phases: sparsify phaseroots and phasesets

As final step of dealing with the holes in the phase numbers, make
phaseroots and phasesets both dictionaries indexed by the phase number.
Further adjust the interface of the C module by pushing the node to
revision mapping down as it is cheaper on the C side to deal with
revision numbers.

Overall, the patch series improves a no-change "hg up" for my NetBSD test
repository from 4.7s to 1.3s.

Differential Revision: https://phab.mercurial-scm.org/D8698

mercurial/cext/parsers.c

@@ -1,762 +1,762
 /*
  parsers.c - efficient content parsing
 
  Copyright 2008 Matt 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 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 inline dirstateTupleObject *make_dirstate_tuple(char state, int mode,
                                                        int size, int mtime)
 {
 	dirstateTupleObject *t =
 	    PyObject_New(dirstateTupleObject, &dirstateTupleType);
 	if (!t) {
 		return NULL;
 	}
 	t->state = state;
 	t->mode = mode;
 	t->size = size;
 	t->mtime = mtime;
 	return t;
 }
 
 static PyObject *dirstate_tuple_new(PyTypeObject *subtype, PyObject *args,
                                     PyObject *kwds)
 {
 	/* We do all the initialization here and not a tp_init function because
 	 * dirstate_tuple is immutable. */
 	dirstateTupleObject *t;
 	char state;
 	int size, mode, mtime;
 	if (!PyArg_ParseTuple(args, "ciii", &state, &mode, &size, &mtime)) {
 		return NULL;
 	}
 
 	t = (dirstateTupleObject *)subtype->tp_alloc(subtype, 1);
 	if (!t) {
 		return NULL;
 	}
 	t->state = state;
 	t->mode = mode;
 	t->size = size;
 	t->mtime = mtime;
 
 	return (PyObject *)t;
 }
 
 static void dirstate_tuple_dealloc(PyObject *o)
 {
 	PyObject_Del(o);
 }
 
 static Py_ssize_t dirstate_tuple_length(PyObject *o)
 {
 	return 4;
 }
 
 static PyObject *dirstate_tuple_item(PyObject *o, Py_ssize_t i)
 {
 	dirstateTupleObject *t = (dirstateTupleObject *)o;
 	switch (i) {
 	case 0:
 		return PyBytes_FromStringAndSize(&t->state, 1);
 	case 1:
 		return PyInt_FromLong(t->mode);
 	case 2:
 		return PyInt_FromLong(t->size);
 	case 3:
 		return PyInt_FromLong(t->mtime);
 	default:
 		PyErr_SetString(PyExc_IndexError, "index out of range");
 		return NULL;
 	}
 }
 
 static PySequenceMethods dirstate_tuple_sq = {
     dirstate_tuple_length, /* sq_length */
     0,                     /* sq_concat */
     0,                     /* sq_repeat */
     dirstate_tuple_item,   /* sq_item */
     0,                     /* sq_ass_item */
     0,                     /* sq_contains */
     0,                     /* sq_inplace_concat */
     0                      /* sq_inplace_repeat */
 };
 
 PyTypeObject dirstateTupleType = {
     PyVarObject_HEAD_INIT(NULL, 0)      /* header */
     "dirstate_tuple",                   /* tp_name */
     sizeof(dirstateTupleObject),        /* tp_basicsize */
     0,                                  /* tp_itemsize */
     (destructor)dirstate_tuple_dealloc, /* tp_dealloc */
     0,                                  /* tp_print */
     0,                                  /* tp_getattr */
     0,                                  /* tp_setattr */
     0,                                  /* tp_compare */
     0,                                  /* tp_repr */
     0,                                  /* tp_as_number */
     &dirstate_tuple_sq,                 /* 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 */
     0,                                  /* 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 */
     0,                                  /* tp_init */
     0,                                  /* tp_alloc */
     dirstate_tuple_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 *)make_dirstate_tuple(state, mode, size, mtime);
 		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;
 }
 
 /*
  * Build a set of non-normal and other parent entries from the dirstate dmap
  */
 static PyObject *nonnormalotherparententries(PyObject *self, PyObject *args)
 {
 	PyObject *dmap, *fname, *v;
 	PyObject *nonnset = NULL, *otherpset = NULL, *result = NULL;
 	Py_ssize_t pos;
 
 	if (!PyArg_ParseTuple(args, "O!:nonnormalentries", &PyDict_Type,
 	                      &dmap)) {
 		goto bail;
 	}
 
 	nonnset = PySet_New(NULL);
 	if (nonnset == NULL) {
 		goto bail;
 	}
 
 	otherpset = PySet_New(NULL);
 	if (otherpset == NULL) {
 		goto bail;
 	}
 
 	pos = 0;
 	while (PyDict_Next(dmap, &pos, &fname, &v)) {
 		dirstateTupleObject *t;
 		if (!dirstate_tuple_check(v)) {
 			PyErr_SetString(PyExc_TypeError,
 			                "expected a dirstate tuple");
 			goto bail;
 		}
 		t = (dirstateTupleObject *)v;
 
 		if (t->state == 'n' && t->size == -2) {
 			if (PySet_Add(otherpset, fname) == -1) {
 				goto bail;
 			}
 		}
 
 		if (t->state == 'n' && t->mtime != -1) {
 			continue;
 		}
 		if (PySet_Add(nonnset, fname) == -1) {
 			goto bail;
 		}
 	}
 
 	result = Py_BuildValue("(OO)", nonnset, otherpset);
 	if (result == NULL) {
 		goto bail;
 	}
 	Py_DECREF(nonnset);
 	Py_DECREF(otherpset);
 	return result;
 bail:
 	Py_XDECREF(nonnset);
 	Py_XDECREF(otherpset);
 	Py_XDECREF(result);
 	return NULL;
 }
 
 /*
  * 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);) {
 		dirstateTupleObject *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 = (dirstateTupleObject *)v;
 
 		state = tuple->state;
 		mode = tuple->mode;
 		size = tuple->size;
 		mtime = tuple->mtime;
 		if (state == 'n' && mtime == now) {
 			/* See pure/parsers.py:pack_dirstate for why we do
 			 * this. */
 			mtime = -1;
 			mtime_unset = (PyObject *)make_dirstate_tuple(
 			    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);
 
 static PyMethodDef methods[] = {
     {"pack_dirstate", pack_dirstate, METH_VARARGS, "pack a dirstate\n"},
     {"nonnormalotherparententries", nonnormalotherparententries, METH_VARARGS,
      "create a set containing non-normal and other parent entries of given "
      "dirstate\n"},
     {"parse_dirstate", parse_dirstate, METH_VARARGS, "parse a dirstate\n"},
     {"parse_index2", parse_index2, METH_VARARGS, "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 = 16;
+static const int version = 17;
 
 static void module_init(PyObject *mod)
 {
 	PyObject *capsule = NULL;
 	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);
 
 	capsule = PyCapsule_New(
 	    make_dirstate_tuple,
 	    "mercurial.cext.parsers.make_dirstate_tuple_CAPI", NULL);
 	if (capsule != NULL)
 		PyModule_AddObject(mod, "make_dirstate_tuple_CAPI", capsule);
 
 	if (PyType_Ready(&dirstateTupleType) < 0) {
 		return;
 	}
 	Py_INCREF(&dirstateTupleType);
 	PyModule_AddObject(mod, "dirstatetuple",
 	                   (PyObject *)&dirstateTupleType);
 }
 
 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/cext/revlog.c

@@ -1,2918 +1,2959
 /*
  parsers.c - efficient content parsing
 
  Copyright 2008 Matt 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 <assert.h>
 #include <ctype.h>
 #include <limits.h>
 #include <stddef.h>
 #include <stdlib.h>
 #include <string.h>
 
 #include "bitmanipulation.h"
 #include "charencode.h"
 #include "revlog.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
 
 typedef struct indexObjectStruct indexObject;
 
 typedef struct {
 	int children[16];
 } nodetreenode;
 
 typedef struct {
 	int abi_version;
 	Py_ssize_t (*index_length)(const indexObject *);
 	const char *(*index_node)(indexObject *, Py_ssize_t);
 	int (*index_parents)(PyObject *, int, int *);
 } Revlog_CAPI;
 
 /*
  * A base-16 trie for fast node->rev mapping.
  *
  * Positive value is index of the next node in the trie
  * Negative value is a leaf: -(rev + 2)
  * Zero is empty
  */
 typedef struct {
 	indexObject *index;
 	nodetreenode *nodes;
 	unsigned length;   /* # nodes in use */
 	unsigned capacity; /* # nodes allocated */
 	int depth;         /* maximum depth of tree */
 	int splits;        /* # splits performed */
 } nodetree;
 
 typedef struct {
 	PyObject_HEAD /* ; */
 	    nodetree nt;
 } nodetreeObject;
 
 /*
  * This class has two behaviors.
  *
  * When used in a list-like way (with integer keys), we decode an
  * entry in a RevlogNG index file on demand. We have limited support for
  * integer-keyed insert and delete, only at elements right before the
  * end.
  *
  * With string keys, we lazily perform a reverse mapping from node to
  * rev, using a base-16 trie.
  */
 struct indexObjectStruct {
 	PyObject_HEAD
 	    /* Type-specific fields go here. */
 	    PyObject *data;     /* raw bytes of index */
 	Py_buffer buf;          /* buffer of data */
 	PyObject **cache;       /* cached tuples */
 	const char **offsets;   /* populated on demand */
 	Py_ssize_t raw_length;  /* original number of elements */
 	Py_ssize_t length;      /* current number of elements */
 	PyObject *added;        /* populated on demand */
 	PyObject *headrevs;     /* cache, invalidated on changes */
 	PyObject *filteredrevs; /* filtered revs set */
 	nodetree nt;            /* base-16 trie */
 	int ntinitialized;      /* 0 or 1 */
 	int ntrev;              /* last rev scanned */
 	int ntlookups;          /* # lookups */
 	int ntmisses;           /* # lookups that miss the cache */
 	int inlined;
 };
 
 static Py_ssize_t index_length(const indexObject *self)
 {
 	if (self->added == NULL)
 		return self->length;
 	return self->length + PyList_GET_SIZE(self->added);
 }
 
 static PyObject *nullentry = NULL;
 static const char nullid[20] = {0};
 static const Py_ssize_t nullrev = -1;
 
 static Py_ssize_t inline_scan(indexObject *self, const char **offsets);
 
+static int index_find_node(indexObject *self, const char *node,
+                           Py_ssize_t nodelen);
+
 #if LONG_MAX == 0x7fffffffL
 static const char *const tuple_format = PY23("Kiiiiiis#", "Kiiiiiiy#");
 #else
 static const char *const tuple_format = PY23("kiiiiiis#", "kiiiiiiy#");
 #endif
 
 /* A RevlogNG v1 index entry is 64 bytes long. */
 static const long v1_hdrsize = 64;
 
 static void raise_revlog_error(void)
 {
 	PyObject *mod = NULL, *dict = NULL, *errclass = NULL;
 
 	mod = PyImport_ImportModule("mercurial.error");
 	if (mod == NULL) {
 		goto cleanup;
 	}
 
 	dict = PyModule_GetDict(mod);
 	if (dict == NULL) {
 		goto cleanup;
 	}
 	Py_INCREF(dict);
 
 	errclass = PyDict_GetItemString(dict, "RevlogError");
 	if (errclass == NULL) {
 		PyErr_SetString(PyExc_SystemError,
 		                "could not find RevlogError");
 		goto cleanup;
 	}
 
 	/* value of exception is ignored by callers */
 	PyErr_SetString(errclass, "RevlogError");
 
 cleanup:
 	Py_XDECREF(dict);
 	Py_XDECREF(mod);
 }
 
 /*
  * Return a pointer to the beginning of a RevlogNG record.
  */
 static const char *index_deref(indexObject *self, Py_ssize_t pos)
 {
 	if (self->inlined && pos > 0) {
 		if (self->offsets == NULL) {
 			Py_ssize_t ret;
 			self->offsets = PyMem_Malloc(self->raw_length *
 			                             sizeof(*self->offsets));
 			if (self->offsets == NULL)
 				return (const char *)PyErr_NoMemory();
 			ret = inline_scan(self, self->offsets);
 			if (ret == -1) {
 				return NULL;
 			};
 		}
 		return self->offsets[pos];
 	}
 
 	return (const char *)(self->buf.buf) + pos * v1_hdrsize;
 }
 
 /*
  * Get parents of the given rev.
  *
  * The specified rev must be valid and must not be nullrev. A returned
  * parent revision may be nullrev, but is guaranteed to be in valid range.
  */
 static inline int index_get_parents(indexObject *self, Py_ssize_t rev, int *ps,
                                     int maxrev)
 {
 	if (rev >= self->length) {
 		long tmp;
 		PyObject *tuple =
 		    PyList_GET_ITEM(self->added, rev - self->length);
 		if (!pylong_to_long(PyTuple_GET_ITEM(tuple, 5), &tmp)) {
 			return -1;
 		}
 		ps[0] = (int)tmp;
 		if (!pylong_to_long(PyTuple_GET_ITEM(tuple, 6), &tmp)) {
 			return -1;
 		}
 		ps[1] = (int)tmp;
 	} else {
 		const char *data = index_deref(self, rev);
 		ps[0] = getbe32(data + 24);
 		ps[1] = getbe32(data + 28);
 	}
 	/* If index file is corrupted, ps[] may point to invalid revisions. So
 	 * there is a risk of buffer overflow to trust them unconditionally. */
 	if (ps[0] < -1 || ps[0] > maxrev || ps[1] < -1 || ps[1] > maxrev) {
 		PyErr_SetString(PyExc_ValueError, "parent out of range");
 		return -1;
 	}
 	return 0;
 }
 
 /*
  * Get parents of the given rev.
  *
  * If the specified rev is out of range, IndexError will be raised. If the
  * revlog entry is corrupted, ValueError may be raised.
  *
  * Returns 0 on success or -1 on failure.
  */
 static int HgRevlogIndex_GetParents(PyObject *op, int rev, int *ps)
 {
 	int tiprev;
 	if (!op || !HgRevlogIndex_Check(op) || !ps) {
 		PyErr_BadInternalCall();
 		return -1;
 	}
 	tiprev = (int)index_length((indexObject *)op) - 1;
 	if (rev < -1 || rev > tiprev) {
 		PyErr_Format(PyExc_IndexError, "rev out of range: %d", rev);
 		return -1;
 	} else if (rev == -1) {
 		ps[0] = ps[1] = -1;
 		return 0;
 	} else {
 		return index_get_parents((indexObject *)op, rev, ps, tiprev);
 	}
 }
 
 static inline int64_t index_get_start(indexObject *self, Py_ssize_t rev)
 {
 	uint64_t offset;
 	if (rev == nullrev) {
 		return 0;
 	}
 	if (rev >= self->length) {
 		PyObject *tuple;
 		PyObject *pylong;
 		PY_LONG_LONG tmp;
 		tuple = PyList_GET_ITEM(self->added, rev - self->length);
 		pylong = PyTuple_GET_ITEM(tuple, 0);
 		tmp = PyLong_AsLongLong(pylong);
 		if (tmp == -1 && PyErr_Occurred()) {
 			return -1;
 		}
 		if (tmp < 0) {
 			PyErr_Format(PyExc_OverflowError,
 			             "revlog entry size out of bound (%lld)",
 			             (long long)tmp);
 			return -1;
 		}
 		offset = (uint64_t)tmp;
 	} else {
 		const char *data = index_deref(self, rev);
 		offset = getbe32(data + 4);
 		if (rev == 0) {
 			/* mask out version number for the first entry */
 			offset &= 0xFFFF;
 		} else {
 			uint32_t offset_high = getbe32(data);
 			offset |= ((uint64_t)offset_high) << 32;
 		}
 	}
 	return (int64_t)(offset >> 16);
 }
 
 static inline int index_get_length(indexObject *self, Py_ssize_t rev)
 {
 	if (rev == nullrev) {
 		return 0;
 	}
 	if (rev >= self->length) {
 		PyObject *tuple;
 		PyObject *pylong;
 		long ret;
 		tuple = PyList_GET_ITEM(self->added, rev - self->length);
 		pylong = PyTuple_GET_ITEM(tuple, 1);
 		ret = PyInt_AsLong(pylong);
 		if (ret == -1 && PyErr_Occurred()) {
 			return -1;
 		}
 		if (ret < 0 || ret > (long)INT_MAX) {
 			PyErr_Format(PyExc_OverflowError,
 			             "revlog entry size out of bound (%ld)",
 			             ret);
 			return -1;
 		}
 		return (int)ret;
 	} else {
 		const char *data = index_deref(self, rev);
 		int tmp = (int)getbe32(data + 8);
 		if (tmp < 0) {
 			PyErr_Format(PyExc_OverflowError,
 			             "revlog entry size out of bound (%d)",
 			             tmp);
 			return -1;
 		}
 		return tmp;
 	}
 }
 
