GetItem(copymap, k);
 		if (o) {
 			*p++ = '\0';
 			l = PyBytes_GET_SIZE(o);
 			memcpy(p, PyBytes_AS_STRING(o), l);
 			p += l;
 			len += l + 1;
 		}
 		putbe32((uint32_t)len, t);
 	}
 
 	pos = p - PyBytes_AS_STRING(packobj);
 	if (pos != nbytes) {
 		PyErr_Format(PyExc_SystemError, "bad dirstate size: %ld != %ld",
 		             (long)pos, (long)nbytes);
 		goto bail;
 	}
 
 	return packobj;
 bail:
 	Py_XDECREF(mtime_unset);
 	Py_XDECREF(packobj);
 	Py_XDECREF(v);
 	return NULL;
 }
 
 #define BUMPED_FIX 1
 #define USING_SHA_256 2
 #define FM1_HEADER_SIZE (4 + 8 + 2 + 2 + 1 + 1 + 1)
 
 static PyObject *readshas(const char *source, unsigned char num,
                           Py_ssize_t hashwidth)
 {
 	int i;
 	PyObject *list = PyTuple_New(num);
 	if (list == NULL) {
 		return NULL;
 	}
 	for (i = 0; i < num; i++) {
 		PyObject *hash = PyBytes_FromStringAndSize(source, hashwidth);
 		if (hash == NULL) {
 			Py_DECREF(list);
 			return NULL;
 		}
 		PyTuple_SET_ITEM(list, i, hash);
 		source += hashwidth;
 	}
 	return list;
 }
 
 static PyObject *fm1readmarker(const char *databegin, const char *dataend,
                                uint32_t *msize)
 {
 	const char *data = databegin;
 	const char *meta;
 
 	double mtime;
 	int16_t tz;
 	uint16_t flags;
 	unsigned char nsuccs, nparents, nmetadata;
 	Py_ssize_t hashwidth = 20;
 
 	PyObject *prec = NULL, *parents = NULL, *succs = NULL;
 	PyObject *metadata = NULL, *ret = NULL;
 	int i;
 
 	if (data + FM1_HEADER_SIZE > dataend) {
 		goto overflow;
 	}
 
 	*msize = getbe32(data);
 	data += 4;
 	mtime = getbefloat64(data);
 	data += 8;
 	tz = getbeint16(data);
 	data += 2;
 	flags = getbeuint16(data);
 	data += 2;
 
 	if (flags & USING_SHA_256) {
 		hashwidth = 32;
 	}
 
 	nsuccs = (unsigned char)(*data++);
 	nparents = (unsigned char)(*data++);
 	nmetadata = (unsigned char)(*data++);
 
 	if (databegin + *msize > dataend) {
 		goto overflow;
 	}
 	dataend = databegin + *msize; /* narrow down to marker size */
 
 	if (data + hashwidth > dataend) {
 		goto overflow;
 	}
 	prec = PyBytes_FromStringAndSize(data, hashwidth);
 	data += hashwidth;
 	if (prec == NULL) {
 		goto bail;
 	}
 
 	if (data + nsuccs * hashwidth > dataend) {
 		goto overflow;
 	}
 	succs = readshas(data, nsuccs, hashwidth);
 	if (succs == NULL) {
 		goto bail;
 	}
 	data += nsuccs * hashwidth;
 
 	if (nparents == 1 || nparents == 2) {
 		if (data + nparents * hashwidth > dataend) {
 			goto overflow;
 		}
 		parents = readshas(data, nparents, hashwidth);
 		if (parents == NULL) {
 			goto bail;
 		}
 		data += nparents * hashwidth;
 	} else {
 		parents = Py_None;
 		Py_INCREF(parents);
 	}
 
 	if (data + 2 * nmetadata > dataend) {
 		goto overflow;
 	}
 	meta = data + (2 * nmetadata);
 	metadata = PyTuple_New(nmetadata);
 	if (metadata == NULL) {
 		goto bail;
 	}
 	for (i = 0; i < nmetadata; i++) {
 		PyObject *tmp, *left = NULL, *right = NULL;
 		Py_ssize_t leftsize = (unsigned char)(*data++);
 		Py_ssize_t rightsize = (unsigned char)(*data++);
 		if (meta + leftsize + rightsize > dataend) {
 			goto overflow;
 		}
 		left = PyBytes_FromStringAndSize(meta, leftsize);
 		meta += leftsize;
 		right = PyBytes_FromStringAndSize(meta, rightsize);
 		meta += rightsize;
 		tmp = PyTuple_New(2);
 		if (!left || !right || !tmp) {
 			Py_XDECREF(left);
 			Py_XDECREF(right);
 			Py_XDECREF(tmp);
 			goto bail;
 		}
 		PyTuple_SET_ITEM(tmp, 0, left);
 		PyTuple_SET_ITEM(tmp, 1, right);
 		PyTuple_SET_ITEM(metadata, i, tmp);
 	}
 	ret = Py_BuildValue("(OOHO(di)O)", prec, succs, flags, metadata, mtime,
 	                    (int)tz * 60, parents);
 	goto bail; /* return successfully */
 
 overflow:
 	PyErr_SetString(PyExc_ValueError, "overflow in obsstore");
 bail:
 	Py_XDECREF(prec);
 	Py_XDECREF(succs);
 	Py_XDECREF(metadata);
 	Py_XDECREF(parents);
 	return ret;
 }
 
 static PyObject *fm1readmarkers(PyObject *self, PyObject *args)
 {
 	const char *data, *dataend;
 	Py_ssize_t datalen, offset, stop;
 	PyObject *markers = NULL;
 
 	if (!PyArg_ParseTuple(args, PY23("s#nn", "y#nn"), &data, &datalen,
 	                      &offset, &stop)) {
 		return NULL;
 	}
 	if (offset < 0) {
 		PyErr_SetString(PyExc_ValueError,
 		                "invalid negative offset in fm1readmarkers");
 		return NULL;
 	}
 	if (stop > datalen) {
 		PyErr_SetString(
 		    PyExc_ValueError,
 		    "stop longer than data length in fm1readmarkers");
 		return NULL;
 	}
 	dataend = data + datalen;
 	data += offset;
 	markers = PyList_New(0);
 	if (!markers) {
 		return NULL;
 	}
 	while (offset < stop) {
 		uint32_t msize;
 		int error;
 		PyObject *record = fm1readmarker(data, dataend, &msize);
 		if (!record) {
 			goto bail;
 		}
 		error = PyList_Append(markers, record);
 		Py_DECREF(record);
 		if (error) {
 			goto bail;
 		}
 		data += msize;
 		offset += msize;
 	}
 	return markers;
 bail:
 	Py_DECREF(markers);
 	return NULL;
 }
 
 static char parsers_doc[] = "Efficient content parsing.";
 
