mac os x: fixes for 10.2 from chris monson <[e-mail]>

mercurial/bdiff.c

@@ -1,361 +1,362
 /*
  bdiff.c - efficient binary diff extension for Mercurial
 
  Copyright 2005 Matt Mackall <[e-mail]>
 
  This software may be used and distributed according to the terms of
  the GNU General Public License, incorporated herein by reference.
 
  Based roughly on Python difflib
 */
 
 #include <Python.h>
+#include <stdint.h>
 #include <stdlib.h>
 #include <string.h>
 
 #ifdef __hpux
 #define inline
 #endif
 
 #ifdef __SUNPRO_C
 # define inline
 #endif 
 
 #ifdef _WIN32
 #ifdef _MSC_VER
 #define inline __inline
 typedef unsigned long uint32_t;
 #else
 #include <stdint.h>
 #endif
 static uint32_t htonl(uint32_t x)
 {
 	return ((x & 0x000000ffUL) << 24) |
 		((x & 0x0000ff00UL) <<  8) |
 		((x & 0x00ff0000UL) >>  8) |
 		((x & 0xff000000UL) >> 24);
 }
 #else
 #include <sys/types.h>
 #include <arpa/inet.h>
 #endif
 
 struct line {
 	int h, len, n, e;
 	const char *l;
 };
 
 struct pos {
 	int pos, len;
 };
 
 struct hunk {
 	int a1, a2, b1, b2;
 };
 
 struct hunklist {
 	struct hunk *base, *head;
 };
 
 static inline uint32_t rol32(uint32_t word, unsigned int shift)
 {
         return (word << shift) | (word >> (32 - shift));
 }
 
 int splitlines(const char *a, int len, struct line **lr)
 {
 	int h, i;
 	const char *p, *b = a;
 	struct line *l;
 
 	/* count the lines */
 	i = 1; /* extra line for sentinel */
 	for (p = a; p < a + len; p++)
 		if (*p == '\n' || p == a + len - 1)
 			i++;
 
 	*lr = l = (struct line *)malloc(sizeof(struct line) * i);
 	if (!l)
 		return -1;
 
 	/* build the line array and calculate hashes */
 	h = 0;
 	for (p = a; p < a + len; p++) {
 		h = *p + rol32(h, 7); /* a simple hash from GNU diff */
 		if (*p == '\n' || p == a + len - 1) {
 			l->len = p - b + 1;
 			l->h = h * l->len;
 			l->l = b;
 			l->n = -1;
 			l++;
 			b = p + 1;
 			h = 0;
 		}
 	}
 
 	/* set up a sentinel */
 	l->h = l->len = 0;
 	l->l = a + len;
 	return i - 1;
 }
 
 int inline cmp(struct line *a, struct line *b)
 {
 	return a->h != b->h || a->len != b->len || memcmp(a->l, b->l, a->len);
 }
 
 static int equatelines(struct line *a, int an, struct line *b, int bn)
 {
 	int i, j, buckets = 1, t;
 	struct pos *h;
 
 	/* build a hash table of the next highest power of 2 */
 	while (buckets < bn + 1)
 		buckets *= 2;
 
 	h = (struct pos *)malloc(buckets * sizeof(struct pos));
 	buckets = buckets - 1;
 	if (!h)
 		return 0;
 
 	/* clear the hash table */
 	for (i = 0; i <= buckets; i++) {
 		h[i].pos = -1;
 		h[i].len = 0;
 	}
 
 	/* add lines to the hash table chains */
 	for (i = bn - 1; i >= 0; i--) {
 		/* find the equivalence class */
 		for (j = b[i].h & buckets; h[j].pos != -1;
 		     j = (j + 1) & buckets)
 			if (!cmp(b + i, b + h[j].pos))
 				break;
 
 		/* add to the head of the equivalence class */
 		b[i].n = h[j].pos;
 		b[i].e = j;
 		h[j].pos = i;
 		h[j].len++; /* keep track of popularity */
 	}
 
 	/* compute popularity threshold */
 	t = (bn >= 200) ? bn / 100 : bn + 1;
 
 	/* match items in a to their equivalence class in b */
 	for (i = 0; i < an; i++) {
 		/* find the equivalence class */
 		for (j = a[i].h & buckets; h[j].pos != -1;
 		     j = (j + 1) & buckets)
 			if (!cmp(a + i, b + h[j].pos))
 				break;
 
 		a[i].e = j; /* use equivalence class for quick compare */
 		if (h[j].len <= t)
 			a[i].n = h[j].pos; /* point to head of match list */
 		else
 			a[i].n = -1; /* too popular */
 	}
 
 	/* discard hash tables */
 	free(h);
 	return 1;
 }
 
 static int longest_match(struct line *a, struct line *b, struct pos *pos,
 			 int a1, int a2, int b1, int b2, int *omi, int *omj)
 {
 	int mi = a1, mj = b1, mk = 0, mb = 0, i, j, k;
 
 	for (i = a1; i < a2; i++) {
 		/* skip things before the current block */
 		for (j = a[i].n; j != -1 && j < b1; j = b[j].n)
 			;
 