 /*
  * RevlogNG format (all in big endian, data may be inlined):
  *    6 bytes: offset
  *    2 bytes: flags
  *    4 bytes: compressed length
  *    4 bytes: uncompressed length
  *    4 bytes: base revision
  *    4 bytes: link revision
  *    4 bytes: parent 1 revision
  *    4 bytes: parent 2 revision
  *   32 bytes: nodeid (only 20 bytes used)
  */
 static PyObject *index_get(indexObject *self, Py_ssize_t pos)
 {
 	uint64_t offset_flags;
 	int comp_len, uncomp_len, base_rev, link_rev, parent_1, parent_2;
 	const char *c_node_id;
 	const char *data;
 	Py_ssize_t length = index_length(self);
 	PyObject *entry;
 
 	if (pos == nullrev) {
 		Py_INCREF(nullentry);
 		return nullentry;
 	}
 
 	if (pos < 0 || pos >= length) {
 		PyErr_SetString(PyExc_IndexError, "revlog index out of range");
 		return NULL;
 	}
 
 	if (pos >= self->length) {
 		PyObject *obj;
 		obj = PyList_GET_ITEM(self->added, pos - self->length);
 		Py_INCREF(obj);
 		return obj;
 	}
 
 	if (self->cache) {
 		if (self->cache[pos]) {
 			Py_INCREF(self->cache[pos]);
 			return self->cache[pos];
 		}
 	} else {
 		self->cache = calloc(self->raw_length, sizeof(PyObject *));
 		if (self->cache == NULL)
 			return PyErr_NoMemory();
 	}
 
 	data = index_deref(self, pos);
 	if (data == NULL)
 		return NULL;
 
 	offset_flags = getbe32(data + 4);
 	if (pos == 0) /* mask out version number for the first entry */
 		offset_flags &= 0xFFFF;
 	else {
 		uint32_t offset_high = getbe32(data);
 		offset_flags |= ((uint64_t)offset_high) << 32;
 	}
 
 	comp_len = getbe32(data + 8);
 	uncomp_len = getbe32(data + 12);
 	base_rev = getbe32(data + 16);
 	link_rev = getbe32(data + 20);
 	parent_1 = getbe32(data + 24);
 	parent_2 = getbe32(data + 28);
 	c_node_id = data + 32;
 
 	entry = Py_BuildValue(tuple_format, offset_flags, comp_len, uncomp_len,
 	                      base_rev, link_rev, parent_1, parent_2, c_node_id,
 	                      (Py_ssize_t)20);
 
 	if (entry) {
 		PyObject_GC_UnTrack(entry);
 		Py_INCREF(entry);
 	}
 
 	self->cache[pos] = entry;
 
 	return entry;
 }
 
 /*
  * Return the 20-byte SHA of the node corresponding to the given rev.
  */
 static const char *index_node(indexObject *self, Py_ssize_t pos)
 {
 	Py_ssize_t length = index_length(self);
 	const char *data;
 
 	if (pos == nullrev)
 		return nullid;
 
 	if (pos >= length)
 		return NULL;
 
 	if (pos >= self->length) {
 		PyObject *tuple, *str;
 		tuple = PyList_GET_ITEM(self->added, pos - self->length);
 		str = PyTuple_GetItem(tuple, 7);
 		return str ? PyBytes_AS_STRING(str) : NULL;
 	}
 
 	data = index_deref(self, pos);
 	return data ? data + 32 : NULL;
 }
 
 /*
  * Return the 20-byte SHA of the node corresponding to the given rev. The
  * rev is assumed to be existing. If not, an exception is set.
  */
 static const char *index_node_existing(indexObject *self, Py_ssize_t pos)
 {
 	const char *node = index_node(self, pos);
 	if (node == NULL) {
 		PyErr_Format(PyExc_IndexError, "could not access rev %d",
 		             (int)pos);
 	}
 	return node;
 }
 
 static int nt_insert(nodetree *self, const char *node, int rev);
 
 static int node_check(PyObject *obj, char **node)
 {
 	Py_ssize_t nodelen;
 	if (PyBytes_AsStringAndSize(obj, node, &nodelen) == -1)
 		return -1;
 	if (nodelen == 20)
 		return 0;
 	PyErr_SetString(PyExc_ValueError, "20-byte hash required");
 	return -1;
 }
 
 static PyObject *index_append(indexObject *self, PyObject *obj)
 {
 	char *node;
 	Py_ssize_t len;
 
 	if (!PyTuple_Check(obj) || PyTuple_GET_SIZE(obj) != 8) {
 		PyErr_SetString(PyExc_TypeError, "8-tuple required");
 		return NULL;
 	}
 
 	if (node_check(PyTuple_GET_ITEM(obj, 7), &node) == -1)
 		return NULL;
 
 	len = index_length(self);
 
 	if (self->added == NULL) {
 		self->added = PyList_New(0);
 		if (self->added == NULL)
 			return NULL;
 	}
 
 	if (PyList_Append(self->added, obj) == -1)
 		return NULL;
 
 	if (self->ntinitialized)
 		nt_insert(&self->nt, node, (int)len);
 
 	Py_CLEAR(self->headrevs);
 	Py_RETURN_NONE;
 }
 
 static PyObject *index_stats(indexObject *self)
 {
 	PyObject *obj = PyDict_New();
 	PyObject *s = NULL;
 	PyObject *t = NULL;
 
 	if (obj == NULL)
 		return NULL;
 
 #define istat(__n, __d)                                                        \
 	do {                                                                   \
 		s = PyBytes_FromString(__d);                                   \
 		t = PyInt_FromSsize_t(self->__n);                              \
 		if (!s || !t)                                                  \
 			goto bail;                                             \
 		if (PyDict_SetItem(obj, s, t) == -1)                           \
 			goto bail;                                             \
 		Py_CLEAR(s);                                                   \
 		Py_CLEAR(t);                                                   \
 	} while (0)
 
 	if (self->added) {
 		Py_ssize_t len = PyList_GET_SIZE(self->added);
 		s = PyBytes_FromString("index entries added");
 		t = PyInt_FromSsize_t(len);
 		if (!s || !t)
 			goto bail;
 		if (PyDict_SetItem(obj, s, t) == -1)
 			goto bail;
 		Py_CLEAR(s);
 		Py_CLEAR(t);
 	}
 
 	if (self->raw_length != self->length)
 		istat(raw_length, "revs on disk");
 	istat(length, "revs in memory");
 	istat(ntlookups, "node trie lookups");
 	istat(ntmisses, "node trie misses");
 	istat(ntrev, "node trie last rev scanned");
 	if (self->ntinitialized) {
 		istat(nt.capacity, "node trie capacity");
 		istat(nt.depth, "node trie depth");
 		istat(nt.length, "node trie count");
 		istat(nt.splits, "node trie splits");
 	}
 
 #undef istat
 
 	return obj;
 
 bail:
 	Py_XDECREF(obj);
 	Py_XDECREF(s);
 	Py_XDECREF(t);
 	return NULL;
 }
 
 /*
  * When we cache a list, we want to be sure the caller can't mutate
  * the cached copy.
  */
 static PyObject *list_copy(PyObject *list)
 {
 	Py_ssize_t len = PyList_GET_SIZE(list);
 	PyObject *newlist = PyList_New(len);
 	Py_ssize_t i;
 
 	if (newlist == NULL)
 		return NULL;
 
 	for (i = 0; i < len; i++) {
 		PyObject *obj = PyList_GET_ITEM(list, i);
 		Py_INCREF(obj);
 		PyList_SET_ITEM(newlist, i, obj);
 	}
 
 	return newlist;
 }
 
 static int check_filter(PyObject *filter, Py_ssize_t arg)
 {
 	if (filter) {
 		PyObject *arglist, *result;
 		int isfiltered;
 
 		arglist = Py_BuildValue("(n)", arg);
 		if (!arglist) {
 			return -1;
 		}
 
 		result = PyEval_CallObject(filter, arglist);
 		Py_DECREF(arglist);
 		if (!result) {
 			return -1;
 		}
 
 		/* PyObject_IsTrue returns 1 if true, 0 if false, -1 if error,
 		 * same as this function, so we can just return it directly.*/
 		isfiltered = PyObject_IsTrue(result);
 		Py_DECREF(result);
 		return isfiltered;
 	} else {
 		return 0;
 	}
 }
 
-static Py_ssize_t add_roots_get_min(indexObject *self, PyObject *list,
-                                    Py_ssize_t marker, char *phases)
-{
-	PyObject *iter = NULL;
-	PyObject *iter_item = NULL;
-	Py_ssize_t min_idx = index_length(self) + 2;
-	long iter_item_long;
-
-	if (PyList_GET_SIZE(list) != 0) {
-		iter = PyObject_GetIter(list);
-		if (iter == NULL)
-			return -2;
-		while ((iter_item = PyIter_Next(iter))) {
-			if (!pylong_to_long(iter_item, &iter_item_long)) {
-				Py_DECREF(iter_item);
-				return -2;
-			}
-			Py_DECREF(iter_item);
-			if (iter_item_long < min_idx)
-				min_idx = iter_item_long;
-			phases[iter_item_long] = (char)marker;
-		}
-		Py_DECREF(iter);
-	}
-
-	return min_idx;
-}
-
 static inline void set_phase_from_parents(char *phases, int parent_1,
                                           int parent_2, Py_ssize_t i)
 {
 	if (parent_1 >= 0 && phases[parent_1] > phases[i])
 		phases[i] = phases[parent_1];
 	if (parent_2 >= 0 && phases[parent_2] > phases[i])
 		phases[i] = phases[parent_2];
 }
 
 static PyObject *reachableroots2(indexObject *self, PyObject *args)
 {
 
 	/* Input */
 	long minroot;
 	PyObject *includepatharg = NULL;
 	int includepath = 0;
 	/* heads and roots are lists */
 	PyObject *heads = NULL;
 	PyObject *roots = NULL;
 	PyObject *reachable = NULL;
 
 	PyObject *val;
 	Py_ssize_t len = index_length(self);
 	long revnum;
 	Py_ssize_t k;
 	Py_ssize_t i;
 	Py_ssize_t l;
 	int r;
 	int parents[2];
 
 	/* Internal data structure:
 	 * tovisit: array of length len+1 (all revs + nullrev), filled upto
 	 * lentovisit
 	 *
 	 * revstates: array of length len+1 (all revs + nullrev) */
 	int *tovisit = NULL;
 	long lentovisit = 0;
 	enum { RS_SEEN = 1, RS_ROOT = 2, RS_REACHABLE = 4 };
 	char *revstates = NULL;
 
 	/* Get arguments */
 	if (!PyArg_ParseTuple(args, "lO!O!O!", &minroot, &PyList_Type, &heads,
 	                      &PyList_Type, &roots, &PyBool_Type,
 	                      &includepatharg))
 		goto bail;
 
 	if (includepatharg == Py_True)
 		includepath = 1;
 
 	/* Initialize return set */
 	reachable = PyList_New(0);
 	if (reachable == NULL)
 		goto bail;
 
 	/* Initialize internal datastructures */
 	tovisit = (int *)malloc((len + 1) * sizeof(int));
 	if (tovisit == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	revstates = (char *)calloc(len + 1, 1);
 	if (revstates == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	l = PyList_GET_SIZE(roots);
 	for (i = 0; i < l; i++) {
 		revnum = PyInt_AsLong(PyList_GET_ITEM(roots, i));
 		if (revnum == -1 && PyErr_Occurred())
 			goto bail;
 		/* If root is out of range, e.g. wdir(), it must be unreachable
 		 * from heads. So we can just ignore it. */
 		if (revnum + 1 < 0 || revnum + 1 >= len + 1)
 			continue;
 		revstates[revnum + 1] |= RS_ROOT;
 	}
 
 	/* Populate tovisit with all the heads */
 	l = PyList_GET_SIZE(heads);
 	for (i = 0; i < l; i++) {
 		revnum = PyInt_AsLong(PyList_GET_ITEM(heads, i));
 		if (revnum == -1 && PyErr_Occurred())
 			goto bail;
 		if (revnum + 1 < 0 || revnum + 1 >= len + 1) {
 			PyErr_SetString(PyExc_IndexError, "head out of range");
 			goto bail;
 		}
 		if (!(revstates[revnum + 1] & RS_SEEN)) {
 			tovisit[lentovisit++] = (int)revnum;
 			revstates[revnum + 1] |= RS_SEEN;
 		}
 	}
 
 	/* Visit the tovisit list and find the reachable roots */
 	k = 0;
 	while (k < lentovisit) {
 		/* Add the node to reachable if it is a root*/
 		revnum = tovisit[k++];
 		if (revstates[revnum + 1] & RS_ROOT) {
 			revstates[revnum + 1] |= RS_REACHABLE;
 			val = PyInt_FromLong(revnum);
 			if (val == NULL)
 				goto bail;
 			r = PyList_Append(reachable, val);
 			Py_DECREF(val);
 			if (r < 0)
 				goto bail;
 			if (includepath == 0)
 				continue;
 		}
 
 		/* Add its parents to the list of nodes to visit */
 		if (revnum == nullrev)
 			continue;
 		r = index_get_parents(self, revnum, parents, (int)len - 1);
 		if (r < 0)
 			goto bail;
 		for (i = 0; i < 2; i++) {
 			if (!(revstates[parents[i] + 1] & RS_SEEN) &&
 			    parents[i] >= minroot) {
 				tovisit[lentovisit++] = parents[i];
 				revstates[parents[i] + 1] |= RS_SEEN;
 			}
 		}
 	}
 
 	/* Find all the nodes in between the roots we found and the heads
 	 * and add them to the reachable set */
 	if (includepath == 1) {
 		long minidx = minroot;
 		if (minidx < 0)
 			minidx = 0;
 		for (i = minidx; i < len; i++) {
 			if (!(revstates[i + 1] & RS_SEEN))
 				continue;
 			r = index_get_parents(self, i, parents, (int)len - 1);
 			/* Corrupted index file, error is set from
 			 * index_get_parents */
 			if (r < 0)
 				goto bail;
 			if (((revstates[parents[0] + 1] |
 			      revstates[parents[1] + 1]) &
 			     RS_REACHABLE) &&
 			    !(revstates[i + 1] & RS_REACHABLE)) {
 				revstates[i + 1] |= RS_REACHABLE;
 				val = PyInt_FromSsize_t(i);
 				if (val == NULL)
 					goto bail;
 				r = PyList_Append(reachable, val);
 				Py_DECREF(val);
 				if (r < 0)
 					goto bail;
 			}
 		}
 	}
 
 	free(revstates);
 	free(tovisit);
 	return reachable;
 bail:
 	Py_XDECREF(reachable);
 	free(revstates);
 	free(tovisit);
 	return NULL;
 }
 
+static int add_roots_get_min(indexObject *self, PyObject *roots, char *phases,
+                             char phase)
+{
+	Py_ssize_t len = index_length(self);
+	PyObject *iter;
+	PyObject *item;
+	PyObject *iterator;
+	int rev, minrev = -1;
+	char *node;
+
+	if (!PySet_Check(roots))
+		return -2;
+	iterator = PyObject_GetIter(roots);
+	if (iterator == NULL)
+		return -2;
+	while ((item = PyIter_Next(iterator))) {
+		if (node_check(item, &node) == -1)
+			goto failed;
+		rev = index_find_node(self, node, 20);
+		/* null is implicitly public, so negative is invalid */
+		if (rev < 0 || rev >= len)
+			goto failed;
+		phases[rev] = phase;
+		if (minrev == -1 || minrev > rev)
+			minrev = rev;
+		Py_DECREF(item);
+	}
+	Py_DECREF(iterator);
+	return minrev;
+failed:
+	Py_DECREF(iterator);
+	Py_DECREF(item);
+	return -2;
+}
+
 static PyObject *compute_phases_map_sets(indexObject *self, PyObject *args)
 {
-	PyObject *roots = Py_None;
+	/* 0: public (untracked), 1: draft, 2: secret, 32: archive,
+	   96: internal */
+	static const char trackedphases[] = {1, 2, 32, 96};
 	PyObject *ret = NULL;
-	PyObject *phasessize = NULL;
+	PyObject *roots = Py_None;
+	PyObject *idx = NULL;
+	PyObject *pyphase = NULL;
+	PyObject *pyrev = NULL;
 	PyObject *phaseroots = NULL;
-	PyObject *phaseset = NULL;
-	PyObject *phasessetlist = NULL;
-	PyObject *rev = NULL;
+	PyObject *phasessize = NULL;
+	PyObject *phasesets[4] = {NULL, NULL, NULL, NULL};
 	Py_ssize_t len = index_length(self);
-	Py_ssize_t numphase = 0;
-	Py_ssize_t minrevallphases = 0;
-	Py_ssize_t minrevphase = 0;
-	Py_ssize_t i = 0;
+	const char *currentphase;
 	char *phases = NULL;
-	long phase;
+	int minphaserev = -1, rev, i;
+	const int numphases = (int)(sizeof(phasesets) / sizeof(phasesets[0]));
 
 	if (!PyArg_ParseTuple(args, "O", &roots))
-		goto done;
-	if (roots == NULL || !PyList_Check(roots)) {
-		PyErr_SetString(PyExc_TypeError, "roots must be a list");
-		goto done;
+		return NULL;
+	if (roots == NULL || !PyDict_Check(roots)) {
+		PyErr_SetString(PyExc_TypeError, "roots must be a dictionary");
+		return NULL;
 	}
 
-	phases = calloc(
-	    len, 1); /* phase per rev: {0: public, 1: draft, 2: secret} */
+	phases = calloc(len, 1);
 	if (phases == NULL) {
 		PyErr_NoMemory();
-		goto done;
+		return NULL;
 	}
-	/* Put the phase information of all the roots in phases */
-	numphase = PyList_GET_SIZE(roots) + 1;
-	minrevallphases = len + 1;
-	phasessetlist = PyList_New(numphase);
-	if (phasessetlist == NULL)
-		goto done;
 
-	PyList_SET_ITEM(phasessetlist, 0, Py_None);
-	Py_INCREF(Py_None);
+	for (i = 0; i < numphases; ++i) {
+		pyphase = PyInt_FromLong(trackedphases[i]);
+		if (pyphase == NULL)
+			goto release;
+		phaseroots = PyDict_GetItem(roots, pyphase);
+		Py_DECREF(pyphase);
+		if (phaseroots == NULL)
+			continue;
+		rev = add_roots_get_min(self, phaseroots, phases, trackedphases[i]);
+		phaseroots = NULL;
+		if (rev == -2)
+			goto release;
+		if (rev != -1 && (minphaserev == -1 || rev < minphaserev))
+			minphaserev = rev;
+	}
 