 PyObject *encodedir(PyObject *self, PyObject *args);
 PyObject *pathencode(PyObject *self, PyObject *args);
 PyObject *lowerencode(PyObject *self, PyObject *args);
 PyObject *parse_index2(PyObject *self, PyObject *args, PyObject *kwargs);
 
 static PyMethodDef methods[] = {
     {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
     {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
     {"parse_index2", (PyCFunction)parse_index2, METH_VARARGS | METH_KEYWORDS,
      "parse a revlog index\n"},
     {"isasciistr", isasciistr, METH_VARARGS, "check if an ASCII string\n"},
     {"asciilower", asciilower, METH_VARARGS, "lowercase an ASCII string\n"},
     {"asciiupper", asciiupper, METH_VARARGS, "uppercase an ASCII string\n"},
     {"dict_new_presized", dict_new_presized, METH_VARARGS,
      "construct a dict with an expected size\n"},
     {"make_file_foldmap", make_file_foldmap, METH_VARARGS,
      "make file foldmap\n"},
     {"jsonescapeu8fast", jsonescapeu8fast, METH_VARARGS,
      "escape a UTF-8 byte string to JSON (fast path)\n"},
     {"encodedir", encodedir, METH_VARARGS, "encodedir a path\n"},
     {"pathencode", pathencode, METH_VARARGS, "fncache-encode a path\n"},
     {"lowerencode", lowerencode, METH_VARARGS, "lower-encode a path\n"},
     {"fm1readmarkers", fm1readmarkers, METH_VARARGS,
      "parse v1 obsolete markers\n"},
     {NULL, NULL}};
 
 void dirs_module_init(PyObject *mod);
 void manifest_module_init(PyObject *mod);
 void revlog_module_init(PyObject *mod);
 
 static const int version = 20;
 
 static void module_init(PyObject *mod)
 {
 	PyModule_AddIntConstant(mod, "version", version);
 
 	/* This module constant has two purposes.  First, it lets us unit test
 	 * the ImportError raised without hard-coding any error text.  This
 	 * means we can change the text in the future without breaking tests,
 	 * even across changesets without a recompile.  Second, its presence
 	 * can be used to determine whether the version-checking logic is
 	 * present, which also helps in testing across changesets without a
 	 * recompile.  Note that this means the pure-Python version of parsers
 	 * should not have this module constant. */
 	PyModule_AddStringConstant(mod, "versionerrortext", versionerrortext);
 
 	dirs_module_init(mod);
 	manifest_module_init(mod);
 	revlog_module_init(mod);
 
 	if (PyType_Ready(&dirstateItemType) < 0) {
 		return;
 	}
 	Py_INCREF(&dirstateItemType);
 	PyModule_AddObject(mod, "DirstateItem", (PyObject *)&dirstateItemType);
 }
 
 static int check_python_version(void)
 {
 	PyObject *sys = PyImport_ImportModule("sys"), *ver;
 	long hexversion;
 	if (!sys) {
 		return -1;
 	}
 	ver = PyObject_GetAttrString(sys, "hexversion");
 	Py_DECREF(sys);
 	if (!ver) {
 		return -1;
 	}
 	hexversion = PyInt_AsLong(ver);
 	Py_DECREF(ver);
 	/* sys.hexversion is a 32-bit number by default, so the -1 case
 	 * should only occur in unusual circumstances (e.g. if sys.hexversion
 	 * is manually set to an invalid value). */
 	if ((hexversion == -1) || (hexversion >> 16 != PY_VERSION_HEX >> 16)) {
 		PyErr_Format(PyExc_ImportError,
 		             "%s: The Mercurial extension "
 		             "modules were compiled with Python " PY_VERSION
 		             ", but "
 		             "Mercurial is currently using Python with "
 		             "sys.hexversion=%ld: "
 		             "Python %s\n at: %s",
 		             versionerrortext, hexversion, Py_GetVersion(),
 		             Py_GetProgramFullPath());
 		return -1;
 	}
 	return 0;
 }
 
 #ifdef IS_PY3K
 static struct PyModuleDef parsers_module = {PyModuleDef_HEAD_INIT, "parsers",
                                             parsers_doc, -1, methods};
 
 PyMODINIT_FUNC PyInit_parsers(void)
 {
 	PyObject *mod;
 
 	if (check_python_version() == -1)
 		return NULL;
 	mod = PyModule_Create(&parsers_module);
 	module_init(mod);
 	return mod;
 }
 #else
 PyMODINIT_FUNC initparsers(void)
 {
 	PyObject *mod;
 
 	if (check_python_version() == -1) {
 		return;
 	}
 	mod = Py_InitModule3("parsers", methods, parsers_doc);
 	module_init(mod);
 }
 #endif

mercurial/dirstateutils/v2.py

@@ -1,118 +1,411
 # v2.py - Pure-Python implementation of the dirstate-v2 file format
 #
 # Copyright Mercurial Contributors
 #
 # 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 struct
 
-from .. import policy
+from ..thirdparty import attr
+from .. import error, policy
 
 parsers = policy.importmod('parsers')
 
 
 # Must match the constant of the same name in
 # `rust/hg-core/src/dirstate_tree/on_disk.rs`
 TREE_METADATA_SIZE = 44
 NODE_SIZE = 43
 
 
 # Must match the `TreeMetadata` Rust struct in
 # `rust/hg-core/src/dirstate_tree/on_disk.rs`. See doc-comments there.
 #
 # * 4 bytes: start offset of root nodes
 # * 4 bytes: number of root nodes
 # * 4 bytes: total number of nodes in the tree that have an entry
 # * 4 bytes: total number of nodes in the tree that have a copy source
 # * 4 bytes: number of bytes in the data file that are not used anymore
 # * 4 bytes: unused
 # * 20 bytes: SHA-1 hash of ignore patterns
 TREE_METADATA = struct.Struct('>LLLLL4s20s')
 