-	for (i = 0; i < numphase - 1; i++) {
-		phaseroots = PyList_GET_ITEM(roots, i);
-		phaseset = PySet_New(NULL);
-		if (phaseset == NULL)
-			goto release;
-		PyList_SET_ITEM(phasessetlist, i + 1, phaseset);
-		if (!PyList_Check(phaseroots)) {
-			PyErr_SetString(PyExc_TypeError,
-			                "roots item must be a list");
+	for (i = 0; i < numphases; ++i) {
+		phasesets[i] = PySet_New(NULL);
+		if (phasesets[i] == NULL)
 			goto release;
 		}
-		minrevphase =
-		    add_roots_get_min(self, phaseroots, i + 1, phases);
-		if (minrevphase == -2) /* Error from add_roots_get_min */
-			goto release;
-		minrevallphases = MIN(minrevallphases, minrevphase);
-	}
-	/* Propagate the phase information from the roots to the revs */
-	if (minrevallphases != -1) {
+
+	if (minphaserev == -1)
+		minphaserev = len;
+	for (rev = minphaserev; rev < len; ++rev) {
 		int parents[2];
-		for (i = minrevallphases; i < len; i++) {
-			if (index_get_parents(self, i, parents, (int)len - 1) <
-			    0)
+		/*
+		 * The parent lookup could be skipped for phaseroots, but
+		 * phase --force would historically not recompute them
+		 * correctly, leaving descendents with a lower phase around.
+		 * As such, unconditionally recompute the phase.
+		 */
+		if (index_get_parents(self, rev, parents, (int)len - 1) < 0)
 				goto release;
-			set_phase_from_parents(phases, parents[0], parents[1],
-			                       i);
+		set_phase_from_parents(phases, parents[0], parents[1], rev);
+		switch (phases[rev]) {
+		case 0:
+			continue;
+		case 1:
+			pyphase = phasesets[0];
+			break;
+		case 2:
+			pyphase = phasesets[1];
+			break;
+		case 32:
+			pyphase = phasesets[2];
+			break;
+		case 96:
+			pyphase = phasesets[3];
+			break;
+		default:
+			goto release;
 		}
+		pyrev = PyInt_FromLong(rev);
+		if (pyrev == NULL)
+			goto release;
+		if (PySet_Add(pyphase, pyrev) == -1) {
+			Py_DECREF(pyrev);
+			goto release;
 	}
-	/* Transform phase list to a python list */
+		Py_DECREF(pyrev);
+	}
+	phaseroots = _dict_new_presized(numphases);
+	if (phaseroots == NULL)
+		goto release;
+	for (int i = 0; i < numphases; ++i) {
+		pyphase = PyInt_FromLong(trackedphases[i]);
+		if (pyphase == NULL)
+			goto release;
+		if (PyDict_SetItem(phaseroots, pyphase, phasesets[i]) == -1) {
+			Py_DECREF(pyphase);
+			goto release;
+		}
+		Py_DECREF(phasesets[i]);
+		phasesets[i] = NULL;
+	}
 	phasessize = PyInt_FromSsize_t(len);
 	if (phasessize == NULL)
 		goto release;
-	for (i = 0; i < len; i++) {
-		phase = phases[i];
-		/* We only store the sets of phase for non public phase, the
-		 * public phase is computed as a difference */
-		if (phase != 0) {
-			phaseset = PyList_GET_ITEM(phasessetlist, phase);
-			rev = PyInt_FromSsize_t(i);
-			if (rev == NULL)
-				goto release;
-			PySet_Add(phaseset, rev);
-			Py_XDECREF(rev);
-		}
-	}
-	ret = PyTuple_Pack(2, phasessize, phasessetlist);
+
+	ret = PyTuple_Pack(2, phasessize, phaseroots);
+	Py_DECREF(phasessize);
+	Py_DECREF(phaseroots);
+	return ret;
 
 release:
-	Py_XDECREF(phasessize);
-	Py_XDECREF(phasessetlist);
-done:
+	for (i = 0; i < numphases; ++i)
+		Py_XDECREF(phasesets[i]);
+	Py_XDECREF(phaseroots);
+
 	free(phases);
-	return ret;
+	return NULL;
 }
 
 static PyObject *index_headrevs(indexObject *self, PyObject *args)
 {
 	Py_ssize_t i, j, len;
 	char *nothead = NULL;
 	PyObject *heads = NULL;
 	PyObject *filter = NULL;
 	PyObject *filteredrevs = Py_None;
 
 	if (!PyArg_ParseTuple(args, "|O", &filteredrevs)) {
 		return NULL;
 	}
 
 	if (self->headrevs && filteredrevs == self->filteredrevs)
 		return list_copy(self->headrevs);
 
 	Py_DECREF(self->filteredrevs);
 	self->filteredrevs = filteredrevs;
 	Py_INCREF(filteredrevs);
 
 	if (filteredrevs != Py_None) {
 		filter = PyObject_GetAttrString(filteredrevs, "__contains__");
 		if (!filter) {
 			PyErr_SetString(
 			    PyExc_TypeError,
 			    "filteredrevs has no attribute __contains__");
 			goto bail;
 		}
 	}
 
 	len = index_length(self);
 	heads = PyList_New(0);
 	if (heads == NULL)
 		goto bail;
 	if (len == 0) {
 		PyObject *nullid = PyInt_FromLong(-1);
 		if (nullid == NULL || PyList_Append(heads, nullid) == -1) {
 			Py_XDECREF(nullid);
 			goto bail;
 		}
 		goto done;
 	}
 
 	nothead = calloc(len, 1);
 	if (nothead == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	for (i = len - 1; i >= 0; i--) {
 		int isfiltered;
 		int parents[2];
 
 		/* If nothead[i] == 1, it means we've seen an unfiltered child
 		 * of this node already, and therefore this node is not
 		 * filtered. So we can skip the expensive check_filter step.
 		 */
 		if (nothead[i] != 1) {
 			isfiltered = check_filter(filter, i);
 			if (isfiltered == -1) {
 				PyErr_SetString(PyExc_TypeError,
 				                "unable to check filter");
 				goto bail;
 			}
 
 			if (isfiltered) {
 				nothead[i] = 1;
 				continue;
 			}
 		}
 
 		if (index_get_parents(self, i, parents, (int)len - 1) < 0)
 			goto bail;
 		for (j = 0; j < 2; j++) {
 			if (parents[j] >= 0)
 				nothead[parents[j]] = 1;
 		}
 	}
 
 	for (i = 0; i < len; i++) {
 		PyObject *head;
 
 		if (nothead[i])
 			continue;
 		head = PyInt_FromSsize_t(i);
 		if (head == NULL || PyList_Append(heads, head) == -1) {
 			Py_XDECREF(head);
 			goto bail;
 		}
 	}
 
 done:
 	self->headrevs = heads;
 	Py_XDECREF(filter);
 	free(nothead);
 	return list_copy(self->headrevs);
 bail:
 	Py_XDECREF(filter);
 	Py_XDECREF(heads);
 	free(nothead);
 	return NULL;
 }
 
 /**
  * Obtain the base revision index entry.
  *
  * Callers must ensure that rev >= 0 or illegal memory access may occur.
  */
 static inline int index_baserev(indexObject *self, int rev)
 {
 	const char *data;
 	int result;
 
 	if (rev >= self->length) {
 		PyObject *tuple =
 		    PyList_GET_ITEM(self->added, rev - self->length);
 		long ret;
 		if (!pylong_to_long(PyTuple_GET_ITEM(tuple, 3), &ret)) {
 			return -2;
 		}
 		result = (int)ret;
 	} else {
 		data = index_deref(self, rev);
 		if (data == NULL) {
 			return -2;
 		}
 
 		result = getbe32(data + 16);
 	}
 	if (result > rev) {
 		PyErr_Format(
 		    PyExc_ValueError,
 		    "corrupted revlog, revision base above revision: %d, %d",
 		    rev, result);
 		return -2;
 	}
 	if (result < -1) {
 		PyErr_Format(
 		    PyExc_ValueError,
 		    "corrupted revlog, revision base out of range: %d, %d", rev,
 		    result);
 		return -2;
 	}
 	return result;
 }
 
 /**
  * Find if a revision is a snapshot or not
  *
  * Only relevant for sparse-revlog case.
  * Callers must ensure that rev is in a valid range.
  */
 static int index_issnapshotrev(indexObject *self, Py_ssize_t rev)
 {
 	int ps[2];
 	Py_ssize_t base;
 	while (rev >= 0) {
 		base = (Py_ssize_t)index_baserev(self, rev);
 		if (base == rev) {
 			base = -1;
 		}
 		if (base == -2) {
 			assert(PyErr_Occurred());
 			return -1;
 		}
 		if (base == -1) {
 			return 1;
 		}
 		if (index_get_parents(self, rev, ps, (int)rev) < 0) {
 			assert(PyErr_Occurred());
 			return -1;
 		};
 		if (base == ps[0] || base == ps[1]) {
 			return 0;
 		}
 		rev = base;
 	}
 	return rev == -1;
 }
 
 static PyObject *index_issnapshot(indexObject *self, PyObject *value)
 {
 	long rev;
 	int issnap;
 	Py_ssize_t length = index_length(self);
 
 	if (!pylong_to_long(value, &rev)) {
 		return NULL;
 	}
 	if (rev < -1 || rev >= length) {
 		PyErr_Format(PyExc_ValueError, "revlog index out of range: %ld",
 		             rev);
 		return NULL;
 	};
 	issnap = index_issnapshotrev(self, (Py_ssize_t)rev);
 	if (issnap < 0) {
 		return NULL;
 	};
 	return PyBool_FromLong((long)issnap);
 }
 
 static PyObject *index_findsnapshots(indexObject *self, PyObject *args)
 {
 	Py_ssize_t start_rev;
 	PyObject *cache;
 	Py_ssize_t base;
 	Py_ssize_t rev;
 	PyObject *key = NULL;
 	PyObject *value = NULL;
 	const Py_ssize_t length = index_length(self);
 	if (!PyArg_ParseTuple(args, "O!n", &PyDict_Type, &cache, &start_rev)) {
 		return NULL;
 	}
 	for (rev = start_rev; rev < length; rev++) {
 		int issnap;
 		PyObject *allvalues = NULL;
 		issnap = index_issnapshotrev(self, rev);
 		if (issnap < 0) {
 			goto bail;
 		}
 		if (issnap == 0) {
 			continue;
 		}
 		base = (Py_ssize_t)index_baserev(self, rev);
 		if (base == rev) {
 			base = -1;
 		}
 		if (base == -2) {
 			assert(PyErr_Occurred());
 			goto bail;
 		}
 		key = PyInt_FromSsize_t(base);
 		allvalues = PyDict_GetItem(cache, key);
 		if (allvalues == NULL && PyErr_Occurred()) {
 			goto bail;
 		}
 		if (allvalues == NULL) {
 			int r;
 			allvalues = PyList_New(0);
 			if (!allvalues) {
 				goto bail;
 			}
 			r = PyDict_SetItem(cache, key, allvalues);
 			Py_DECREF(allvalues);
 			if (r < 0) {
 				goto bail;
 			}
 		}
 		value = PyInt_FromSsize_t(rev);
 		if (PyList_Append(allvalues, value)) {
 			goto bail;
 		}
 		Py_CLEAR(key);
 		Py_CLEAR(value);
 	}
 	Py_RETURN_NONE;
 bail:
 	Py_XDECREF(key);
 	Py_XDECREF(value);
 	return NULL;
 }
 
 static PyObject *index_deltachain(indexObject *self, PyObject *args)
 {
 	int rev, generaldelta;
 	PyObject *stoparg;
 	int stoprev, iterrev, baserev = -1;
 	int stopped;
 	PyObject *chain = NULL, *result = NULL;
 	const Py_ssize_t length = index_length(self);
 
 	if (!PyArg_ParseTuple(args, "iOi", &rev, &stoparg, &generaldelta)) {
 		return NULL;
 	}
 
 	if (PyInt_Check(stoparg)) {
 		stoprev = (int)PyInt_AsLong(stoparg);
 		if (stoprev == -1 && PyErr_Occurred()) {
 			return NULL;
 		}
 	} else if (stoparg == Py_None) {
 		stoprev = -2;
 	} else {
 		PyErr_SetString(PyExc_ValueError,
 		                "stoprev must be integer or None");
 		return NULL;
 	}
 
 	if (rev < 0 || rev >= length) {
 		PyErr_SetString(PyExc_ValueError, "revlog index out of range");
 		return NULL;
 	}
 
 	chain = PyList_New(0);
 	if (chain == NULL) {
 		return NULL;
 	}
 
 	baserev = index_baserev(self, rev);
 
 	/* This should never happen. */
 	if (baserev <= -2) {
 		/* Error should be set by index_deref() */
 		assert(PyErr_Occurred());
 		goto bail;
 	}
 
 	iterrev = rev;
 
 	while (iterrev != baserev && iterrev != stoprev) {
 		PyObject *value = PyInt_FromLong(iterrev);
 		if (value == NULL) {
 			goto bail;
 		}
 		if (PyList_Append(chain, value)) {
 			Py_DECREF(value);
 			goto bail;
 		}
 		Py_DECREF(value);
 
 		if (generaldelta) {
 			iterrev = baserev;
 		} else {
 			iterrev--;
 		}
 
 		if (iterrev < 0) {
 			break;
 		}
 
 		if (iterrev >= length) {
 			PyErr_SetString(PyExc_IndexError,
 			                "revision outside index");
 			return NULL;
 		}
 
 		baserev = index_baserev(self, iterrev);
 
 		/* This should never happen. */
 		if (baserev <= -2) {
 			/* Error should be set by index_deref() */
 			assert(PyErr_Occurred());
 			goto bail;
 		}
 	}
 
 	if (iterrev == stoprev) {
 		stopped = 1;
 	} else {
 		PyObject *value = PyInt_FromLong(iterrev);
 		if (value == NULL) {
 			goto bail;
 		}
 		if (PyList_Append(chain, value)) {
 			Py_DECREF(value);
 			goto bail;
 		}
 		Py_DECREF(value);
 
 		stopped = 0;
 	}
 
 	if (PyList_Reverse(chain)) {
 		goto bail;
 	}
 
 	result = Py_BuildValue("OO", chain, stopped ? Py_True : Py_False);
 	Py_DECREF(chain);
 	return result;
 
 bail:
 	Py_DECREF(chain);
 	return NULL;
 }
 
 static inline int64_t
 index_segment_span(indexObject *self, Py_ssize_t start_rev, Py_ssize_t end_rev)
 {
 	int64_t start_offset;
 	int64_t end_offset;
 	int end_size;
 	start_offset = index_get_start(self, start_rev);
 	if (start_offset < 0) {
 		return -1;
 	}
 	end_offset = index_get_start(self, end_rev);
 	if (end_offset < 0) {
 		return -1;
 	}
 	end_size = index_get_length(self, end_rev);
 	if (end_size < 0) {
 		return -1;
 	}
 	if (end_offset < start_offset) {
 		PyErr_Format(PyExc_ValueError,
 		             "corrupted revlog index: inconsistent offset "
 		             "between revisions (%zd) and (%zd)",
 		             start_rev, end_rev);
 		return -1;
 	}
 	return (end_offset - start_offset) + (int64_t)end_size;
 }
 
 /* returns endidx so that revs[startidx:endidx] has no empty trailing revs */
 static Py_ssize_t trim_endidx(indexObject *self, const Py_ssize_t *revs,
                               Py_ssize_t startidx, Py_ssize_t endidx)
 {
 	int length;
 	while (endidx > 1 && endidx > startidx) {
 		length = index_get_length(self, revs[endidx - 1]);
 		if (length < 0) {
 			return -1;
 		}
 		if (length != 0) {
 			break;
 		}
 		endidx -= 1;
 	}
 	return endidx;
 }
 
 struct Gap {
 	int64_t size;
 	Py_ssize_t idx;
 };
 
 static int gap_compare(const void *left, const void *right)
 {
 	const struct Gap *l_left = ((const struct Gap *)left);
 	const struct Gap *l_right = ((const struct Gap *)right);
 	if (l_left->size < l_right->size) {
 		return -1;
 	} else if (l_left->size > l_right->size) {
 		return 1;
 	}
 	return 0;
 }
 static int Py_ssize_t_compare(const void *left, const void *right)
 {
 	const Py_ssize_t l_left = *(const Py_ssize_t *)left;
 	const Py_ssize_t l_right = *(const Py_ssize_t *)right;
 	if (l_left < l_right) {
 		return -1;
 	} else if (l_left > l_right) {
 		return 1;
 	}
 	return 0;
 }
 
 static PyObject *index_slicechunktodensity(indexObject *self, PyObject *args)
 {
 	/* method arguments */
 	PyObject *list_revs = NULL; /* revisions in the chain */
 	double targetdensity = 0;   /* min density to achieve */
 	Py_ssize_t mingapsize = 0;  /* threshold to ignore gaps */
 
 	/* other core variables */
 	Py_ssize_t idxlen = index_length(self);
 	Py_ssize_t i;            /* used for various iteration */
 	PyObject *result = NULL; /* the final return of the function */
 
 	/* generic information about the delta chain being slice */
 	Py_ssize_t num_revs = 0;    /* size of the full delta chain */
 	Py_ssize_t *revs = NULL;    /* native array of revision in the chain */
 	int64_t chainpayload = 0;   /* sum of all delta in the chain */
 	int64_t deltachainspan = 0; /* distance from first byte to last byte */
 
 	/* variable used for slicing the delta chain */
 	int64_t readdata = 0; /* amount of data currently planned to be read */
 	double density = 0;   /* ration of payload data compared to read ones */
 	int64_t previous_end;
 	struct Gap *gaps = NULL; /* array of notable gap in the chain */
 	Py_ssize_t num_gaps =
 	    0; /* total number of notable gap recorded so far */
 	Py_ssize_t *selected_indices = NULL; /* indices of gap skipped over */
 	Py_ssize_t num_selected = 0;         /* number of gaps skipped */
 	PyObject *chunk = NULL;              /* individual slice */
 	PyObject *allchunks = NULL;          /* all slices */
 	Py_ssize_t previdx;
 
 	/* parsing argument */
 	if (!PyArg_ParseTuple(args, "O!dn", &PyList_Type, &list_revs,
 	                      &targetdensity, &mingapsize)) {
 		goto bail;
 	}
 
 	/* If the delta chain contains a single element, we do not need slicing
 	 */
 	num_revs = PyList_GET_SIZE(list_revs);
 	if (num_revs <= 1) {
 		result = PyTuple_Pack(1, list_revs);
 		goto done;
 	}
 
 	/* Turn the python list into a native integer array (for efficiency) */
 	revs = (Py_ssize_t *)calloc(num_revs, sizeof(Py_ssize_t));
 	if (revs == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 	for (i = 0; i < num_revs; i++) {
 		Py_ssize_t revnum = PyInt_AsLong(PyList_GET_ITEM(list_revs, i));
 		if (revnum == -1 && PyErr_Occurred()) {
 			goto bail;
 		}
 		if (revnum < nullrev || revnum >= idxlen) {
 			PyErr_Format(PyExc_IndexError,
 			             "index out of range: %zd", revnum);
 			goto bail;
 		}
 		revs[i] = revnum;
 	}
 
 	/* Compute and check various property of the unsliced delta chain */
 	deltachainspan = index_segment_span(self, revs[0], revs[num_revs - 1]);
 	if (deltachainspan < 0) {
 		goto bail;
 	}
 
 	if (deltachainspan <= mingapsize) {
 		result = PyTuple_Pack(1, list_revs);
 		goto done;
 	}
 	chainpayload = 0;
 	for (i = 0; i < num_revs; i++) {
 		int tmp = index_get_length(self, revs[i]);
 		if (tmp < 0) {
 			goto bail;
 		}
 		chainpayload += tmp;
 	}
 
 	readdata = deltachainspan;
 	density = 1.0;
 
 	if (0 < deltachainspan) {
 		density = (double)chainpayload / (double)deltachainspan;
 	}
 
 	if (density >= targetdensity) {
 		result = PyTuple_Pack(1, list_revs);
 		goto done;
 	}
 
 	/* if chain is too sparse, look for relevant gaps */
 	gaps = (struct Gap *)calloc(num_revs, sizeof(struct Gap));
 	if (gaps == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	previous_end = -1;
 	for (i = 0; i < num_revs; i++) {
 		int64_t revstart;
 		int revsize;
 		revstart = index_get_start(self, revs[i]);
 		if (revstart < 0) {
 			goto bail;
 		};
 		revsize = index_get_length(self, revs[i]);
 		if (revsize < 0) {
 			goto bail;
 		};
 		if (revsize == 0) {
 			continue;
 		}
 		if (previous_end >= 0) {
 			int64_t gapsize = revstart - previous_end;
 			if (gapsize > mingapsize) {
 				gaps[num_gaps].size = gapsize;
 				gaps[num_gaps].idx = i;
 				num_gaps += 1;
 			}
 		}
 		previous_end = revstart + revsize;
 	}
 	if (num_gaps == 0) {
 		result = PyTuple_Pack(1, list_revs);
 		goto done;
 	}
 	qsort(gaps, num_gaps, sizeof(struct Gap), &gap_compare);
 
 	/* Slice the largest gap first, they improve the density the most */
 	selected_indices =
 	    (Py_ssize_t *)malloc((num_gaps + 1) * sizeof(Py_ssize_t));
 	if (selected_indices == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	for (i = num_gaps - 1; i >= 0; i--) {
 		selected_indices[num_selected] = gaps[i].idx;
 		readdata -= gaps[i].size;
 		num_selected += 1;
 		if (readdata <= 0) {
 			density = 1.0;
 		} else {
 			density = (double)chainpayload / (double)readdata;
 		}
 		if (density >= targetdensity) {
 			break;
 		}
 	}
 	qsort(selected_indices, num_selected, sizeof(Py_ssize_t),
 	      &Py_ssize_t_compare);
 
 	/* create the resulting slice */
 	allchunks = PyList_New(0);
 	if (allchunks == NULL) {
 		goto bail;
 	}
 	previdx = 0;
 	selected_indices[num_selected] = num_revs;
 	for (i = 0; i <= num_selected; i++) {
 		Py_ssize_t idx = selected_indices[i];
 		Py_ssize_t endidx = trim_endidx(self, revs, previdx, idx);
 		if (endidx < 0) {
 			goto bail;
 		}
 		if (previdx < endidx) {
 			chunk = PyList_GetSlice(list_revs, previdx, endidx);
 			if (chunk == NULL) {
 				goto bail;
 			}
 			if (PyList_Append(allchunks, chunk) == -1) {
 				goto bail;
 			}
 			Py_DECREF(chunk);
 			chunk = NULL;
 		}
 		previdx = idx;
 	}
 	result = allchunks;
 	goto done;
 
 bail:
 	Py_XDECREF(allchunks);
 	Py_XDECREF(chunk);
 done:
 	free(revs);
 	free(gaps);
 	free(selected_indices);
 	return result;
 }
 
 static inline int nt_level(const char *node, Py_ssize_t level)
 {
 	int v = node[level >> 1];
 	if (!(level & 1))
 		v >>= 4;
 	return v & 0xf;
 }
 
 /*
  * Return values:
  *
  *   -4: match is ambiguous (multiple candidates)
  *   -2: not found
  * rest: valid rev
  */
 static int nt_find(nodetree *self, const char *node, Py_ssize_t nodelen,
                    int hex)
 {
 	int (*getnybble)(const char *, Py_ssize_t) = hex ? hexdigit : nt_level;
 	int level, maxlevel, off;
 
 	if (nodelen == 20 && node[0] == '\0' && memcmp(node, nullid, 20) == 0)
 		return -1;
 
 	if (hex)
 		maxlevel = nodelen > 40 ? 40 : (int)nodelen;
 	else
 		maxlevel = nodelen > 20 ? 40 : ((int)nodelen * 2);
 
 	for (level = off = 0; level < maxlevel; level++) {
 		int k = getnybble(node, level);
 		nodetreenode *n = &self->nodes[off];
 		int v = n->children[k];
 
 		if (v < 0) {
 			const char *n;
 			Py_ssize_t i;
 
 			v = -(v + 2);
 			n = index_node(self->index, v);
 			if (n == NULL)
 				return -2;
 			for (i = level; i < maxlevel; i++)
 				if (getnybble(node, i) != nt_level(n, i))
 					return -2;
 			return v;
 		}
 		if (v == 0)
 			return -2;
 		off = v;
 	}
 	/* multiple matches against an ambiguous prefix */
 	return -4;
 }
 
 static int nt_new(nodetree *self)
 {
 	if (self->length == self->capacity) {
 		unsigned newcapacity;
 		nodetreenode *newnodes;
 		newcapacity = self->capacity * 2;
 		if (newcapacity >= INT_MAX / sizeof(nodetreenode)) {
 			PyErr_SetString(PyExc_MemoryError,
 			                "overflow in nt_new");
 			return -1;
 		}
 		newnodes =
 		    realloc(self->nodes, newcapacity * sizeof(nodetreenode));
 		if (newnodes == NULL) {
 			PyErr_SetString(PyExc_MemoryError, "out of memory");
 			return -1;
 		}
 		self->capacity = newcapacity;
 		self->nodes = newnodes;
 		memset(&self->nodes[self->length], 0,
 		       sizeof(nodetreenode) * (self->capacity - self->length));
 	}
 	return self->length++;
 }
 
 static int nt_insert(nodetree *self, const char *node, int rev)
 {
 	int level = 0;
 	int off = 0;
 
 	while (level < 40) {
 		int k = nt_level(node, level);
 		nodetreenode *n;
 		int v;
 
 		n = &self->nodes[off];
 		v = n->children[k];
 
 		if (v == 0) {
 			n->children[k] = -rev - 2;
 			return 0;
 		}
 		if (v < 0) {
 			const char *oldnode =
 			    index_node_existing(self->index, -(v + 2));
 			int noff;
 
 			if (oldnode == NULL)
 				return -1;
 			if (!memcmp(oldnode, node, 20)) {
 				n->children[k] = -rev - 2;
 				return 0;
 			}
 			noff = nt_new(self);
 			if (noff == -1)
 				return -1;
 			/* self->nodes may have been changed by realloc */
 			self->nodes[off].children[k] = noff;
 			off = noff;
 			n = &self->nodes[off];
 			n->children[nt_level(oldnode, ++level)] = v;
 			if (level > self->depth)
 				self->depth = level;
 			self->splits += 1;
 		} else {
 			level += 1;
 			off = v;
 		}
 	}
 
 	return -1;
 }
 
 static PyObject *ntobj_insert(nodetreeObject *self, PyObject *args)
 {
 	Py_ssize_t rev;
 	const char *node;
 	Py_ssize_t length;
 	if (!PyArg_ParseTuple(args, "n", &rev))
 		return NULL;
 	length = index_length(self->nt.index);
 	if (rev < 0 || rev >= length) {
 		PyErr_SetString(PyExc_ValueError, "revlog index out of range");
 		return NULL;
 	}
 	node = index_node_existing(self->nt.index, rev);
 	if (nt_insert(&self->nt, node, (int)rev) == -1)
 		return NULL;
 	Py_RETURN_NONE;
 }
 
 static int nt_delete_node(nodetree *self, const char *node)
 {
 	/* rev==-2 happens to get encoded as 0, which is interpreted as not set
 	 */
 	return nt_insert(self, node, -2);
 }
 
 static int nt_init(nodetree *self, indexObject *index, unsigned capacity)
 {
 	/* Initialize before overflow-checking to avoid nt_dealloc() crash. */
 	self->nodes = NULL;
 
 	self->index = index;
 	/* The input capacity is in terms of revisions, while the field is in
 	 * terms of nodetree nodes. */
 	self->capacity = (capacity < 4 ? 4 : capacity / 2);
 	self->depth = 0;
 	self->splits = 0;
 	if ((size_t)self->capacity > INT_MAX / sizeof(nodetreenode)) {
 		PyErr_SetString(PyExc_ValueError, "overflow in init_nt");
 		return -1;
 	}
 	self->nodes = calloc(self->capacity, sizeof(nodetreenode));
 	if (self->nodes == NULL) {
 		PyErr_NoMemory();
 		return -1;
 	}
 	self->length = 1;
 	return 0;
 }
 
 static int ntobj_init(nodetreeObject *self, PyObject *args)
 {
 	PyObject *index;
 	unsigned capacity;
 	if (!PyArg_ParseTuple(args, "O!I", &HgRevlogIndex_Type, &index,
 	                      &capacity))
 		return -1;
 	Py_INCREF(index);
 	return nt_init(&self->nt, (indexObject *)index, capacity);
 }
 
 static int nt_partialmatch(nodetree *self, const char *node, Py_ssize_t nodelen)
 {
 	return nt_find(self, node, nodelen, 1);
 }
 
 /*
  * Find the length of the shortest unique prefix of node.
  *
  * Return values:
  *
  *   -3: error (exception set)
  *   -2: not found (no exception set)
  * rest: length of shortest prefix
  */
 static int nt_shortest(nodetree *self, const char *node)
 {
 	int level, off;
 
 	for (level = off = 0; level < 40; level++) {
 		int k, v;
 		nodetreenode *n = &self->nodes[off];
 		k = nt_level(node, level);
 		v = n->children[k];
 		if (v < 0) {
 			const char *n;
 			v = -(v + 2);
 			n = index_node_existing(self->index, v);
 			if (n == NULL)
 				return -3;
 			if (memcmp(node, n, 20) != 0)
 				/*
 				 * Found a unique prefix, but it wasn't for the
 				 * requested node (i.e the requested node does
 				 * not exist).
 				 */
 				return -2;
 			return level + 1;
 		}
 		if (v == 0)
 			return -2;
 		off = v;
 	}
 	/*
 	 * The node was still not unique after 40 hex digits, so this won't
 	 * happen. Also, if we get here, then there's a programming error in
 	 * this file that made us insert a node longer than 40 hex digits.
 	 */
 	PyErr_SetString(PyExc_Exception, "broken node tree");
 	return -3;
 }
 
 static PyObject *ntobj_shortest(nodetreeObject *self, PyObject *args)
 {
 	PyObject *val;
 	char *node;
 	int length;
 
 	if (!PyArg_ParseTuple(args, "O", &val))
 		return NULL;
 	if (node_check(val, &node) == -1)
 		return NULL;
 
 	length = nt_shortest(&self->nt, node);
 	if (length == -3)
 		return NULL;
 	if (length == -2) {
 		raise_revlog_error();
 		return NULL;
 	}
 	return PyInt_FromLong(length);
 }
 
 static void nt_dealloc(nodetree *self)
 {
 	free(self->nodes);
 	self->nodes = NULL;
 }
 
 static void ntobj_dealloc(nodetreeObject *self)
 {
 	Py_XDECREF(self->nt.index);
 	nt_dealloc(&self->nt);
 	PyObject_Del(self);
 }
 
 static PyMethodDef ntobj_methods[] = {
     {"insert", (PyCFunction)ntobj_insert, METH_VARARGS,
      "insert an index entry"},
     {"shortest", (PyCFunction)ntobj_shortest, METH_VARARGS,
      "find length of shortest hex nodeid of a binary ID"},
     {NULL} /* Sentinel */
 };
 
 static PyTypeObject nodetreeType = {
     PyVarObject_HEAD_INIT(NULL, 0) /* header */
     "parsers.nodetree",            /* tp_name */
     sizeof(nodetreeObject),        /* tp_basicsize */
     0,                             /* tp_itemsize */
     (destructor)ntobj_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 */
     "nodetree",                    /* tp_doc */
     0,                             /* tp_traverse */
     0,                             /* tp_clear */
     0,                             /* tp_richcompare */
     0,                             /* tp_weaklistoffset */
     0,                             /* tp_iter */
     0,                             /* tp_iternext */
     ntobj_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)ntobj_init,          /* tp_init */
     0,                             /* tp_alloc */
 };
 
 static int index_init_nt(indexObject *self)
 {
 	if (!self->ntinitialized) {
 		if (nt_init(&self->nt, self, (int)self->raw_length) == -1) {
 			nt_dealloc(&self->nt);
 			return -1;
 		}
 		if (nt_insert(&self->nt, nullid, -1) == -1) {
 			nt_dealloc(&self->nt);
 			return -1;
 		}
 		self->ntinitialized = 1;
 		self->ntrev = (int)index_length(self);
 		self->ntlookups = 1;
 		self->ntmisses = 0;
 	}
 	return 0;
 }
 
 /*
  * Return values:
  *
  *   -3: error (exception set)
  *   -2: not found (no exception set)
  * rest: valid rev
  */
 static int index_find_node(indexObject *self, const char *node,
                            Py_ssize_t nodelen)
 {
 	int rev;
 
 	if (index_init_nt(self) == -1)
 		return -3;
 
 	self->ntlookups++;
 	rev = nt_find(&self->nt, node, nodelen, 0);
 	if (rev >= -1)
 		return rev;
 
 	/*
 	 * For the first handful of lookups, we scan the entire index,
 	 * and cache only the matching nodes. This optimizes for cases
 	 * like "hg tip", where only a few nodes are accessed.
 	 *
 	 * After that, we cache every node we visit, using a single
 	 * scan amortized over multiple lookups.  This gives the best
 	 * bulk performance, e.g. for "hg log".
 	 */
 	if (self->ntmisses++ < 4) {
 		for (rev = self->ntrev - 1; rev >= 0; rev--) {
 			const char *n = index_node_existing(self, rev);
 			if (n == NULL)
 				return -3;
 			if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
 				if (nt_insert(&self->nt, n, rev) == -1)
 					return -3;
 				break;
 			}
 		}
 	} else {
 		for (rev = self->ntrev - 1; rev >= 0; rev--) {
 			const char *n = index_node_existing(self, rev);
 			if (n == NULL)
 				return -3;
 			if (nt_insert(&self->nt, n, rev) == -1) {
 				self->ntrev = rev + 1;
 				return -3;
 			}
 			if (memcmp(node, n, nodelen > 20 ? 20 : nodelen) == 0) {
 				break;
 			}
 		}
 		self->ntrev = rev;
 	}
 
 	if (rev >= 0)
 		return rev;
 	return -2;
 }
 
 static PyObject *index_getitem(indexObject *self, PyObject *value)
 {
 	char *node;
 	int rev;
 
 	if (PyInt_Check(value)) {
 		long idx;
 		if (!pylong_to_long(value, &idx)) {
 			return NULL;
 		}
 		return index_get(self, idx);
 	}
 
 	if (node_check(value, &node) == -1)
 		return NULL;
 	rev = index_find_node(self, node, 20);
 	if (rev >= -1)
 		return PyInt_FromLong(rev);
 	if (rev == -2)
 		raise_revlog_error();
 	return NULL;
 }
 