 
 # Must match the `Node` Rust struct in
 # `rust/hg-core/src/dirstate_tree/on_disk.rs`. See doc-comments there.
 #
 # * 4 bytes: start offset of full path
 # * 2 bytes: length of the full path
 # * 2 bytes: length within the full path before its "base name"
 # * 4 bytes: start offset of the copy source if any, or zero for no copy source
 # * 2 bytes: length of the copy source if any, or unused
 # * 4 bytes: start offset of child nodes
 # * 4 bytes: number of child nodes
 # * 4 bytes: number of descendant nodes that have an entry
 # * 4 bytes: number of descendant nodes that have a "tracked" state
 # * 1 byte: flags
 # * 4 bytes: expected size
 # * 4 bytes: mtime seconds
 # * 4 bytes: mtime nanoseconds
 NODE = struct.Struct('>LHHLHLLLLBlll')
 
 
 assert TREE_METADATA_SIZE == TREE_METADATA.size
 assert NODE_SIZE == NODE.size
 
 
 def parse_dirstate(map, copy_map, data, tree_metadata):
     """parse a full v2-dirstate from a binary data into dictionnaries:
 
     - map: a {path: entry} mapping that will be filled
     - copy_map: a {path: copy-source} mapping that will be filled
     - data: a binary blob contains v2 nodes data
     - tree_metadata:: a binary blob of the top level node (from the docket)
     """
     (
         root_nodes_start,
         root_nodes_len,
         _nodes_with_entry_count,
         _nodes_with_copy_source_count,
         _unreachable_bytes,
         _unused,
         _ignore_patterns_hash,
     ) = TREE_METADATA.unpack(tree_metadata)
     parse_nodes(map, copy_map, data, root_nodes_start, root_nodes_len)
 
 
 def parse_nodes(map, copy_map, data, start, len):
     """parse <len> nodes from <data> starting at offset <start>
 
     This is used by parse_dirstate to recursively fill `map` and `copy_map`.
     """
     for i in range(len):
         node_start = start + NODE_SIZE * i
         node_bytes = slice_with_len(data, node_start, NODE_SIZE)
         (
             path_start,
             path_len,
             _basename_start,
             copy_source_start,
             copy_source_len,
             children_start,
             children_count,
             _descendants_with_entry_count,
             _tracked_descendants_count,
             flags,
             size,
             mtime_s,
             _mtime_ns,
         ) = NODE.unpack(node_bytes)
 
         # Parse child nodes of this node recursively
         parse_nodes(map, copy_map, data, children_start, children_count)
 
         item = parsers.DirstateItem.from_v2_data(flags, size, mtime_s)
         if not item.any_tracked:
             continue
         path = slice_with_len(data, path_start, path_len)
         map[path] = item
         if copy_source_start:
             copy_map[path] = slice_with_len(
                 data, copy_source_start, copy_source_len
             )
 