 /*
  * Fully populate the radix tree.
  */
 static int index_populate_nt(indexObject *self)
 {
 	int rev;
 	if (self->ntrev > 0) {
 		for (rev = self->ntrev - 1; rev >= 0; rev--) {
 			const char *n = index_node_existing(self, rev);
 			if (n == NULL)
 				return -1;
 			if (nt_insert(&self->nt, n, rev) == -1)
 				return -1;
 		}
 		self->ntrev = -1;
 	}
 	return 0;
 }
 
 static PyObject *index_partialmatch(indexObject *self, PyObject *args)
 {
 	const char *fullnode;
 	Py_ssize_t nodelen;
 	char *node;
 	int rev, i;
 
 	if (!PyArg_ParseTuple(args, PY23("s#", "y#"), &node, &nodelen))
 		return NULL;
 
 	if (nodelen < 1) {
 		PyErr_SetString(PyExc_ValueError, "key too short");
 		return NULL;
 	}
 
 	if (nodelen > 40) {
 		PyErr_SetString(PyExc_ValueError, "key too long");
 		return NULL;
 	}
 
 	for (i = 0; i < nodelen; i++)
 		hexdigit(node, i);
 	if (PyErr_Occurred()) {
 		/* input contains non-hex characters */
 		PyErr_Clear();
 		Py_RETURN_NONE;
 	}
 
 	if (index_init_nt(self) == -1)
 		return NULL;
 	if (index_populate_nt(self) == -1)
 		return NULL;
 	rev = nt_partialmatch(&self->nt, node, nodelen);
 
 	switch (rev) {
 	case -4:
 		raise_revlog_error();
 		return NULL;
 	case -2:
 		Py_RETURN_NONE;
 	case -1:
 		return PyBytes_FromStringAndSize(nullid, 20);
 	}
 
 	fullnode = index_node_existing(self, rev);
 	if (fullnode == NULL) {
 		return NULL;
 	}
 	return PyBytes_FromStringAndSize(fullnode, 20);
 }
 
 static PyObject *index_shortest(indexObject *self, PyObject *args)
 {
 	PyObject *val;
 	char *node;
 	int length;
 
 	if (!PyArg_ParseTuple(args, "O", &val))
 		return NULL;
 	if (node_check(val, &node) == -1)
 		return NULL;
 
 	self->ntlookups++;
 	if (index_init_nt(self) == -1)
 		return NULL;
 	if (index_populate_nt(self) == -1)
 		return NULL;
 	length = nt_shortest(&self->nt, node);
 	if (length == -3)
 		return NULL;
 	if (length == -2) {
 		raise_revlog_error();
 		return NULL;
 	}
 	return PyInt_FromLong(length);
 }
 
 static PyObject *index_m_get(indexObject *self, PyObject *args)
 {
 	PyObject *val;
 	char *node;
 	int rev;
 
 	if (!PyArg_ParseTuple(args, "O", &val))
 		return NULL;
 	if (node_check(val, &node) == -1)
 		return NULL;
 	rev = index_find_node(self, node, 20);
 	if (rev == -3)
 		return NULL;
 	if (rev == -2)
 		Py_RETURN_NONE;
 	return PyInt_FromLong(rev);
 }
 
 static int index_contains(indexObject *self, PyObject *value)
 {
 	char *node;
 
 	if (PyInt_Check(value)) {
 		long rev;
 		if (!pylong_to_long(value, &rev)) {
 			return -1;
 		}
 		return rev >= -1 && rev < index_length(self);
 	}
 
 	if (node_check(value, &node) == -1)
 		return -1;
 
 	switch (index_find_node(self, node, 20)) {
 	case -3:
 		return -1;
 	case -2:
 		return 0;
 	default:
 		return 1;
 	}
 }
 
 static PyObject *index_m_has_node(indexObject *self, PyObject *args)
 {
 	int ret = index_contains(self, args);
 	if (ret < 0)
 		return NULL;
 	return PyBool_FromLong((long)ret);
 }
 
 static PyObject *index_m_rev(indexObject *self, PyObject *val)
 {
 	char *node;
 	int rev;
 
 	if (node_check(val, &node) == -1)
 		return NULL;
 	rev = index_find_node(self, node, 20);
 	if (rev >= -1)
 		return PyInt_FromLong(rev);
 	if (rev == -2)
 		raise_revlog_error();
 	return NULL;
 }
 
 typedef uint64_t bitmask;
 
 /*
  * Given a disjoint set of revs, return all candidates for the
  * greatest common ancestor. In revset notation, this is the set
  * "heads(::a and ::b and ...)"
  */
 static PyObject *find_gca_candidates(indexObject *self, const int *revs,
                                      int revcount)
 {
 	const bitmask allseen = (1ull << revcount) - 1;
 	const bitmask poison = 1ull << revcount;
 	PyObject *gca = PyList_New(0);
 	int i, v, interesting;
 	int maxrev = -1;
 	bitmask sp;
 	bitmask *seen;
 
 	if (gca == NULL)
 		return PyErr_NoMemory();
 
 	for (i = 0; i < revcount; i++) {
 		if (revs[i] > maxrev)
 			maxrev = revs[i];
 	}
 
 	seen = calloc(sizeof(*seen), maxrev + 1);
 	if (seen == NULL) {
 		Py_DECREF(gca);
 		return PyErr_NoMemory();
 	}
 
 	for (i = 0; i < revcount; i++)
 		seen[revs[i]] = 1ull << i;
 
 	interesting = revcount;
 
 	for (v = maxrev; v >= 0 && interesting; v--) {
 		bitmask sv = seen[v];
 		int parents[2];
 
 		if (!sv)
 			continue;
 
 		if (sv < poison) {
 			interesting -= 1;
 			if (sv == allseen) {
 				PyObject *obj = PyInt_FromLong(v);
 				if (obj == NULL)
 					goto bail;
 				if (PyList_Append(gca, obj) == -1) {
 					Py_DECREF(obj);
 					goto bail;
 				}
 				sv |= poison;
 				for (i = 0; i < revcount; i++) {
 					if (revs[i] == v)
 						goto done;
 				}
 			}
 		}
 		if (index_get_parents(self, v, parents, maxrev) < 0)
 			goto bail;
 
 		for (i = 0; i < 2; i++) {
 			int p = parents[i];
 			if (p == -1)
 				continue;
 			sp = seen[p];
 			if (sv < poison) {
 				if (sp == 0) {
 					seen[p] = sv;
 					interesting++;
 				} else if (sp != sv)
 					seen[p] |= sv;
 			} else {
 				if (sp && sp < poison)
 					interesting--;
 				seen[p] = sv;
 			}
 		}
 	}
 
 done:
 	free(seen);
 	return gca;
 bail:
 	free(seen);
 	Py_XDECREF(gca);
 	return NULL;
 }
 
 /*
  * Given a disjoint set of revs, return the subset with the longest
  * path to the root.
  */
 static PyObject *find_deepest(indexObject *self, PyObject *revs)
 {
 	const Py_ssize_t revcount = PyList_GET_SIZE(revs);
 	static const Py_ssize_t capacity = 24;
 	int *depth, *interesting = NULL;
 	int i, j, v, ninteresting;
 	PyObject *dict = NULL, *keys = NULL;
 	long *seen = NULL;
 	int maxrev = -1;
 	long final;
 
 	if (revcount > capacity) {
 		PyErr_Format(PyExc_OverflowError,
 		             "bitset size (%ld) > capacity (%ld)",
 		             (long)revcount, (long)capacity);
 		return NULL;
 	}
 
 	for (i = 0; i < revcount; i++) {
 		int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
 		if (n > maxrev)
 			maxrev = n;
 	}
 
 	depth = calloc(sizeof(*depth), maxrev + 1);
 	if (depth == NULL)
 		return PyErr_NoMemory();
 
 	seen = calloc(sizeof(*seen), maxrev + 1);
 	if (seen == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	interesting = calloc(sizeof(*interesting), ((size_t)1) << revcount);
 	if (interesting == NULL) {
 		PyErr_NoMemory();
 		goto bail;
 	}
 
 	if (PyList_Sort(revs) == -1)
 		goto bail;
 
 	for (i = 0; i < revcount; i++) {
 		int n = (int)PyInt_AsLong(PyList_GET_ITEM(revs, i));
 		long b = 1l << i;
 		depth[n] = 1;
 		seen[n] = b;
 		interesting[b] = 1;
 	}
 
 	/* invariant: ninteresting is the number of non-zero entries in
 	 * interesting. */
 	ninteresting = (int)revcount;
 
 	for (v = maxrev; v >= 0 && ninteresting > 1; v--) {
 		int dv = depth[v];
 		int parents[2];
 		long sv;
 
 		if (dv == 0)
 			continue;
 
 		sv = seen[v];
 		if (index_get_parents(self, v, parents, maxrev) < 0)
 			goto bail;
 
 		for (i = 0; i < 2; i++) {
 			int p = parents[i];
 			long sp;
 			int dp;
 
 			if (p == -1)
 				continue;
 
 			dp = depth[p];
 			sp = seen[p];
 			if (dp <= dv) {
 				depth[p] = dv + 1;
 				if (sp != sv) {
 					interesting[sv] += 1;
 					seen[p] = sv;
 					if (sp) {
 						interesting[sp] -= 1;
 						if (interesting[sp] == 0)
 							ninteresting -= 1;
 					}
 				}
 			} else if (dv == dp - 1) {
 				long nsp = sp | sv;
 				if (nsp == sp)
 					continue;
 				seen[p] = nsp;
 				interesting[sp] -= 1;
 				if (interesting[sp] == 0)
 					ninteresting -= 1;
 				if (interesting[nsp] == 0)
 					ninteresting += 1;
 				interesting[nsp] += 1;
 			}
 		}
 		interesting[sv] -= 1;
 		if (interesting[sv] == 0)
 			ninteresting -= 1;
 	}
 
 	final = 0;
 	j = ninteresting;
 	for (i = 0; i < (int)(2 << revcount) && j > 0; i++) {
 		if (interesting[i] == 0)
 			continue;
 		final |= i;
 		j -= 1;
 	}
 	if (final == 0) {
 		keys = PyList_New(0);
 		goto bail;
 	}
 
 	dict = PyDict_New();
 	if (dict == NULL)
 		goto bail;
 
 	for (i = 0; i < revcount; i++) {
 		PyObject *key;
 
 		if ((final & (1 << i)) == 0)
 			continue;
 
 		key = PyList_GET_ITEM(revs, i);
 		Py_INCREF(key);
 		Py_INCREF(Py_None);
 		if (PyDict_SetItem(dict, key, Py_None) == -1) {
 			Py_DECREF(key);
 			Py_DECREF(Py_None);
 			goto bail;
 		}
 	}
 
 	keys = PyDict_Keys(dict);
 
 bail:
 	free(depth);
 	free(seen);
 	free(interesting);
 	Py_XDECREF(dict);
 
 	return keys;
 }
 
 /*
  * Given a (possibly overlapping) set of revs, return all the
  * common ancestors heads: heads(::args[0] and ::a[1] and ...)
  */
 static PyObject *index_commonancestorsheads(indexObject *self, PyObject *args)
 {
 	PyObject *ret = NULL;
 	Py_ssize_t argcount, i, len;
 	bitmask repeat = 0;
 	int revcount = 0;
 	int *revs;
 
 	argcount = PySequence_Length(args);
 	revs = PyMem_Malloc(argcount * sizeof(*revs));
 	if (argcount > 0 && revs == NULL)
 		return PyErr_NoMemory();
 	len = index_length(self);
 
 	for (i = 0; i < argcount; i++) {
 		static const int capacity = 24;
 		PyObject *obj = PySequence_GetItem(args, i);
 		bitmask x;
 		long val;
 
 		if (!PyInt_Check(obj)) {
 			PyErr_SetString(PyExc_TypeError,
 			                "arguments must all be ints");
 			Py_DECREF(obj);
 			goto bail;
 		}
 		val = PyInt_AsLong(obj);
 		Py_DECREF(obj);
 		if (val == -1) {
 			ret = PyList_New(0);
 			goto done;
 		}
 		if (val < 0 || val >= len) {
 			PyErr_SetString(PyExc_IndexError, "index out of range");
 			goto bail;
 		}
 		/* this cheesy bloom filter lets us avoid some more
 		 * expensive duplicate checks in the common set-is-disjoint
 		 * case */
 		x = 1ull << (val & 0x3f);
 		if (repeat & x) {
 			int k;
 			for (k = 0; k < revcount; k++) {
 				if (val == revs[k])
 					goto duplicate;
 			}
 		} else
 			repeat |= x;
 		if (revcount >= capacity) {
 			PyErr_Format(PyExc_OverflowError,
 			             "bitset size (%d) > capacity (%d)",
 			             revcount, capacity);
 			goto bail;
 		}
 		revs[revcount++] = (int)val;
 	duplicate:;
 	}
 
 	if (revcount == 0) {
 		ret = PyList_New(0);
 		goto done;
 	}
 	if (revcount == 1) {
 		PyObject *obj;
 		ret = PyList_New(1);
 		if (ret == NULL)
 			goto bail;
 		obj = PyInt_FromLong(revs[0]);
 		if (obj == NULL)
 			goto bail;
 		PyList_SET_ITEM(ret, 0, obj);
 		goto done;
 	}
 
 	ret = find_gca_candidates(self, revs, revcount);
 	if (ret == NULL)
 		goto bail;
 
 done:
 	PyMem_Free(revs);
 	return ret;
 
 bail:
 	PyMem_Free(revs);
 	Py_XDECREF(ret);
 	return NULL;
 }
 
 /*
  * Given a (possibly overlapping) set of revs, return the greatest
  * common ancestors: those with the longest path to the root.
  */
 static PyObject *index_ancestors(indexObject *self, PyObject *args)
 {
 	PyObject *ret;
 	PyObject *gca = index_commonancestorsheads(self, args);
 	if (gca == NULL)
 		return NULL;
 
 	if (PyList_GET_SIZE(gca) <= 1) {
 		return gca;
 	}
 
 	ret = find_deepest(self, gca);
 	Py_DECREF(gca);
 	return ret;
 }
 
 /*
  * Invalidate any trie entries introduced by added revs.
  */
 static void index_invalidate_added(indexObject *self, Py_ssize_t start)
 {
 	Py_ssize_t i, len = PyList_GET_SIZE(self->added);
 
 	for (i = start; i < len; i++) {
 		PyObject *tuple = PyList_GET_ITEM(self->added, i);
 		PyObject *node = PyTuple_GET_ITEM(tuple, 7);
 
 		nt_delete_node(&self->nt, PyBytes_AS_STRING(node));
 	}
 
 	if (start == 0)
 		Py_CLEAR(self->added);
 }
 
 /*
  * Delete a numeric range of revs, which must be at the end of the
  * range.
  */
 static int index_slice_del(indexObject *self, PyObject *item)
 {
 	Py_ssize_t start, stop, step, slicelength;
 	Py_ssize_t length = index_length(self) + 1;
 	int ret = 0;
 
 /* Argument changed from PySliceObject* to PyObject* in Python 3. */
 #ifdef IS_PY3K
 	if (PySlice_GetIndicesEx(item, length, &start, &stop, &step,
 	                         &slicelength) < 0)
 #else
 	if (PySlice_GetIndicesEx((PySliceObject *)item, length, &start, &stop,
 	                         &step, &slicelength) < 0)
 #endif
 		return -1;
 
 	if (slicelength <= 0)
 		return 0;
 
 	if ((step < 0 && start < stop) || (step > 0 && start > stop))
 		stop = start;
 
 	if (step < 0) {
 		stop = start + 1;
 		start = stop + step * (slicelength - 1) - 1;
 		step = -step;
 	}
 
 	if (step != 1) {
 		PyErr_SetString(PyExc_ValueError,
 		                "revlog index delete requires step size of 1");
 		return -1;
 	}
 
 	if (stop != length - 1) {
 		PyErr_SetString(PyExc_IndexError,
 		                "revlog index deletion indices are invalid");
 		return -1;
 	}
 
 	if (start < self->length) {
 		if (self->ntinitialized) {
 			Py_ssize_t i;
 
 			for (i = start; i < self->length; i++) {
 				const char *node = index_node_existing(self, i);
 				if (node == NULL)
 					return -1;
 
 				nt_delete_node(&self->nt, node);
 			}
 			if (self->added)
 				index_invalidate_added(self, 0);
 			if (self->ntrev > start)
 				self->ntrev = (int)start;
 		} else if (self->added) {
 			Py_CLEAR(self->added);
 		}
 
 		self->length = start;
 		if (start < self->raw_length) {
 			if (self->cache) {
 				Py_ssize_t i;
 				for (i = start; i < self->raw_length; i++)
 					Py_CLEAR(self->cache[i]);
 			}
 			self->raw_length = start;
 		}
 		goto done;
 	}
 
 	if (self->ntinitialized) {
 		index_invalidate_added(self, start - self->length);
 		if (self->ntrev > start)
 			self->ntrev = (int)start;
 	}
 	if (self->added)
 		ret = PyList_SetSlice(self->added, start - self->length,
 		                      PyList_GET_SIZE(self->added), NULL);
 done:
 	Py_CLEAR(self->headrevs);
 	return ret;
 }
 
 /*
  * Supported ops:
  *
  * slice deletion
  * string assignment (extend node->rev mapping)
  * string deletion (shrink node->rev mapping)
  */
 static int index_assign_subscript(indexObject *self, PyObject *item,
                                   PyObject *value)
 {
 	char *node;
 	long rev;
 
 	if (PySlice_Check(item) && value == NULL)
 		return index_slice_del(self, item);
 
 	if (node_check(item, &node) == -1)
 		return -1;
 
 	if (value == NULL)
 		return self->ntinitialized ? nt_delete_node(&self->nt, node)
 		                           : 0;
 	rev = PyInt_AsLong(value);
 	if (rev > INT_MAX || rev < 0) {
 		if (!PyErr_Occurred())
 			PyErr_SetString(PyExc_ValueError, "rev out of range");
 		return -1;
 	}
 
 	if (index_init_nt(self) == -1)
 		return -1;
 	return nt_insert(&self->nt, node, (int)rev);
 }
 
 /*
  * Find all RevlogNG entries in an index that has inline data. Update
  * the optional "offsets" table with those entries.
  */
 static Py_ssize_t inline_scan(indexObject *self, const char **offsets)
 {
 	const char *data = (const char *)self->buf.buf;
 	Py_ssize_t pos = 0;
 	Py_ssize_t end = self->buf.len;
 	long incr = v1_hdrsize;
 	Py_ssize_t len = 0;
 
 	while (pos + v1_hdrsize <= end && pos >= 0) {
 		uint32_t comp_len;
 		/* 3rd element of header is length of compressed inline data */
 		comp_len = getbe32(data + pos + 8);
 		incr = v1_hdrsize + comp_len;
 		if (offsets)
 			offsets[len] = data + pos;
 		len++;
 		pos += incr;
 	}
 
 	if (pos != end) {
 		if (!PyErr_Occurred())
 			PyErr_SetString(PyExc_ValueError, "corrupt index file");
 		return -1;
 	}
 
 	return len;
 }
 
 static int index_init(indexObject *self, PyObject *args)
 {
 	PyObject *data_obj, *inlined_obj;
 	Py_ssize_t size;
 
 	/* Initialize before argument-checking to avoid index_dealloc() crash.
 	 */
 	self->raw_length = 0;
 	self->added = NULL;
 	self->cache = NULL;
 	self->data = NULL;
 	memset(&self->buf, 0, sizeof(self->buf));
 	self->headrevs = NULL;
 	self->filteredrevs = Py_None;
 	Py_INCREF(Py_None);
 	self->ntinitialized = 0;
 	self->offsets = NULL;
 
 	if (!PyArg_ParseTuple(args, "OO", &data_obj, &inlined_obj))
 		return -1;
 	if (!PyObject_CheckBuffer(data_obj)) {
 		PyErr_SetString(PyExc_TypeError,
 		                "data does not support buffer interface");
 		return -1;
 	}
 
 	if (PyObject_GetBuffer(data_obj, &self->buf, PyBUF_SIMPLE) == -1)
 		return -1;
 	size = self->buf.len;
 
 	self->inlined = inlined_obj && PyObject_IsTrue(inlined_obj);
 	self->data = data_obj;
 
 	self->ntlookups = self->ntmisses = 0;
 	self->ntrev = -1;
 	Py_INCREF(self->data);
 
 	if (self->inlined) {
 		Py_ssize_t len = inline_scan(self, NULL);
 		if (len == -1)
 			goto bail;
 		self->raw_length = len;
 		self->length = len;
 	} else {
 		if (size % v1_hdrsize) {
 			PyErr_SetString(PyExc_ValueError, "corrupt index file");
 			goto bail;
 		}
 		self->raw_length = size / v1_hdrsize;
 		self->length = self->raw_length;
 	}
 
 	return 0;
 bail:
 	return -1;
 }
 
 static PyObject *index_nodemap(indexObject *self)
 {
 	Py_INCREF(self);
 	return (PyObject *)self;
 }
 
 static void _index_clearcaches(indexObject *self)
 {
 	if (self->cache) {
 		Py_ssize_t i;
 
 		for (i = 0; i < self->raw_length; i++)
 			Py_CLEAR(self->cache[i]);
 		free(self->cache);
 		self->cache = NULL;
 	}
 	if (self->offsets) {
 		PyMem_Free((void *)self->offsets);
 		self->offsets = NULL;
 	}
 	if (self->ntinitialized) {
 		nt_dealloc(&self->nt);
 	}
 	self->ntinitialized = 0;
 	Py_CLEAR(self->headrevs);
 }
 
 static PyObject *index_clearcaches(indexObject *self)
 {
 	_index_clearcaches(self);
 	self->ntrev = -1;
 	self->ntlookups = self->ntmisses = 0;
 	Py_RETURN_NONE;
 }
 
 static void index_dealloc(indexObject *self)
 {
 	_index_clearcaches(self);
 	Py_XDECREF(self->filteredrevs);
 	if (self->buf.buf) {
 		PyBuffer_Release(&self->buf);
 		memset(&self->buf, 0, sizeof(self->buf));
 	}
 	Py_XDECREF(self->data);
 	Py_XDECREF(self->added);
 	PyObject_Del(self);
 }
 
 static PySequenceMethods index_sequence_methods = {
     (lenfunc)index_length,      /* sq_length */
     0,                          /* sq_concat */
     0,                          /* sq_repeat */
     (ssizeargfunc)index_get,    /* sq_item */
     0,                          /* sq_slice */
     0,                          /* sq_ass_item */
     0,                          /* sq_ass_slice */
     (objobjproc)index_contains, /* sq_contains */
 };
 
 static PyMappingMethods index_mapping_methods = {
     (lenfunc)index_length,                 /* mp_length */
     (binaryfunc)index_getitem,             /* mp_subscript */
     (objobjargproc)index_assign_subscript, /* mp_ass_subscript */
 };
 
 static PyMethodDef index_methods[] = {
     {"ancestors", (PyCFunction)index_ancestors, METH_VARARGS,
      "return the gca set of the given revs"},
     {"commonancestorsheads", (PyCFunction)index_commonancestorsheads,
      METH_VARARGS,
      "return the heads of the common ancestors of the given revs"},
     {"clearcaches", (PyCFunction)index_clearcaches, METH_NOARGS,
      "clear the index caches"},
     {"get", (PyCFunction)index_m_get, METH_VARARGS, "get an index entry"},
     {"get_rev", (PyCFunction)index_m_get, METH_VARARGS,
      "return `rev` associated with a node or None"},
     {"has_node", (PyCFunction)index_m_has_node, METH_O,
      "return True if the node exist in the index"},
     {"rev", (PyCFunction)index_m_rev, METH_O,
      "return `rev` associated with a node or raise RevlogError"},
     {"computephasesmapsets", (PyCFunction)compute_phases_map_sets, METH_VARARGS,
      "compute phases"},
     {"reachableroots2", (PyCFunction)reachableroots2, METH_VARARGS,
      "reachableroots"},
     {"headrevs", (PyCFunction)index_headrevs, METH_VARARGS,
      "get head revisions"}, /* Can do filtering since 3.2 */
     {"headrevsfiltered", (PyCFunction)index_headrevs, METH_VARARGS,
      "get filtered head revisions"}, /* Can always do filtering */
     {"issnapshot", (PyCFunction)index_issnapshot, METH_O,
      "True if the object is a snapshot"},
     {"findsnapshots", (PyCFunction)index_findsnapshots, METH_VARARGS,
      "Gather snapshot data in a cache dict"},
     {"deltachain", (PyCFunction)index_deltachain, METH_VARARGS,
      "determine revisions with deltas to reconstruct fulltext"},
     {"slicechunktodensity", (PyCFunction)index_slicechunktodensity,
      METH_VARARGS, "determine revisions with deltas to reconstruct fulltext"},
     {"append", (PyCFunction)index_append, METH_O, "append an index entry"},
     {"partialmatch", (PyCFunction)index_partialmatch, METH_VARARGS,
      "match a potentially ambiguous node ID"},
     {"shortest", (PyCFunction)index_shortest, METH_VARARGS,
      "find length of shortest hex nodeid of a binary ID"},
     {"stats", (PyCFunction)index_stats, METH_NOARGS, "stats for the index"},
     {NULL} /* Sentinel */
 };
 
 static PyGetSetDef index_getset[] = {
     {"nodemap", (getter)index_nodemap, NULL, "nodemap", NULL},
     {NULL} /* Sentinel */
 };
 
 PyTypeObject HgRevlogIndex_Type = {
     PyVarObject_HEAD_INIT(NULL, 0) /* header */
     "parsers.index",               /* tp_name */
     sizeof(indexObject),           /* tp_basicsize */
     0,                             /* tp_itemsize */
     (destructor)index_dealloc,     /* tp_dealloc */
     0,                             /* tp_print */
     0,                             /* tp_getattr */
     0,                             /* tp_setattr */
     0,                             /* tp_compare */
     0,                             /* tp_repr */
     0,                             /* tp_as_number */
     &index_sequence_methods,       /* tp_as_sequence */
     &index_mapping_methods,        /* 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 */
     "revlog index",                /* tp_doc */
     0,                             /* tp_traverse */
     0,                             /* tp_clear */
     0,                             /* tp_richcompare */
     0,                             /* tp_weaklistoffset */
     0,                             /* tp_iter */
     0,                             /* tp_iternext */
     index_methods,                 /* tp_methods */
     0,                             /* tp_members */
     index_getset,                  /* tp_getset */
     0,                             /* tp_base */
     0,                             /* tp_dict */
     0,                             /* tp_descr_get */
     0,                             /* tp_descr_set */
     0,                             /* tp_dictoffset */
     (initproc)index_init,          /* tp_init */
     0,                             /* tp_alloc */
 };
 
 /*
  * returns a tuple of the form (index, index, cache) with elements as
  * follows:
  *
  * index: an index object that lazily parses RevlogNG records
  * cache: if data is inlined, a tuple (0, index_file_content), else None
  *        index_file_content could be a string, or a buffer
  *
  * added complications are for backwards compatibility
  */
 PyObject *parse_index2(PyObject *self, PyObject *args)
 {
 	PyObject *tuple = NULL, *cache = NULL;
 	indexObject *idx;
 	int ret;
 
 	idx = PyObject_New(indexObject, &HgRevlogIndex_Type);
 	if (idx == NULL)
 		goto bail;
 
 	ret = index_init(idx, args);
 	if (ret == -1)
 		goto bail;
 
 	if (idx->inlined) {
 		cache = Py_BuildValue("iO", 0, idx->data);
 		if (cache == NULL)
 			goto bail;
 	} else {
 		cache = Py_None;
 		Py_INCREF(cache);
 	}
 
 	tuple = Py_BuildValue("NN", idx, cache);
 	if (!tuple)
 		goto bail;
 	return tuple;
 
 bail:
 	Py_XDECREF(idx);
 	Py_XDECREF(cache);
 	Py_XDECREF(tuple);
 	return NULL;
 }
 
 static Revlog_CAPI CAPI = {
     /* increment the abi_version field upon each change in the Revlog_CAPI
        struct or in the ABI of the listed functions */
     2,
     index_length,
     index_node,
     HgRevlogIndex_GetParents,
 };
 
 void revlog_module_init(PyObject *mod)
 {
 	PyObject *caps = NULL;
 	HgRevlogIndex_Type.tp_new = PyType_GenericNew;
 	if (PyType_Ready(&HgRevlogIndex_Type) < 0)
 		return;
 	Py_INCREF(&HgRevlogIndex_Type);
 	PyModule_AddObject(mod, "index", (PyObject *)&HgRevlogIndex_Type);
 
 	nodetreeType.tp_new = PyType_GenericNew;
 	if (PyType_Ready(&nodetreeType) < 0)
 		return;
 	Py_INCREF(&nodetreeType);
 	PyModule_AddObject(mod, "nodetree", (PyObject *)&nodetreeType);
 
 	if (!nullentry) {
 		nullentry =
 		    Py_BuildValue(PY23("iiiiiiis#", "iiiiiiiy#"), 0, 0, 0, -1,
 		                  -1, -1, -1, nullid, (Py_ssize_t)20);
 	}
 	if (nullentry)
 		PyObject_GC_UnTrack(nullentry);
 
 	caps = PyCapsule_New(&CAPI, "mercurial.cext.parsers.revlog_CAPI", NULL);
 	if (caps != NULL)
 		PyModule_AddObject(mod, "revlog_CAPI", caps);
 }

mercurial/phases.py

@@ -1,924 +1,929
 """ Mercurial phases support code
 
     ---
 
     Copyright 2011 Pierre-Yves David <[e-mail]>
                    Logilab SA        <[e-mail]>
                    Augie Fackler     <[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.
 
     ---
 
 This module implements most phase logic in mercurial.
 
 
 Basic Concept
 =============
 
 A 'changeset phase' is an indicator that tells us how a changeset is
 manipulated and communicated. The details of each phase is described
 below, here we describe the properties they have in common.
 
 Like bookmarks, phases are not stored in history and thus are not
 permanent and leave no audit trail.
 
 First, no changeset can be in two phases at once. Phases are ordered,
 so they can be considered from lowest to highest. The default, lowest
 phase is 'public' - this is the normal phase of existing changesets. A
 child changeset can not be in a lower phase than its parents.
 
 These phases share a hierarchy of traits:
 
             immutable shared
     public:     X        X
     draft:               X
     secret:
 
 Local commits are draft by default.
 
 Phase Movement and Exchange
 ===========================
 
 Phase data is exchanged by pushkey on pull and push. Some servers have
 a publish option set, we call such a server a "publishing server".
 Pushing a draft changeset to a publishing server changes the phase to
 public.
 
 A small list of fact/rules define the exchange of phase:
 
 * old client never changes server states
 * pull never changes server states
 * publish and old server changesets are seen as public by client
 * any secret changeset seen in another repository is lowered to at
   least draft
 
 Here is the final table summing up the 49 possible use cases of phase
 exchange:
 
                            server
                   old     publish      non-publish
                  N   X    N   D   P    N   D   P
     old client
     pull
      N           -   X/X  -   X/D X/P  -   X/D X/P
      X           -   X/X  -   X/D X/P  -   X/D X/P
     push
      X           X/X X/X  X/P X/P X/P  X/D X/D X/P
     new client
     pull
      N           -   P/X  -   P/D P/P  -   D/D P/P
      D           -   P/X  -   P/D P/P  -   D/D P/P
      P           -   P/X  -   P/D P/P  -   P/D P/P
     push
      D           P/X P/X  P/P P/P P/P  D/D D/D P/P
      P           P/X P/X  P/P P/P P/P  P/P P/P P/P
 
 Legend:
 
     A/B = final state on client / state on server
 
     * N = new/not present,
     * P = public,
     * D = draft,
     * X = not tracked (i.e., the old client or server has no internal
           way of recording the phase.)
 
     passive = only pushes
 
 
     A cell here can be read like this:
 
     "When a new client pushes a draft changeset (D) to a publishing
     server where it's not present (N), it's marked public on both
     sides (P/P)."
 
 Note: old client behave as a publishing server with draft only content
 - other people see it as public
 - content is pushed as draft
 
 """
 
 from __future__ import absolute_import
 
 import errno
 import struct
 
 from .i18n import _
 from .node import (
     bin,
     hex,
     nullid,
     nullrev,
     short,
     wdirrev,
 )
 from .pycompat import (
     getattr,
     setattr,
 )
 from . import (
     error,
     pycompat,
     smartset,
     txnutil,
     util,
 )
 
 _fphasesentry = struct.Struct(b'>i20s')
 
 # record phase index
 public, draft, secret = range(3)
 archived = 32  # non-continuous for compatibility
 internal = 96  # non-continuous for compatibility
 allphases = (public, draft, secret, archived, internal)
 trackedphases = (draft, secret, archived, internal)
 # record phase names
 cmdphasenames = [b'public', b'draft', b'secret']  # known to `hg phase` command
 phasenames = dict(enumerate(cmdphasenames))
 phasenames[archived] = b'archived'
 phasenames[internal] = b'internal'
 # map phase name to phase number
 phasenumber = {name: phase for phase, name in phasenames.items()}
 # like phasenumber, but also include maps for the numeric and binary
 # phase number to the phase number
 phasenumber2 = phasenumber.copy()
 phasenumber2.update({phase: phase for phase in phasenames})
 phasenumber2.update({b'%i' % phase: phase for phase in phasenames})
 # record phase property
 mutablephases = (draft, secret, archived, internal)
 remotehiddenphases = (secret, archived, internal)
 localhiddenphases = (internal, archived)
 
 
 def supportinternal(repo):
     """True if the internal phase can be used on a repository"""
     return b'internal-phase' in repo.requirements
 
 
 def _readroots(repo, phasedefaults=None):
     """Read phase roots from disk
 
     phasedefaults is a list of fn(repo, roots) callable, which are
     executed if the phase roots file does not exist. When phases are
     being initialized on an existing repository, this could be used to
     set selected changesets phase to something else than public.
 
     Return (roots, dirty) where dirty is true if roots differ from
     what is being stored.
     """
     repo = repo.unfiltered()
     dirty = False
-    roots = [set() for i in range(max(allphases) + 1)]
+    roots = {i: set() for i in allphases}
     try:
         f, pending = txnutil.trypending(repo.root, repo.svfs, b'phaseroots')
         try:
             for line in f:
                 phase, nh = line.split()
                 roots[int(phase)].add(bin(nh))
         finally:
             f.close()
     except IOError as inst:
         if inst.errno != errno.ENOENT:
             raise
         if phasedefaults:
             for f in phasedefaults:
                 roots = f(repo, roots)
         dirty = True
     return roots, dirty
 
 
 def binaryencode(phasemapping):
     """encode a 'phase -> nodes' mapping into a binary stream
 
     The revision lists are encoded as (phase, root) pairs.
     """
     binarydata = []
     for phase, nodes in pycompat.iteritems(phasemapping):
         for head in nodes:
             binarydata.append(_fphasesentry.pack(phase, head))
     return b''.join(binarydata)
 
 
 def binarydecode(stream):
     """decode a binary stream into a 'phase -> nodes' mapping
 
     The (phase, root) pairs are turned back into a dictionary with
     the phase as index and the aggregated roots of that phase as value."""
     headsbyphase = {i: [] for i in allphases}
     entrysize = _fphasesentry.size
     while True:
         entry = stream.read(entrysize)
         if len(entry) < entrysize:
             if entry:
                 raise error.Abort(_(b'bad phase-heads stream'))
             break
         phase, node = _fphasesentry.unpack(entry)
         headsbyphase[phase].append(node)
     return headsbyphase
 
 
 def _sortedrange_insert(data, idx, rev, t):
     merge_before = False
     if idx:
         r1, t1 = data[idx - 1]
         merge_before = r1[-1] + 1 == rev and t1 == t
     merge_after = False
     if idx < len(data):
         r2, t2 = data[idx]
         merge_after = r2[0] == rev + 1 and t2 == t
 
     if merge_before and merge_after:
         data[idx - 1] = (pycompat.xrange(r1[0], r2[-1] + 1), t)
         data.pop(idx)
     elif merge_before:
         data[idx - 1] = (pycompat.xrange(r1[0], rev + 1), t)
     elif merge_after:
         data[idx] = (pycompat.xrange(rev, r2[-1] + 1), t)
     else:
         data.insert(idx, (pycompat.xrange(rev, rev + 1), t))
 
 
 def _sortedrange_split(data, idx, rev, t):
     r1, t1 = data[idx]
     if t == t1:
         return
     t = (t1[0], t[1])
     if len(r1) == 1:
         data.pop(idx)
         _sortedrange_insert(data, idx, rev, t)
     elif r1[0] == rev:
         data[idx] = (pycompat.xrange(rev + 1, r1[-1] + 1), t1)
         _sortedrange_insert(data, idx, rev, t)
     elif r1[-1] == rev:
         data[idx] = (pycompat.xrange(r1[0], rev), t1)
         _sortedrange_insert(data, idx + 1, rev, t)
     else:
         data[idx : idx + 1] = [
             (pycompat.xrange(r1[0], rev), t1),
             (pycompat.xrange(rev, rev + 1), t),
             (pycompat.xrange(rev + 1, r1[-1] + 1), t1),
         ]
 
 
 def _trackphasechange(data, rev, old, new):
     """add a phase move to the <data> list of ranges
 