 
 def slice_with_len(data, start, len):
     return data[start : start + len]
+
+
+@attr.s
+class Node(object):
+    path = attr.ib()
+    entry = attr.ib()
+    parent = attr.ib(default=None)
+    children_count = attr.ib(default=0)
+    children_offset = attr.ib(default=0)
+    descendants_with_entry = attr.ib(default=0)
+    tracked_descendants = attr.ib(default=0)
+
+    def pack(self, copy_map, paths_offset):
+        path = self.path
+        copy = copy_map.get(path)
+        entry = self.entry
+
+        path_start = paths_offset
+        path_len = len(path)
+        basename_start = path.rfind(b'/') + 1  # 0 if rfind returns -1
+        if copy is not None:
+            copy_source_start = paths_offset + len(path)
+            copy_source_len = len(copy)
+        else:
+            copy_source_start = 0
+            copy_source_len = 0
+        if entry is not None:
+            flags, size, mtime_s = entry.v2_data()
+            mtime_ns = 0
+        else:
+            # There are no mtime-cached directories in the Python implementation
+            flags = 0
+            mode = 0
+            size = 0
+            mtime_s = 0
+            mtime_ns = 0
+        return NODE.pack(
+            path_start,
+            path_len,
+            basename_start,
+            copy_source_start,
+            copy_source_len,
+            self.children_offset,
+            self.children_count,
+            self.descendants_with_entry,
+            self.tracked_descendants,
+            flags,
+            size,
+            mtime_s,
+            mtime_ns,
+        )
+
+
+def pack_dirstate(map, copy_map, now):
+    """
+    Pack `map` and `copy_map` into the dirstate v2 binary format and return
+    the bytearray.
+    `now` is a timestamp of the current filesystem time used to detect race
+    conditions in writing the dirstate to disk, see inline comment.
+
+    The on-disk format expects a tree-like structure where the leaves are
+    written first (and sorted per-directory), going up levels until the root
+    node and writing that one to the docket. See more details on the on-disk
+    format in `mercurial/helptext/internals/dirstate-v2`.
+
+    Since both `map` and `copy_map` are flat dicts we need to figure out the
+    hierarchy. This algorithm does so without having to build the entire tree
+    in-memory: it only keeps the minimum number of nodes around to satisfy the
+    format.
+
+    # Algorithm explanation
+
+    This explanation does not talk about the different counters for tracked
+    descendents and storing the copies, but that work is pretty simple once this
+    algorithm is in place.
+
+    ## Building a subtree
+
+    First, sort `map`: this makes it so the leaves of the tree are contiguous
+    per directory (i.e. a/b/c and a/b/d will be next to each other in the list),
+    and enables us to use the ordering of folders to have a "cursor" of the
+    current folder we're in without ever going twice in the same branch of the
+    tree. The cursor is a node that remembers its parent and any information
+    relevant to the format (see the `Node` class), building the relevant part
+    of the tree lazily.
+    Then, for each file in `map`, move the cursor into the tree to the
+    corresponding folder of the file: for example, if the very first file
+    is "a/b/c", we start from `Node[""]`, create `Node["a"]` which points to
+    its parent `Node[""]`, then create `Node["a/b"]`, which points to its parent
+    `Node["a"]`. These nodes are kept around in a stack.
+    If the next file in `map` is in the same subtree ("a/b/d" or "a/b/e/f"), we
+    add it to the stack and keep looping with the same logic of creating the
+    tree nodes as needed. If however the next file in `map` is *not* in the same
+    subtree ("a/other", if we're still in the "a/b" folder), then we know that
+    the subtree we're in is complete.
+
+    ## Writing the subtree
+
+    We have the entire subtree in the stack, so we start writing it to disk
+    folder by folder. The way we write a folder is to pop the stack into a list
+    until the folder changes, revert this list of direct children (to satisfy
+    the format requirement that children be sorted). This process repeats until
+    we hit the "other" subtree.
+
+    An example:
+        a
+        dir1/b
+        dir1/c
+        dir2/dir3/d
+        dir2/dir3/e
+        dir2/f
+
+    Would have us:
+        - add to the stack until "dir2/dir3/e"
+        - realize that "dir2/f" is in a different subtree
+        - pop "dir2/dir3/e", "dir2/dir3/d", reverse them so they're sorted and
+          pack them since the next entry is "dir2/dir3"
+        - go back up to "dir2"
+        - add "dir2/f" to the stack
+        - realize we're done with the map
+        - pop "dir2/f", "dir2/dir3" from the stack, reverse and pack them
+        - go up to the root node, do the same to write "a", "dir1" and "dir2" in
+          that order
+
+    ## Special case for the root node
+
+    The root node is not serialized in the format, but its information is
+    written to the docket. Again, see more details on the on-disk format in
+    `mercurial/helptext/internals/dirstate-v2`.
+    """
+    now = int(now)
+    data = bytearray()
+    root_nodes_start = 0
+    root_nodes_len = 0
+    nodes_with_entry_count = 0
+    nodes_with_copy_source_count = 0
+    # Will always be 0 since this implementation always re-writes everything
+    # to disk
+    unreachable_bytes = 0
+    unused = b'\x00' * 4
+    # This is an optimization that's only useful for the Rust implementation
+    ignore_patterns_hash = b'\x00' * 20
+
+    if len(map) == 0:
+        tree_metadata = TREE_METADATA.pack(
+            root_nodes_start,
+            root_nodes_len,
+            nodes_with_entry_count,
+            nodes_with_copy_source_count,
+            unreachable_bytes,
+            unused,
+            ignore_patterns_hash,
+        )
+        return data, tree_metadata
+
+    sorted_map = sorted(map.items(), key=lambda x: x[0])
+
+    # Use a stack to not have to only remember the nodes we currently need
+    # instead of building the entire tree in memory
+    stack = []
+    current_node = Node(b"", None)
+    stack.append(current_node)
+
+    for index, (path, entry) in enumerate(sorted_map, 1):
+        if entry.need_delay(now):
+            # The file was last modified "simultaneously" with the current
+            # write to dirstate (i.e. within the same second for file-
+            # systems with a granularity of 1 sec). This commonly happens
+            # for at least a couple of files on 'update'.
+            # The user could change the file without changing its size
+            # within the same second. Invalidate the file's mtime in
+            # dirstate, forcing future 'status' calls to compare the
+            # contents of the file if the size is the same. This prevents
+            # mistakenly treating such files as clean.
+            entry.set_possibly_dirty()
+        nodes_with_entry_count += 1
+        if path in copy_map:
+            nodes_with_copy_source_count += 1
+        current_folder = get_folder(path)
+        current_node = move_to_correct_node_in_tree(
+            current_folder, current_node, stack
+        )
+
+        current_node.children_count += 1
+        # Entries from `map` are never `None`
+        if entry.tracked:
+            current_node.tracked_descendants += 1
+        current_node.descendants_with_entry += 1
+        stack.append(Node(path, entry, current_node))
+
+        should_pack = True
+        next_path = None
+        if index < len(sorted_map):
+            # Determine if the next entry is in the same sub-tree, if so don't
+            # pack yet
+            next_path = sorted_map[index][0]
+            should_pack = not get_folder(next_path).startswith(current_folder)
+        if should_pack:
+            pack_directory_children(current_node, copy_map, data, stack)
+            while stack and current_node.path != b"":
+                # Go up the tree and write until we reach the folder of the next
+                # entry (if any, otherwise the root)
+                parent = current_node.parent
+                in_parent_folder_of_next_entry = next_path is not None and (
+                    get_folder(next_path).startswith(get_folder(stack[-1].path))
+                )
+                if parent is None or in_parent_folder_of_next_entry:
+                    break
+                pack_directory_children(parent, copy_map, data, stack)
+                current_node = parent
+
+    # Special case for the root node since we don't write it to disk, only its
+    # children to the docket
+    current_node = stack.pop()
+    assert current_node.path == b"", current_node.path
+    assert len(stack) == 0, len(stack)
+
+    tree_metadata = TREE_METADATA.pack(
+        current_node.children_offset,
+        current_node.children_count,
+        nodes_with_entry_count,
+        nodes_with_copy_source_count,
+        unreachable_bytes,
+        unused,
+        ignore_patterns_hash,
+    )
+
+    return data, tree_metadata
+
+
+def get_folder(path):
+    """
+    Return the folder of the path that's given, an empty string for root paths.
+    """
+    return path.rsplit(b'/', 1)[0] if b'/' in path else b''
+
+
+def move_to_correct_node_in_tree(target_folder, current_node, stack):
+    """
+    Move inside the dirstate node tree to the node corresponding to
+    `target_folder`, creating the missing nodes along the way if needed.
+    """
+    while target_folder != current_node.path:
+        if target_folder.startswith(current_node.path):
+            # We need to go down a folder
+            prefix = target_folder[len(current_node.path) :].lstrip(b'/')
+            subfolder_name = prefix.split(b'/', 1)[0]
+            if current_node.path:
+                subfolder_path = current_node.path + b'/' + subfolder_name
+            else:
+                subfolder_path = subfolder_name
+            next_node = stack[-1]
+            if next_node.path == target_folder:
+                # This folder is now a file and only contains removed entries
+                # merge with the last node
+                current_node = next_node
+            else:
+                current_node.children_count += 1
+                current_node = Node(subfolder_path, None, current_node)
+                stack.append(current_node)
+        else:
+            # We need to go up a folder
+            current_node = current_node.parent
+    return current_node
+
+
+def pack_directory_children(node, copy_map, data, stack):
+    """
+    Write the binary representation of the direct sorted children of `node` to
+    `data`
+    """
+    direct_children = []
+
+    while stack[-1].path != b"" and get_folder(stack[-1].path) == node.path:
+        direct_children.append(stack.pop())
+    if not direct_children:
+        raise error.ProgrammingError(b"no direct children for %r" % node.path)
+
+    # Reverse the stack to get the correct sorted order
+    direct_children.reverse()
+    packed_children = bytearray()
+    # Write the paths to `data`. Pack child nodes but don't write them yet
+    for child in direct_children:
+        packed = child.pack(copy_map=copy_map, paths_offset=len(data))
+        packed_children.extend(packed)
+        data.extend(child.path)
+        data.extend(copy_map.get(child.path, b""))
+        node.tracked_descendants += child.tracked_descendants
+        node.descendants_with_entry += child.descendants_with_entry
+    # Write the fixed-size child nodes all together
+    node.children_offset = len(data)
+    data.extend(packed_children)