     If data is None, nothing happens.
     """
     if data is None:
         return
 
     # If data is empty, create a one-revision range and done
     if not data:
         data.insert(0, (pycompat.xrange(rev, rev + 1), (old, new)))
         return
 
     low = 0
     high = len(data)
     t = (old, new)
     while low < high:
         mid = (low + high) // 2
         revs = data[mid][0]
 
         if rev in revs:
             _sortedrange_split(data, mid, rev, t)
             return
 
         if revs[0] == rev + 1:
             if mid and data[mid - 1][0][-1] == rev:
                 _sortedrange_split(data, mid - 1, rev, t)
             else:
                 _sortedrange_insert(data, mid, rev, t)
             return
 
         if revs[-1] == rev - 1:
             if mid + 1 < len(data) and data[mid + 1][0][0] == rev:
                 _sortedrange_split(data, mid + 1, rev, t)
             else:
                 _sortedrange_insert(data, mid + 1, rev, t)
             return
 
         if revs[0] > rev:
             high = mid
         else:
             low = mid + 1
 
     if low == len(data):
         data.append((pycompat.xrange(rev, rev + 1), t))
         return
 
     r1, t1 = data[low]
     if r1[0] > rev:
         data.insert(low, (pycompat.xrange(rev, rev + 1), t))
     else:
         data.insert(low + 1, (pycompat.xrange(rev, rev + 1), t))
 
 
 class phasecache(object):
     def __init__(self, repo, phasedefaults, _load=True):
         if _load:
             # Cheap trick to allow shallow-copy without copy module
             self.phaseroots, self.dirty = _readroots(repo, phasedefaults)
             self._loadedrevslen = 0
             self._phasesets = None
             self.filterunknown(repo)
             self.opener = repo.svfs
 
     def hasnonpublicphases(self, repo):
         """detect if there are revisions with non-public phase"""
         repo = repo.unfiltered()
         cl = repo.changelog
         if len(cl) >= self._loadedrevslen:
             self.invalidate()
             self.loadphaserevs(repo)
-        return any(self.phaseroots[1:])
+        return any(
+            revs
+            for phase, revs in pycompat.iteritems(self.phaseroots)
+            if phase != public
+        )
 
     def nonpublicphaseroots(self, repo):
         """returns the roots of all non-public phases
 
         The roots are not minimized, so if the secret revisions are
         descendants of draft revisions, their roots will still be present.
         """
         repo = repo.unfiltered()
         cl = repo.changelog
         if len(cl) >= self._loadedrevslen:
             self.invalidate()
             self.loadphaserevs(repo)
-        return set().union(*[roots for roots in self.phaseroots[1:] if roots])
+        return set().union(
+            *[
+                revs
+                for phase, revs in pycompat.iteritems(self.phaseroots)
+                if phase != public
+            ]
+        )
 
     def getrevset(self, repo, phases, subset=None):
         """return a smartset for the given phases"""
         self.loadphaserevs(repo)  # ensure phase's sets are loaded
         phases = set(phases)
         publicphase = public in phases
 
         if publicphase:
             # In this case, phases keeps all the *other* phases.
             phases = set(allphases).difference(phases)
             if not phases:
                 return smartset.fullreposet(repo)
 
         # fast path: _phasesets contains the interesting sets,
         # might only need a union and post-filtering.
         revsneedscopy = False
         if len(phases) == 1:
             [p] = phases
             revs = self._phasesets[p]
             revsneedscopy = True  # Don't modify _phasesets
         else:
             # revs has the revisions in all *other* phases.
             revs = set.union(*[self._phasesets[p] for p in phases])
 
         def _addwdir(wdirsubset, wdirrevs):
             if wdirrev in wdirsubset and repo[None].phase() in phases:
                 if revsneedscopy:
                     wdirrevs = wdirrevs.copy()
                 # The working dir would never be in the # cache, but it was in
                 # the subset being filtered for its phase (or filtered out,
                 # depending on publicphase), so add it to the output to be
                 # included (or filtered out).
                 wdirrevs.add(wdirrev)
             return wdirrevs
 
         if not publicphase:
             if repo.changelog.filteredrevs:
                 revs = revs - repo.changelog.filteredrevs
 
             if subset is None:
                 return smartset.baseset(revs)
             else:
                 revs = _addwdir(subset, revs)
                 return subset & smartset.baseset(revs)
         else:
             if subset is None:
                 subset = smartset.fullreposet(repo)
 
             revs = _addwdir(subset, revs)
 
             if not revs:
                 return subset
             return subset.filter(lambda r: r not in revs)
 
     def copy(self):
         # Shallow copy meant to ensure isolation in
         # advance/retractboundary(), nothing more.
         ph = self.__class__(None, None, _load=False)
-        ph.phaseroots = self.phaseroots[:]
+        ph.phaseroots = self.phaseroots.copy()
         ph.dirty = self.dirty
         ph.opener = self.opener
         ph._loadedrevslen = self._loadedrevslen
         ph._phasesets = self._phasesets
         return ph
 
     def replace(self, phcache):
         """replace all values in 'self' with content of phcache"""
         for a in (
             b'phaseroots',
             b'dirty',
             b'opener',
             b'_loadedrevslen',
             b'_phasesets',
         ):
             setattr(self, a, getattr(phcache, a))
 
     def _getphaserevsnative(self, repo):
         repo = repo.unfiltered()
-        nativeroots = []
-        for phase in trackedphases:
-            nativeroots.append(
-                pycompat.maplist(repo.changelog.rev, self.phaseroots[phase])
-            )
-        revslen, phasesets = repo.changelog.computephases(nativeroots)
-        phasesets2 = [set() for phase in range(max(allphases) + 1)]
-        for phase, phaseset in zip(allphases, phasesets):
-            phasesets2[phase] = phaseset
-        return revslen, phasesets2
+        return repo.changelog.computephases(self.phaseroots)
 
     def _computephaserevspure(self, repo):
         repo = repo.unfiltered()
         cl = repo.changelog
-        self._phasesets = [set() for phase in range(max(allphases) + 1)]
+        self._phasesets = {phase: set() for phase in allphases}
         lowerroots = set()
         for phase in reversed(trackedphases):
             roots = pycompat.maplist(cl.rev, self.phaseroots[phase])
             if roots:
                 ps = set(cl.descendants(roots))
                 for root in roots:
                     ps.add(root)
                 ps.difference_update(lowerroots)
                 lowerroots.update(ps)
                 self._phasesets[phase] = ps
         self._loadedrevslen = len(cl)
 
     def loadphaserevs(self, repo):
         """ensure phase information is loaded in the object"""
         if self._phasesets is None:
             try:
                 res = self._getphaserevsnative(repo)
                 self._loadedrevslen, self._phasesets = res
             except AttributeError:
                 self._computephaserevspure(repo)
 
     def invalidate(self):
         self._loadedrevslen = 0
         self._phasesets = None
 
     def phase(self, repo, rev):
         # We need a repo argument here to be able to build _phasesets
         # if necessary. The repository instance is not stored in
         # phasecache to avoid reference cycles. The changelog instance
         # is not stored because it is a filecache() property and can
         # be replaced without us being notified.
         if rev == nullrev:
             return public
         if rev < nullrev:
             raise ValueError(_(b'cannot lookup negative revision'))
         if rev >= self._loadedrevslen:
             self.invalidate()
             self.loadphaserevs(repo)
         for phase in trackedphases:
             if rev in self._phasesets[phase]:
                 return phase
         return public
 
     def write(self):
         if not self.dirty:
             return
         f = self.opener(b'phaseroots', b'w', atomictemp=True, checkambig=True)
         try:
             self._write(f)
         finally:
             f.close()
 
     def _write(self, fp):
-        for phase, roots in enumerate(self.phaseroots):
+        for phase, roots in pycompat.iteritems(self.phaseroots):
             for h in sorted(roots):
                 fp.write(b'%i %s\n' % (phase, hex(h)))
         self.dirty = False
 
     def _updateroots(self, phase, newroots, tr):
         self.phaseroots[phase] = newroots
         self.invalidate()
         self.dirty = True
 
         tr.addfilegenerator(b'phase', (b'phaseroots',), self._write)
         tr.hookargs[b'phases_moved'] = b'1'
 
     def registernew(self, repo, tr, targetphase, nodes):
         repo = repo.unfiltered()
         self._retractboundary(repo, tr, targetphase, nodes)
         if tr is not None and b'phases' in tr.changes:
             phasetracking = tr.changes[b'phases']
             torev = repo.changelog.rev
             phase = self.phase
             revs = [torev(node) for node in nodes]
             revs.sort()
             for rev in revs:
                 revphase = phase(repo, rev)
                 _trackphasechange(phasetracking, rev, None, revphase)
         repo.invalidatevolatilesets()
 
     def advanceboundary(self, repo, tr, targetphase, nodes, dryrun=None):
         """Set all 'nodes' to phase 'targetphase'
 
         Nodes with a phase lower than 'targetphase' are not affected.
 
         If dryrun is True, no actions will be performed
 
         Returns a set of revs whose phase is changed or should be changed
         """
         # Be careful to preserve shallow-copied values: do not update
         # phaseroots values, replace them.
         if tr is None:
             phasetracking = None
         else:
             phasetracking = tr.changes.get(b'phases')
 
         repo = repo.unfiltered()
 
         changes = set()  # set of revisions to be changed
         delroots = []  # set of root deleted by this path
         for phase in (phase for phase in allphases if phase > targetphase):
             # filter nodes that are not in a compatible phase already
             nodes = [
                 n for n in nodes if self.phase(repo, repo[n].rev()) >= phase
             ]
             if not nodes:
                 break  # no roots to move anymore
 
             olds = self.phaseroots[phase]
 
             affected = repo.revs(b'%ln::%ln', olds, nodes)
             changes.update(affected)
             if dryrun:
                 continue
             for r in affected:
                 _trackphasechange(
                     phasetracking, r, self.phase(repo, r), targetphase
                 )
 
             roots = {
                 ctx.node()
                 for ctx in repo.set(b'roots((%ln::) - %ld)', olds, affected)
             }
             if olds != roots:
                 self._updateroots(phase, roots, tr)
                 # some roots may need to be declared for lower phases
                 delroots.extend(olds - roots)
         if not dryrun:
             # declare deleted root in the target phase
             if targetphase != 0:
                 self._retractboundary(repo, tr, targetphase, delroots)
             repo.invalidatevolatilesets()
         return changes
 
     def retractboundary(self, repo, tr, targetphase, nodes):
-        oldroots = self.phaseroots[: targetphase + 1]
+        oldroots = {
+            phase: revs
+            for phase, revs in pycompat.iteritems(self.phaseroots)
+            if phase <= targetphase
+        }
         if tr is None:
             phasetracking = None
         else:
             phasetracking = tr.changes.get(b'phases')
         repo = repo.unfiltered()
         if (
             self._retractboundary(repo, tr, targetphase, nodes)
             and phasetracking is not None
         ):
 
             # find the affected revisions
             new = self.phaseroots[targetphase]
             old = oldroots[targetphase]
             affected = set(repo.revs(b'(%ln::) - (%ln::)', new, old))
 
             # find the phase of the affected revision
             for phase in pycompat.xrange(targetphase, -1, -1):
                 if phase:
-                    roots = oldroots[phase]
+                    roots = oldroots.get(phase, [])
                     revs = set(repo.revs(b'%ln::%ld', roots, affected))
                     affected -= revs
                 else:  # public phase
                     revs = affected
                 for r in sorted(revs):
                     _trackphasechange(phasetracking, r, phase, targetphase)
         repo.invalidatevolatilesets()
 
     def _retractboundary(self, repo, tr, targetphase, nodes):
         # Be careful to preserve shallow-copied values: do not update
         # phaseroots values, replace them.
         if targetphase in (archived, internal) and not supportinternal(repo):
             name = phasenames[targetphase]
             msg = b'this repository does not support the %s phase' % name
             raise error.ProgrammingError(msg)
 
         repo = repo.unfiltered()
         torev = repo.changelog.rev
         tonode = repo.changelog.node
         currentroots = {torev(node) for node in self.phaseroots[targetphase]}
         finalroots = oldroots = set(currentroots)
         newroots = [torev(node) for node in nodes]
         newroots = [
             rev for rev in newroots if self.phase(repo, rev) < targetphase
         ]
 
         if newroots:
             if nullrev in newroots:
                 raise error.Abort(_(b'cannot change null revision phase'))
             currentroots.update(newroots)
 
             # Only compute new roots for revs above the roots that are being
             # retracted.
             minnewroot = min(newroots)
             aboveroots = [rev for rev in currentroots if rev >= minnewroot]
             updatedroots = repo.revs(b'roots(%ld::)', aboveroots)
 
             finalroots = {rev for rev in currentroots if rev < minnewroot}
             finalroots.update(updatedroots)
         if finalroots != oldroots:
             self._updateroots(
                 targetphase, {tonode(rev) for rev in finalroots}, tr
             )
             return True
         return False
 
     def filterunknown(self, repo):
         """remove unknown nodes from the phase boundary
 
         Nothing is lost as unknown nodes only hold data for their descendants.
         """
         filtered = False
         has_node = repo.changelog.index.has_node  # to filter unknown nodes
-        for phase, nodes in enumerate(self.phaseroots):
+        for phase, nodes in pycompat.iteritems(self.phaseroots):
             missing = sorted(node for node in nodes if not has_node(node))
             if missing:
                 for mnode in missing:
                     repo.ui.debug(
                         b'removing unknown node %s from %i-phase boundary\n'
                         % (short(mnode), phase)
                     )
                 nodes.symmetric_difference_update(missing)
                 filtered = True
         if filtered:
             self.dirty = True
         # filterunknown is called by repo.destroyed, we may have no changes in
         # root but _phasesets contents is certainly invalid (or at least we
         # have not proper way to check that). related to issue 3858.
         #
         # The other caller is __init__ that have no _phasesets initialized
         # anyway. If this change we should consider adding a dedicated
         # "destroyed" function to phasecache or a proper cache key mechanism
         # (see branchmap one)
         self.invalidate()
 
 
 def advanceboundary(repo, tr, targetphase, nodes, dryrun=None):
     """Add nodes to a phase changing other nodes phases if necessary.
 
     This function move boundary *forward* this means that all nodes
     are set in the target phase or kept in a *lower* phase.
 
     Simplify boundary to contains phase roots only.
 
     If dryrun is True, no actions will be performed
 
     Returns a set of revs whose phase is changed or should be changed
     """
     phcache = repo._phasecache.copy()
     changes = phcache.advanceboundary(
         repo, tr, targetphase, nodes, dryrun=dryrun
     )
     if not dryrun:
         repo._phasecache.replace(phcache)
     return changes
 
 
 def retractboundary(repo, tr, targetphase, nodes):
     """Set nodes back to a phase changing other nodes phases if
     necessary.
 
     This function move boundary *backward* this means that all nodes
     are set in the target phase or kept in a *higher* phase.
 
     Simplify boundary to contains phase roots only."""
     phcache = repo._phasecache.copy()
     phcache.retractboundary(repo, tr, targetphase, nodes)
     repo._phasecache.replace(phcache)
 
 
 def registernew(repo, tr, targetphase, nodes):
     """register a new revision and its phase
 
     Code adding revisions to the repository should use this function to
     set new changeset in their target phase (or higher).
     """
     phcache = repo._phasecache.copy()
     phcache.registernew(repo, tr, targetphase, nodes)
     repo._phasecache.replace(phcache)
 
 
 def listphases(repo):
     """List phases root for serialization over pushkey"""
     # Use ordered dictionary so behavior is deterministic.
     keys = util.sortdict()
     value = b'%i' % draft
     cl = repo.unfiltered().changelog
     for root in repo._phasecache.phaseroots[draft]:
         if repo._phasecache.phase(repo, cl.rev(root)) <= draft:
             keys[hex(root)] = value
 
     if repo.publishing():
         # Add an extra data to let remote know we are a publishing
         # repo. Publishing repo can't just pretend they are old repo.
         # When pushing to a publishing repo, the client still need to
         # push phase boundary
         #
         # Push do not only push changeset. It also push phase data.
         # New phase data may apply to common changeset which won't be
         # push (as they are common). Here is a very simple example:
         #
         # 1) repo A push changeset X as draft to repo B
         # 2) repo B make changeset X public
         # 3) repo B push to repo A. X is not pushed but the data that
         #    X as now public should
         #
         # The server can't handle it on it's own as it has no idea of
         # client phase data.
         keys[b'publishing'] = b'True'
     return keys
 
 
 def pushphase(repo, nhex, oldphasestr, newphasestr):
     """List phases root for serialization over pushkey"""
     repo = repo.unfiltered()
     with repo.lock():
         currentphase = repo[nhex].phase()
         newphase = abs(int(newphasestr))  # let's avoid negative index surprise
         oldphase = abs(int(oldphasestr))  # let's avoid negative index surprise
         if currentphase == oldphase and newphase < oldphase:
             with repo.transaction(b'pushkey-phase') as tr:
                 advanceboundary(repo, tr, newphase, [bin(nhex)])
             return True
         elif currentphase == newphase:
             # raced, but got correct result
             return True
         else:
             return False
 
 
 def subsetphaseheads(repo, subset):
     """Finds the phase heads for a subset of a history
 
     Returns a list indexed by phase number where each item is a list of phase
     head nodes.
     """
     cl = repo.changelog
 
     headsbyphase = {i: [] for i in allphases}
     # No need to keep track of secret phase; any heads in the subset that
     # are not mentioned are implicitly secret.
     for phase in allphases[:secret]:
         revset = b"heads(%%ln & %s())" % phasenames[phase]
         headsbyphase[phase] = [cl.node(r) for r in repo.revs(revset, subset)]
     return headsbyphase
 
 
 def updatephases(repo, trgetter, headsbyphase):
     """Updates the repo with the given phase heads"""
     # Now advance phase boundaries of all phases
     #
     # run the update (and fetch transaction) only if there are actually things
     # to update. This avoid creating empty transaction during no-op operation.
 
     for phase in allphases:
         revset = b'%ln - _phase(%s)'
         heads = [c.node() for c in repo.set(revset, headsbyphase[phase], phase)]
         if heads:
             advanceboundary(repo, trgetter(), phase, heads)
 
 
 def analyzeremotephases(repo, subset, roots):
     """Compute phases heads and root in a subset of node from root dict
 
     * subset is heads of the subset
     * roots is {<nodeid> => phase} mapping. key and value are string.
 