mercurial/pure/parsers.py

@@ -1,770 +1,789
 # parsers.py - Python implementation of parsers.c
 #
 # Copyright 2009 Olivia Mackall <[e-mail]> and others
 #
 # 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 stat
 import struct
 import zlib
 
 from ..node import (
     nullrev,
     sha1nodeconstants,
 )
 from ..thirdparty import attr
 from .. import (
     error,
     pycompat,
     revlogutils,
     util,
 )
 
 from ..revlogutils import nodemap as nodemaputil
 from ..revlogutils import constants as revlog_constants
 
 stringio = pycompat.bytesio
 
 
 _pack = struct.pack
 _unpack = struct.unpack
 _compress = zlib.compress
 _decompress = zlib.decompress
 
 
 # a special value used internally for `size` if the file come from the other parent
 FROM_P2 = -2
 
 # a special value used internally for `size` if the file is modified/merged/added
 NONNORMAL = -1
 
 # a special value used internally for `time` if the time is ambigeous
 AMBIGUOUS_TIME = -1
 
 # Bits of the `flags` byte inside a node in the file format
 DIRSTATE_V2_WDIR_TRACKED = 1 << 0
 DIRSTATE_V2_P1_TRACKED = 1 << 1
 DIRSTATE_V2_P2_INFO = 1 << 2
 DIRSTATE_V2_HAS_MODE_AND_SIZE = 1 << 3
 DIRSTATE_V2_HAS_MTIME = 1 << 4
 DIRSTATE_V2_MODE_EXEC_PERM = 1 << 5
 DIRSTATE_V2_MODE_IS_SYMLINK = 1 << 6
 
 
 @attr.s(slots=True, init=False)
 class DirstateItem(object):
     """represent a dirstate entry
 
     It hold multiple attributes
 
     # about file tracking
     - wc_tracked: is the file tracked by the working copy
     - p1_tracked: is the file tracked in working copy first parent
     - p2_info: the file has been involved in some merge operation. Either
                because it was actually merged, or because the p2 version was
                ahead, or because some rename moved it there. In either case
                `hg status` will want it displayed as modified.
 
     # about the file state expected from p1 manifest:
     - mode: the file mode in p1
     - size: the file size in p1
 
     These value can be set to None, which mean we don't have a meaningful value
     to compare with. Either because we don't really care about them as there
     `status` is known without having to look at the disk or because we don't
     know these right now and a full comparison will be needed to find out if
     the file is clean.
 
     # about the file state on disk last time we saw it:
     - mtime: the last known clean mtime for the file.
 
     This value can be set to None if no cachable state exist. Either because we
     do not care (see previous section) or because we could not cache something
     yet.
     """
 
     _wc_tracked = attr.ib()
     _p1_tracked = attr.ib()
     _p2_info = attr.ib()
     _mode = attr.ib()
     _size = attr.ib()
     _mtime = attr.ib()
 
     def __init__(
         self,
         wc_tracked=False,
         p1_tracked=False,
         p2_info=False,
         has_meaningful_data=True,
         has_meaningful_mtime=True,
         parentfiledata=None,
     ):
         self._wc_tracked = wc_tracked
         self._p1_tracked = p1_tracked
         self._p2_info = p2_info
 
         self._mode = None
         self._size = None
         self._mtime = None
         if parentfiledata is None:
             has_meaningful_mtime = False
             has_meaningful_data = False
         if has_meaningful_data:
             self._mode = parentfiledata[0]
             self._size = parentfiledata[1]
         if has_meaningful_mtime:
             self._mtime = parentfiledata[2]
 
     @classmethod
     def from_v2_data(cls, flags, size, mtime):
         """Build a new DirstateItem object from V2 data"""
         has_mode_size = bool(flags & DIRSTATE_V2_HAS_MODE_AND_SIZE)
         mode = None
         if has_mode_size:
             assert stat.S_IXUSR == 0o100
             if flags & DIRSTATE_V2_MODE_EXEC_PERM:
                 mode = 0o755
             else:
                 mode = 0o644
             if flags & DIRSTATE_V2_MODE_IS_SYMLINK:
                 mode |= stat.S_IFLNK
             else:
                 mode |= stat.S_IFREG
         return cls(
             wc_tracked=bool(flags & DIRSTATE_V2_WDIR_TRACKED),
             p1_tracked=bool(flags & DIRSTATE_V2_P1_TRACKED),
             p2_info=bool(flags & DIRSTATE_V2_P2_INFO),
             has_meaningful_data=has_mode_size,
             has_meaningful_mtime=bool(flags & DIRSTATE_V2_HAS_MTIME),
             parentfiledata=(mode, size, mtime),
         )
 
     @classmethod
     def from_v1_data(cls, state, mode, size, mtime):
         """Build a new DirstateItem object from V1 data
 
         Since the dirstate-v1 format is frozen, the signature of this function
         is not expected to change, unlike the __init__ one.
         """
         if state == b'm':
             return cls(wc_tracked=True, p1_tracked=True, p2_info=True)
         elif state == b'a':
             return cls(wc_tracked=True)
         elif state == b'r':
             if size == NONNORMAL:
                 p1_tracked = True
                 p2_info = True
             elif size == FROM_P2:
                 p1_tracked = False
                 p2_info = True
             else:
                 p1_tracked = True
                 p2_info = False
             return cls(p1_tracked=p1_tracked, p2_info=p2_info)
         elif state == b'n':
             if size == FROM_P2:
                 return cls(wc_tracked=True, p2_info=True)
             elif size == NONNORMAL:
                 return cls(wc_tracked=True, p1_tracked=True)
             elif mtime == AMBIGUOUS_TIME:
                 return cls(
                     wc_tracked=True,
                     p1_tracked=True,
                     has_meaningful_mtime=False,
                     parentfiledata=(mode, size, 42),
                 )
             else:
                 return cls(
                     wc_tracked=True,
                     p1_tracked=True,
                     parentfiledata=(mode, size, mtime),
                 )
         else:
             raise RuntimeError(b'unknown state: %s' % state)
 
     def set_possibly_dirty(self):
         """Mark a file as "possibly dirty"
 
         This means the next status call will have to actually check its content
         to make sure it is correct.
         """
         self._mtime = None
 
     def set_clean(self, mode, size, mtime):
         """mark a file as "clean" cancelling potential "possibly dirty call"
 
         Note: this function is a descendant of `dirstate.normal` and is
         currently expected to be call on "normal" entry only. There are not
         reason for this to not change in the future as long as the ccode is
         updated to preserve the proper state of the non-normal files.
         """
         self._wc_tracked = True
         self._p1_tracked = True
         self._mode = mode
         self._size = size
         self._mtime = mtime
 
     def set_tracked(self):
         """mark a file as tracked in the working copy
 
         This will ultimately be called by command like `hg add`.
         """
         self._wc_tracked = True
         # `set_tracked` is replacing various `normallookup` call. So we mark
         # the files as needing lookup
         #
         # Consider dropping this in the future in favor of something less broad.
         self._mtime = None
 
     def set_untracked(self):
         """mark a file as untracked in the working copy
 
         This will ultimately be called by command like `hg remove`.
         """
         self._wc_tracked = False
         self._mode = None
         self._size = None
         self._mtime = None
 
     def drop_merge_data(self):
         """remove all "merge-only" from a DirstateItem
 
         This is to be call by the dirstatemap code when the second parent is dropped
         """
         if self._p2_info:
             self._p2_info = False
             self._mode = None
             self._size = None
             self._mtime = None
 
     @property
     def mode(self):
         return self.v1_mode()
 
     @property
     def size(self):
         return self.v1_size()
 
     @property
     def mtime(self):
         return self.v1_mtime()
 
     @property
     def state(self):
         """
         States are:
           n  normal
           m  needs merging
           r  marked for removal
           a  marked for addition
 
         XXX This "state" is a bit obscure and mostly a direct expression of the
         dirstatev1 format. It would make sense to ultimately deprecate it in
         favor of the more "semantic" attributes.
         """
         if not self.any_tracked:
             return b'?'
         return self.v1_state()
 
     @property
     def tracked(self):
         """True is the file is tracked in the working copy"""
         return self._wc_tracked
 
     @property
     def any_tracked(self):
         """True is the file is tracked anywhere (wc or parents)"""
         return self._wc_tracked or self._p1_tracked or self._p2_info
 
     @property
     def added(self):
         """True if the file has been added"""
         return self._wc_tracked and not (self._p1_tracked or self._p2_info)
 
     @property
     def maybe_clean(self):
         """True if the file has a chance to be in the "clean" state"""
         if not self._wc_tracked:
             return False
         elif not self._p1_tracked:
             return False
         elif self._p2_info:
             return False
         return True
 
     @property
     def p1_tracked(self):
         """True if the file is tracked in the first parent manifest"""
         return self._p1_tracked
 
     @property
     def p2_info(self):
         """True if the file needed to merge or apply any input from p2
 
         See the class documentation for details.
         """
         return self._wc_tracked and self._p2_info
 
     @property
     def removed(self):
         """True if the file has been removed"""
         return not self._wc_tracked and (self._p1_tracked or self._p2_info)
 
+    def v2_data(self):
+        """Returns (flags, mode, size, mtime) for v2 serialization"""
+        flags = 0
+        if self._wc_tracked:
+            flags |= DIRSTATE_V2_WDIR_TRACKED
+        if self._p1_tracked:
+            flags |= DIRSTATE_V2_P1_TRACKED
+        if self._p2_info:
+            flags |= DIRSTATE_V2_P2_INFO
+        if self.mode is not None and self.size is not None:
+            flags |= DIRSTATE_V2_HAS_MODE_AND_SIZE
+            if self.mode & stat.S_IXUSR:
+                flags |= DIRSTATE_V2_MODE_EXEC_PERM
+            if stat.S_ISLNK(self.mode):
+                flags |= DIRSTATE_V2_MODE_IS_SYMLINK
+        if self.mtime is not None:
+            flags |= DIRSTATE_V2_HAS_MTIME
+        return (flags, self.size or 0, self.mtime or 0)
+
     def v1_state(self):
         """return a "state" suitable for v1 serialization"""
         if not self.any_tracked:
             # the object has no state to record, this is -currently-
             # unsupported
             raise RuntimeError('untracked item')
         elif self.removed:
             return b'r'
         elif self._p1_tracked and self._p2_info:
             return b'm'
         elif self.added:
             return b'a'
         else:
             return b'n'
 
     def v1_mode(self):
         """return a "mode" suitable for v1 serialization"""
         return self._mode if self._mode is not None else 0
 
     def v1_size(self):
         """return a "size" suitable for v1 serialization"""
         if not self.any_tracked:
             # the object has no state to record, this is -currently-
             # unsupported
             raise RuntimeError('untracked item')
         elif self.removed and self._p1_tracked and self._p2_info:
             return NONNORMAL
         elif self._p2_info:
             return FROM_P2
         elif self.removed:
             return 0
         elif self.added:
             return NONNORMAL
         elif self._size is None:
             return NONNORMAL
         else:
             return self._size
 
     def v1_mtime(self):
         """return a "mtime" suitable for v1 serialization"""
         if not self.any_tracked:
             # the object has no state to record, this is -currently-
             # unsupported
             raise RuntimeError('untracked item')
         elif self.removed:
             return 0
         elif self._mtime is None:
             return AMBIGUOUS_TIME
         elif self._p2_info:
             return AMBIGUOUS_TIME
         elif not self._p1_tracked:
             return AMBIGUOUS_TIME
         else:
             return self._mtime
 