     Accept unknown element input
     """
     repo = repo.unfiltered()
     # build list from dictionary
     draftroots = []
     has_node = repo.changelog.index.has_node  # to filter unknown nodes
     for nhex, phase in pycompat.iteritems(roots):
         if nhex == b'publishing':  # ignore data related to publish option
             continue
         node = bin(nhex)
         phase = int(phase)
         if phase == public:
             if node != nullid:
                 repo.ui.warn(
                     _(
                         b'ignoring inconsistent public root'
                         b' from remote: %s\n'
                     )
                     % nhex
                 )
         elif phase == draft:
             if has_node(node):
                 draftroots.append(node)
         else:
             repo.ui.warn(
                 _(b'ignoring unexpected root from remote: %i %s\n')
                 % (phase, nhex)
             )
     # compute heads
     publicheads = newheads(repo, subset, draftroots)
     return publicheads, draftroots
 
 
 class remotephasessummary(object):
     """summarize phase information on the remote side
 
     :publishing: True is the remote is publishing
     :publicheads: list of remote public phase heads (nodes)
     :draftheads: list of remote draft phase heads (nodes)
     :draftroots: list of remote draft phase root (nodes)
     """
 
     def __init__(self, repo, remotesubset, remoteroots):
         unfi = repo.unfiltered()
         self._allremoteroots = remoteroots
 
         self.publishing = remoteroots.get(b'publishing', False)
 
         ana = analyzeremotephases(repo, remotesubset, remoteroots)
         self.publicheads, self.draftroots = ana
         # Get the list of all "heads" revs draft on remote
         dheads = unfi.set(b'heads(%ln::%ln)', self.draftroots, remotesubset)
         self.draftheads = [c.node() for c in dheads]
 
 
 def newheads(repo, heads, roots):
     """compute new head of a subset minus another
 
     * `heads`: define the first subset
     * `roots`: define the second we subtract from the first"""
     # prevent an import cycle
     # phases > dagop > patch > copies > scmutil > obsolete > obsutil > phases
     from . import dagop
 
     repo = repo.unfiltered()
     cl = repo.changelog
     rev = cl.index.get_rev
     if not roots:
         return heads
     if not heads or heads == [nullid]:
         return []
     # The logic operated on revisions, convert arguments early for convenience
     new_heads = {rev(n) for n in heads if n != nullid}
     roots = [rev(n) for n in roots]
     # compute the area we need to remove
     affected_zone = repo.revs(b"(%ld::%ld)", roots, new_heads)
     # heads in the area are no longer heads
     new_heads.difference_update(affected_zone)
     # revisions in the area have children outside of it,
     # They might be new heads
     candidates = repo.revs(
         b"parents(%ld + (%ld and merge())) and not null", roots, affected_zone
     )
     candidates -= affected_zone
     if new_heads or candidates:
         # remove candidate that are ancestors of other heads
         new_heads.update(candidates)
         prunestart = repo.revs(b"parents(%ld) and not null", new_heads)
         pruned = dagop.reachableroots(repo, candidates, prunestart)
         new_heads.difference_update(pruned)
 
     return pycompat.maplist(cl.node, sorted(new_heads))
 
 
 def newcommitphase(ui):
     """helper to get the target phase of new commit
 
     Handle all possible values for the phases.new-commit options.
 
     """
     v = ui.config(b'phases', b'new-commit')
     try:
         return phasenumber2[v]
     except KeyError:
         raise error.ConfigError(
             _(b"phases.new-commit: not a valid phase name ('%s')") % v
         )
 
 
 def hassecret(repo):
     """utility function that check if a repo have any secret changeset."""
     return bool(repo._phasecache.phaseroots[secret])
 
 
 def preparehookargs(node, old, new):
     if old is None:
         old = b''
     else:
         old = phasenames[old]
     return {b'node': node, b'oldphase': old, b'phase': phasenames[new]}

mercurial/policy.py

@@ -1,157 +1,157
 # policy.py - module policy logic for Mercurial.
 #
 # Copyright 2015 Gregory Szorc <[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 os
 import sys
 
 from .pycompat import getattr
 
 # Rules for how modules can be loaded. Values are:
 #
 #    c - require C extensions
 #    rust+c - require Rust and C extensions
 #    rust+c-allow - allow Rust and C extensions with fallback to pure Python
 #                   for each
 #    allow - allow pure Python implementation when C loading fails
 #    cffi - required cffi versions (implemented within pure module)
 #    cffi-allow - allow pure Python implementation if cffi version is missing
 #    py - only load pure Python modules
 #
 # By default, fall back to the pure modules so the in-place build can
 # run without recompiling the C extensions. This will be overridden by
 # __modulepolicy__ generated by setup.py.
 policy = b'allow'
 _packageprefs = {
     # policy: (versioned package, pure package)
     b'c': ('cext', None),
     b'allow': ('cext', 'pure'),
     b'cffi': ('cffi', None),
     b'cffi-allow': ('cffi', 'pure'),
     b'py': (None, 'pure'),
     # For now, rust policies impact importrust only
     b'rust+c': ('cext', None),
     b'rust+c-allow': ('cext', 'pure'),
 }
 
 try:
     from . import __modulepolicy__
 
     policy = __modulepolicy__.modulepolicy
 except ImportError:
     pass
 
 # PyPy doesn't load C extensions.
 #
 # The canonical way to do this is to test platform.python_implementation().
 # But we don't import platform and don't bloat for it here.
 if '__pypy__' in sys.builtin_module_names:
     policy = b'cffi'
 
 # Environment variable can always force settings.
 if sys.version_info[0] >= 3:
     if 'HGMODULEPOLICY' in os.environ:
         policy = os.environ['HGMODULEPOLICY'].encode('utf-8')
 else:
     policy = os.environ.get('HGMODULEPOLICY', policy)
 
 
 def _importfrom(pkgname, modname):
     # from .<pkgname> import <modname> (where . is looked through this module)
     fakelocals = {}
     pkg = __import__(pkgname, globals(), fakelocals, [modname], level=1)
     try:
         fakelocals[modname] = mod = getattr(pkg, modname)
     except AttributeError:
         raise ImportError('cannot import name %s' % modname)
     # force import; fakelocals[modname] may be replaced with the real module
     getattr(mod, '__doc__', None)
     return fakelocals[modname]
 
 
 # keep in sync with "version" in C modules
 _cextversions = {
     ('cext', 'base85'): 1,
     ('cext', 'bdiff'): 3,
     ('cext', 'mpatch'): 1,
     ('cext', 'osutil'): 4,
-    ('cext', 'parsers'): 16,
+    ('cext', 'parsers'): 17,
 }
 
 # map import request to other package or module
 _modredirects = {
     ('cext', 'charencode'): ('cext', 'parsers'),
     ('cffi', 'base85'): ('pure', 'base85'),
     ('cffi', 'charencode'): ('pure', 'charencode'),
     ('cffi', 'parsers'): ('pure', 'parsers'),
 }
 
 
 def _checkmod(pkgname, modname, mod):
     expected = _cextversions.get((pkgname, modname))
     actual = getattr(mod, 'version', None)
     if actual != expected:
         raise ImportError(
             'cannot import module %s.%s '
             '(expected version: %d, actual: %r)'
             % (pkgname, modname, expected, actual)
         )
 
 
 def importmod(modname):
     """Import module according to policy and check API version"""
     try:
         verpkg, purepkg = _packageprefs[policy]
     except KeyError:
         raise ImportError('invalid HGMODULEPOLICY %r' % policy)
     assert verpkg or purepkg
     if verpkg:
         pn, mn = _modredirects.get((verpkg, modname), (verpkg, modname))
         try:
             mod = _importfrom(pn, mn)
             if pn == verpkg:
                 _checkmod(pn, mn, mod)
             return mod
         except ImportError:
             if not purepkg:
                 raise
     pn, mn = _modredirects.get((purepkg, modname), (purepkg, modname))
     return _importfrom(pn, mn)
 
 
 def _isrustpermissive():
     """Assuming the policy is a Rust one, tell if it's permissive."""
     return policy.endswith(b'-allow')
 
 
 def importrust(modname, member=None, default=None):
     """Import Rust module according to policy and availability.
 
     If policy isn't a Rust one, this returns `default`.
 
     If either the module or its member is not available, this returns `default`
     if policy is permissive and raises `ImportError` if not.
     """
     if not policy.startswith(b'rust'):
         return default
 
     try:
         mod = _importfrom('rustext', modname)
     except ImportError:
         if _isrustpermissive():
             return default
         raise
     if member is None:
         return mod
 
     try:
         return getattr(mod, member)
     except AttributeError:
         if _isrustpermissive():
             return default
         raise ImportError("Cannot import name %s" % member)

tests/test-parseindex.t

@@ -1,223 +1,223
 revlog.parseindex must be able to parse the index file even if
 an index entry is split between two 64k blocks.  The ideal test
 would be to create an index file with inline data where
 64k < size < 64k + 64 (64k is the size of the read buffer, 64 is
 the size of an index entry) and with an index entry starting right
 before the 64k block boundary, and try to read it.
 We approximate that by reducing the read buffer to 1 byte.
 
   $ hg init a
   $ cd a
   $ echo abc > foo
   $ hg add foo
   $ hg commit -m 'add foo'
   $ echo >> foo
   $ hg commit -m 'change foo'
   $ hg log -r 0:
   changeset:   0:7c31755bf9b5
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     add foo
   
   changeset:   1:26333235a41c
   tag:         tip
   user:        test
   date:        Thu Jan 01 00:00:00 1970 +0000
   summary:     change foo
   
   $ cat >> test.py << EOF
   > from __future__ import print_function
   > from mercurial import changelog, node, pycompat, vfs
   > 
   > class singlebyteread(object):
   >     def __init__(self, real):
   >         self.real = real
   > 
   >     def read(self, size=-1):
   >         if size == 65536:
   >             size = 1
   >         return self.real.read(size)
   > 
   >     def __getattr__(self, key):
   >         return getattr(self.real, key)
   > 
   >     def __enter__(self):
   >         self.real.__enter__()
   >         return self
   > 
   >     def __exit__(self, *args, **kwargs):
   >         return self.real.__exit__(*args, **kwargs)
   > 
   > def opener(*args):
   >     o = vfs.vfs(*args)
   >     def wrapper(*a, **kwargs):
   >         f = o(*a, **kwargs)
   >         return singlebyteread(f)
   >     wrapper.options = o.options
   >     return wrapper
   > 
   > cl = changelog.changelog(opener(b'.hg/store'))
   > print(len(cl), 'revisions:')
   > for r in cl:
   >     print(pycompat.sysstr(node.short(cl.node(r))))
   > EOF
   $ "$PYTHON" test.py
   2 revisions:
   7c31755bf9b5
   26333235a41c
 
   $ cd ..
 
 #if no-pure
 
 Test SEGV caused by bad revision passed to reachableroots() (issue4775):
 
   $ cd a
 
   $ "$PYTHON" <<EOF
   > from __future__ import print_function
   > from mercurial import changelog, vfs
   > cl = changelog.changelog(vfs.vfs(b'.hg/store'))
   > print('good heads:')
   > for head in [0, len(cl) - 1, -1]:
   >     print('%s: %r' % (head, cl.reachableroots(0, [head], [0])))
   > print('bad heads:')
   > for head in [len(cl), 10000, -2, -10000, None]:
   >     print('%s:' % head, end=' ')
   >     try:
   >         cl.reachableroots(0, [head], [0])
   >         print('uncaught buffer overflow?')
   >     except (IndexError, TypeError) as inst:
   >         print(inst)
   > print('good roots:')
   > for root in [0, len(cl) - 1, -1]:
   >     print('%s: %r' % (root, cl.reachableroots(root, [len(cl) - 1], [root])))
   > print('out-of-range roots are ignored:')
   > for root in [len(cl), 10000, -2, -10000]:
   >     print('%s: %r' % (root, cl.reachableroots(root, [len(cl) - 1], [root])))
   > print('bad roots:')
   > for root in [None]:
   >     print('%s:' % root, end=' ')
   >     try:
   >         cl.reachableroots(root, [len(cl) - 1], [root])
   >         print('uncaught error?')
   >     except TypeError as inst:
   >         print(inst)
   > EOF
   good heads:
   0: [0]
   1: [0]
   -1: []
   bad heads:
   2: head out of range
   10000: head out of range
   -2: head out of range
   -10000: head out of range
   None: an integer is required( .got type NoneType.)? (re)
   good roots:
   0: [0]
   1: [1]
   -1: [-1]
   out-of-range roots are ignored:
   2: []
   10000: []
   -2: []
   -10000: []
   bad roots:
   None: an integer is required( .got type NoneType.)? (re)
 
   $ cd ..
 
 Test corrupted p1/p2 fields that could cause SEGV at parsers.c:
 
   $ mkdir invalidparent
   $ cd invalidparent
 
   $ hg clone --pull -q --config phases.publish=False ../a limit --config format.sparse-revlog=no
   $ hg clone --pull -q --config phases.publish=False ../a neglimit --config format.sparse-revlog=no
   $ hg clone --pull -q --config phases.publish=False ../a segv --config format.sparse-revlog=no
   $ rm -R limit/.hg/cache neglimit/.hg/cache segv/.hg/cache
 
   $ "$PYTHON" <<EOF
   > data = open("limit/.hg/store/00changelog.i", "rb").read()
   > poisons = [
   >     (b'limit', b'\0\0\0\x02'),
   >     (b'neglimit', b'\xff\xff\xff\xfe'),
   >     (b'segv', b'\0\x01\0\0'),
   > ]
   > for n, p in poisons:
   >     # corrupt p1 at rev0 and p2 at rev1
   >     d = data[:24] + p + data[28:127 + 28] + p + data[127 + 32:]
   >     open(n + b"/.hg/store/00changelog.i", "wb").write(d)
   > EOF
 
   $ hg -R limit debugrevlogindex -f1 -c
      rev flag     size   link     p1     p2       nodeid
        0 0000       62      0      2     -1 7c31755bf9b5
        1 0000       65      1      0      2 26333235a41c
 
   $ hg -R limit debugdeltachain -c
       rev  chain# chainlen     prev   delta       size    rawsize  chainsize     ratio   lindist extradist extraratio
         0       1        1       -1    base         63         62         63   1.01613        63         0    0.00000
         1       2        1       -1    base         66         65         66   1.01538        66         0    0.00000
 
   $ hg -R neglimit debugrevlogindex -f1 -c
      rev flag     size   link     p1     p2       nodeid
        0 0000       62      0     -2     -1 7c31755bf9b5
        1 0000       65      1      0     -2 26333235a41c
 
   $ hg -R segv debugrevlogindex -f1 -c
      rev flag     size   link     p1     p2       nodeid
        0 0000       62      0  65536     -1 7c31755bf9b5
        1 0000       65      1      0  65536 26333235a41c
 
   $ hg -R segv debugdeltachain -c
       rev  chain# chainlen     prev   delta       size    rawsize  chainsize     ratio   lindist extradist extraratio
         0       1        1       -1    base         63         62         63   1.01613        63         0    0.00000
         1       2        1       -1    base         66         65         66   1.01538        66         0    0.00000
 
   $ cat <<EOF > test.py
   > from __future__ import print_function
   > import sys
   > from mercurial import changelog, pycompat, vfs
   > cl = changelog.changelog(vfs.vfs(pycompat.fsencode(sys.argv[1])))
   > n0, n1 = cl.node(0), cl.node(1)
   > ops = [
   >     ('reachableroots',
   >      lambda: cl.index.reachableroots2(0, [1], [0], False)),
-  >     ('compute_phases_map_sets', lambda: cl.computephases([[0], []])),
+  >     ('compute_phases_map_sets', lambda: cl.computephases({1: {cl.node(0)}})),
   >     ('index_headrevs', lambda: cl.headrevs()),
   >     ('find_gca_candidates', lambda: cl.commonancestorsheads(n0, n1)),
   >     ('find_deepest', lambda: cl.ancestor(n0, n1)),
   >     ]
   > for l, f in ops:
   >     print(l + ':', end=' ')
   >     try:
   >         f()
   >         print('uncaught buffer overflow?')
   >     except ValueError as inst:
   >         print(inst)
   > EOF
 
   $ "$PYTHON" test.py limit/.hg/store
   reachableroots: parent out of range
   compute_phases_map_sets: parent out of range
   index_headrevs: parent out of range
   find_gca_candidates: parent out of range
   find_deepest: parent out of range
   $ "$PYTHON" test.py neglimit/.hg/store
   reachableroots: parent out of range
   compute_phases_map_sets: parent out of range
   index_headrevs: parent out of range
   find_gca_candidates: parent out of range
   find_deepest: parent out of range
   $ "$PYTHON" test.py segv/.hg/store
   reachableroots: parent out of range
   compute_phases_map_sets: parent out of range
   index_headrevs: parent out of range
   find_gca_candidates: parent out of range
   find_deepest: parent out of range
 
   $ cd ..
 
 #endif