     def need_delay(self, now):
         """True if the stored mtime would be ambiguous with the current time"""
         return self.v1_state() == b'n' and self.v1_mtime() == now
 
 
 def gettype(q):
     return int(q & 0xFFFF)
 
 
 class BaseIndexObject(object):
     # Can I be passed to an algorithme implemented in Rust ?
     rust_ext_compat = 0
     # Format of an index entry according to Python's `struct` language
     index_format = revlog_constants.INDEX_ENTRY_V1
     # Size of a C unsigned long long int, platform independent
     big_int_size = struct.calcsize(b'>Q')
     # Size of a C long int, platform independent
     int_size = struct.calcsize(b'>i')
     # An empty index entry, used as a default value to be overridden, or nullrev
     null_item = (
         0,
         0,
         0,
         -1,
         -1,
         -1,
         -1,
         sha1nodeconstants.nullid,
         0,
         0,
         revlog_constants.COMP_MODE_INLINE,
         revlog_constants.COMP_MODE_INLINE,
     )
 
     @util.propertycache
     def entry_size(self):
         return self.index_format.size
 
     @property
     def nodemap(self):
         msg = b"index.nodemap is deprecated, use index.[has_node|rev|get_rev]"
         util.nouideprecwarn(msg, b'5.3', stacklevel=2)
         return self._nodemap
 
     @util.propertycache
     def _nodemap(self):
         nodemap = nodemaputil.NodeMap({sha1nodeconstants.nullid: nullrev})
         for r in range(0, len(self)):
             n = self[r][7]
             nodemap[n] = r
         return nodemap
 
     def has_node(self, node):
         """return True if the node exist in the index"""
         return node in self._nodemap
 
     def rev(self, node):
         """return a revision for a node
 
         If the node is unknown, raise a RevlogError"""
         return self._nodemap[node]
 
     def get_rev(self, node):
         """return a revision for a node
 
         If the node is unknown, return None"""
         return self._nodemap.get(node)
 
     def _stripnodes(self, start):
         if '_nodemap' in vars(self):
             for r in range(start, len(self)):
                 n = self[r][7]
                 del self._nodemap[n]
 
     def clearcaches(self):
         self.__dict__.pop('_nodemap', None)
 
     def __len__(self):
         return self._lgt + len(self._extra)
 
     def append(self, tup):
         if '_nodemap' in vars(self):
             self._nodemap[tup[7]] = len(self)
         data = self._pack_entry(len(self), tup)
         self._extra.append(data)
 
     def _pack_entry(self, rev, entry):
         assert entry[8] == 0
         assert entry[9] == 0
         return self.index_format.pack(*entry[:8])
 
     def _check_index(self, i):
         if not isinstance(i, int):
             raise TypeError(b"expecting int indexes")
         if i < 0 or i >= len(self):
             raise IndexError
 
     def __getitem__(self, i):
         if i == -1:
             return self.null_item
         self._check_index(i)
         if i >= self._lgt:
             data = self._extra[i - self._lgt]
         else:
             index = self._calculate_index(i)
             data = self._data[index : index + self.entry_size]
         r = self._unpack_entry(i, data)
         if self._lgt and i == 0:
             offset = revlogutils.offset_type(0, gettype(r[0]))
             r = (offset,) + r[1:]
         return r
 
     def _unpack_entry(self, rev, data):
         r = self.index_format.unpack(data)
         r = r + (
             0,
             0,
             revlog_constants.COMP_MODE_INLINE,
             revlog_constants.COMP_MODE_INLINE,
         )
         return r
 
     def pack_header(self, header):
         """pack header information as binary"""
         v_fmt = revlog_constants.INDEX_HEADER
         return v_fmt.pack(header)
 
     def entry_binary(self, rev):
         """return the raw binary string representing a revision"""
         entry = self[rev]
         p = revlog_constants.INDEX_ENTRY_V1.pack(*entry[:8])
         if rev == 0:
             p = p[revlog_constants.INDEX_HEADER.size :]
         return p
 
 
 class IndexObject(BaseIndexObject):
     def __init__(self, data):
         assert len(data) % self.entry_size == 0, (
             len(data),
             self.entry_size,
             len(data) % self.entry_size,
         )
         self._data = data
         self._lgt = len(data) // self.entry_size
         self._extra = []
 
     def _calculate_index(self, i):
         return i * self.entry_size
 
     def __delitem__(self, i):
         if not isinstance(i, slice) or not i.stop == -1 or i.step is not None:
             raise ValueError(b"deleting slices only supports a:-1 with step 1")
         i = i.start
         self._check_index(i)
         self._stripnodes(i)
         if i < self._lgt:
             self._data = self._data[: i * self.entry_size]
             self._lgt = i
             self._extra = []
         else:
             self._extra = self._extra[: i - self._lgt]
 
 
 class PersistentNodeMapIndexObject(IndexObject):
     """a Debug oriented class to test persistent nodemap
 
     We need a simple python object to test API and higher level behavior. See
     the Rust implementation for  more serious usage. This should be used only
     through the dedicated `devel.persistent-nodemap` config.
     """
 
     def nodemap_data_all(self):
         """Return bytes containing a full serialization of a nodemap
 
         The nodemap should be valid for the full set of revisions in the
         index."""
         return nodemaputil.persistent_data(self)
 
     def nodemap_data_incremental(self):
         """Return bytes containing a incremental update to persistent nodemap
 
         This containst the data for an append-only update of the data provided
         in the last call to `update_nodemap_data`.
         """
         if self._nm_root is None:
             return None
         docket = self._nm_docket
         changed, data = nodemaputil.update_persistent_data(
             self, self._nm_root, self._nm_max_idx, self._nm_docket.tip_rev
         )
 
         self._nm_root = self._nm_max_idx = self._nm_docket = None
         return docket, changed, data
 
     def update_nodemap_data(self, docket, nm_data):
         """provide full block of persisted binary data for a nodemap
 
         The data are expected to come from disk. See `nodemap_data_all` for a
         produceur of such data."""
         if nm_data is not None:
             self._nm_root, self._nm_max_idx = nodemaputil.parse_data(nm_data)
             if self._nm_root:
                 self._nm_docket = docket
             else:
                 self._nm_root = self._nm_max_idx = self._nm_docket = None
 
 
 class InlinedIndexObject(BaseIndexObject):
     def __init__(self, data, inline=0):
         self._data = data
         self._lgt = self._inline_scan(None)
         self._inline_scan(self._lgt)
         self._extra = []
 
     def _inline_scan(self, lgt):
         off = 0
         if lgt is not None:
             self._offsets = [0] * lgt
         count = 0
         while off <= len(self._data) - self.entry_size:
             start = off + self.big_int_size
             (s,) = struct.unpack(
                 b'>i',
                 self._data[start : start + self.int_size],
             )
             if lgt is not None:
                 self._offsets[count] = off
             count += 1
             off += self.entry_size + s
         if off != len(self._data):
             raise ValueError(b"corrupted data")
         return count
 
     def __delitem__(self, i):
         if not isinstance(i, slice) or not i.stop == -1 or i.step is not None:
             raise ValueError(b"deleting slices only supports a:-1 with step 1")
         i = i.start
         self._check_index(i)
         self._stripnodes(i)
         if i < self._lgt:
             self._offsets = self._offsets[:i]
             self._lgt = i
             self._extra = []
         else:
             self._extra = self._extra[: i - self._lgt]
 
     def _calculate_index(self, i):
         return self._offsets[i]
 
 
 def parse_index2(data, inline, revlogv2=False):
     if not inline:
         cls = IndexObject2 if revlogv2 else IndexObject
         return cls(data), None
     cls = InlinedIndexObject
     return cls(data, inline), (0, data)
 
 
 def parse_index_cl_v2(data):
     return IndexChangelogV2(data), None
 
 
 class IndexObject2(IndexObject):
     index_format = revlog_constants.INDEX_ENTRY_V2
 
     def replace_sidedata_info(
         self,
         rev,
         sidedata_offset,
         sidedata_length,
         offset_flags,
         compression_mode,
     ):
         """
         Replace an existing index entry's sidedata offset and length with new
         ones.
         This cannot be used outside of the context of sidedata rewriting,
         inside the transaction that creates the revision `rev`.
         """
         if rev < 0:
             raise KeyError
         self._check_index(rev)
         if rev < self._lgt:
             msg = b"cannot rewrite entries outside of this transaction"
             raise KeyError(msg)
         else:
             entry = list(self[rev])
             entry[0] = offset_flags
             entry[8] = sidedata_offset
             entry[9] = sidedata_length
             entry[11] = compression_mode
             entry = tuple(entry)
             new = self._pack_entry(rev, entry)
             self._extra[rev - self._lgt] = new
 
     def _unpack_entry(self, rev, data):
         data = self.index_format.unpack(data)
         entry = data[:10]
         data_comp = data[10] & 3
         sidedata_comp = (data[10] & (3 << 2)) >> 2
         return entry + (data_comp, sidedata_comp)
 
     def _pack_entry(self, rev, entry):
         data = entry[:10]
         data_comp = entry[10] & 3
         sidedata_comp = (entry[11] & 3) << 2
         data += (data_comp | sidedata_comp,)
 
         return self.index_format.pack(*data)
 
     def entry_binary(self, rev):
         """return the raw binary string representing a revision"""
         entry = self[rev]
         return self._pack_entry(rev, entry)
 
     def pack_header(self, header):
         """pack header information as binary"""
         msg = 'version header should go in the docket, not the index: %d'
         msg %= header
         raise error.ProgrammingError(msg)
 
 
 class IndexChangelogV2(IndexObject2):
     index_format = revlog_constants.INDEX_ENTRY_CL_V2
 
     def _unpack_entry(self, rev, data, r=True):
         items = self.index_format.unpack(data)
         entry = items[:3] + (rev, rev) + items[3:8]
         data_comp = items[8] & 3
         sidedata_comp = (items[8] >> 2) & 3
         return entry + (data_comp, sidedata_comp)
 
     def _pack_entry(self, rev, entry):
         assert entry[3] == rev, entry[3]
         assert entry[4] == rev, entry[4]
         data = entry[:3] + entry[5:10]
         data_comp = entry[10] & 3
         sidedata_comp = (entry[11] & 3) << 2
         data += (data_comp | sidedata_comp,)
         return self.index_format.pack(*data)
 
 
 def parse_index_devel_nodemap(data, inline):
     """like parse_index2, but alway return a PersistentNodeMapIndexObject"""
     return PersistentNodeMapIndexObject(data), None
 
 
 def parse_dirstate(dmap, copymap, st):
     parents = [st[:20], st[20:40]]
     # dereference fields so they will be local in loop
     format = b">cllll"
     e_size = struct.calcsize(format)
     pos1 = 40
     l = len(st)
 
     # the inner loop
     while pos1 < l:
         pos2 = pos1 + e_size
         e = _unpack(b">cllll", st[pos1:pos2])  # a literal here is faster
         pos1 = pos2 + e[4]
         f = st[pos2:pos1]
         if b'\0' in f:
             f, c = f.split(b'\0')
             copymap[f] = c
         dmap[f] = DirstateItem.from_v1_data(*e[:4])
     return parents
 
 
 def pack_dirstate(dmap, copymap, pl, now):
     now = int(now)
     cs = stringio()
     write = cs.write
     write(b"".join(pl))
     for f, e in pycompat.iteritems(dmap):
         if e.need_delay(now):
             # The file was last modified "simultaneously" with the current
             # write to dirstate (i.e. within the same second for file-
             # systems with a granularity of 1 sec). This commonly happens
             # for at least a couple of files on 'update'.
             # The user could change the file without changing its size
             # within the same second. Invalidate the file's mtime in
             # dirstate, forcing future 'status' calls to compare the
             # contents of the file if the size is the same. This prevents
             # mistakenly treating such files as clean.
             e.set_possibly_dirty()
 
         if f in copymap:
             f = b"%s\0%s" % (f, copymap[f])
         e = _pack(
             b">cllll",
             e.v1_state(),
             e.v1_mode(),
             e.v1_size(),
             e.v1_mtime(),
             len(f),
         )
         write(e)
         write(f)
     return cs.getvalue()