 		/* loop through all lines match a[i] in b */
 		for (; j != -1 && j < b2; j = b[j].n) {
 			/* does this extend an earlier match? */
 			if (i > a1 && j > b1 && pos[j - 1].pos == i - 1)
 				k = pos[j - 1].len + 1;
 			else
 				k = 1;
 			pos[j].pos = i;
 			pos[j].len = k;
 
 			/* best match so far? */
 			if (k > mk) {
 				mi = i;
 				mj = j;
 				mk = k;
 			}
 		}
 	}
 
 	if (mk) {
 		mi = mi - mk + 1;
 		mj = mj - mk + 1;
 	}
 
 	/* expand match to include neighboring popular lines */
 	while (mi - mb > a1 && mj - mb > b1 &&
 	       a[mi - mb - 1].e == b[mj - mb - 1].e)
 		mb++;
 	while (mi + mk < a2 && mj + mk < b2 &&
 	       a[mi + mk].e == b[mj + mk].e)
 		mk++;
 
 	*omi = mi - mb;
 	*omj = mj - mb;
 	return mk + mb;
 }
 
 static void recurse(struct line *a, struct line *b, struct pos *pos,
 		    int a1, int a2, int b1, int b2, struct hunklist *l)
 {
 	int i, j, k;
 
 	/* find the longest match in this chunk */
 	k = longest_match(a, b, pos, a1, a2, b1, b2, &i, &j);
 	if (!k)
 		return;
 
 	/* and recurse on the remaining chunks on either side */
 	recurse(a, b, pos, a1, i, b1, j, l);
 	l->head->a1 = i;
 	l->head->a2 = i + k;
 	l->head->b1 = j;
 	l->head->b2 = j + k;
 	l->head++;
 	recurse(a, b, pos, i + k, a2, j + k, b2, l);
 }
 
 static struct hunklist diff(struct line *a, int an, struct line *b, int bn)
 {
 	struct hunklist l;
 	struct pos *pos;
 	int t;
 
 	/* allocate and fill arrays */
 	t = equatelines(a, an, b, bn);
 	pos = (struct pos *)calloc(bn, sizeof(struct pos));
 	/* we can't have more matches than lines in the shorter file */
 	l.head = l.base = (struct hunk *)malloc(sizeof(struct hunk) *
 	                                        ((an<bn ? an:bn) + 1));
 
 	if (pos && l.base && t) {
 		/* generate the matching block list */
 		recurse(a, b, pos, 0, an, 0, bn, &l);
 		l.head->a1 = an;
 		l.head->b1 = bn;
 		l.head++;
 	}
 
 	free(pos);
 	return l;
 }
 
 static PyObject *blocks(PyObject *self, PyObject *args)
 {
 	PyObject *sa, *sb, *rl = NULL, *m;
 	struct line *a, *b;
 	struct hunklist l = {NULL, NULL};
 	struct hunk *h;
 	int an, bn, pos = 0;
 
 	if (!PyArg_ParseTuple(args, "SS:bdiff", &sa, &sb))
 		return NULL;
 
 	an = splitlines(PyString_AsString(sa), PyString_Size(sa), &a);
 	bn = splitlines(PyString_AsString(sb), PyString_Size(sb), &b);
 	if (!a || !b)
 		goto nomem;
 
 	l = diff(a, an, b, bn);
 	rl = PyList_New(l.head - l.base);
 	if (!l.head || !rl)
 		goto nomem;
 
 	for (h = l.base; h != l.head; h++) {
 		m = Py_BuildValue("iiii", h->a1, h->a2, h->b1, h->b2);
 		PyList_SetItem(rl, pos, m);
 		pos++;
 	}
 
 nomem:
 	free(a);
 	free(b);
 	free(l.base);
 	return rl ? rl : PyErr_NoMemory();
 }
 
 static PyObject *bdiff(PyObject *self, PyObject *args)
 {
 	PyObject *sa, *sb, *result = NULL;
 	struct line *al, *bl;
 	struct hunklist l = {NULL, NULL};
 	struct hunk *h;
 	char encode[12], *rb;
 	int an, bn, len = 0, la = 0, lb = 0;
 
 	if (!PyArg_ParseTuple(args, "SS:bdiff", &sa, &sb))
 		return NULL;
 
 	an = splitlines(PyString_AsString(sa), PyString_Size(sa), &al);
 	bn = splitlines(PyString_AsString(sb), PyString_Size(sb), &bl);
 	if (!al || !bl)
 		goto nomem;
 
 	l = diff(al, an, bl, bn);
 	if (!l.head)
 		goto nomem;
 
 	/* calculate length of output */
 	for (h = l.base; h != l.head; h++) {
 		if (h->a1 != la || h->b1 != lb)
 			len += 12 + bl[h->b1].l - bl[lb].l;
 		la = h->a2;
 		lb = h->b2;
 	}
 
 	result = PyString_FromStringAndSize(NULL, len);
 	if (!result)
 		goto nomem;
 
 	/* build binary patch */
 	rb = PyString_AsString(result);
 	la = lb = 0;
 
 	for (h = l.base; h != l.head; h++) {
 		if (h->a1 != la || h->b1 != lb) {
 			len = bl[h->b1].l - bl[lb].l;
 			*(uint32_t *)(encode)     = htonl(al[la].l - al->l);
 			*(uint32_t *)(encode + 4) = htonl(al[h->a1].l - al->l);
 			*(uint32_t *)(encode + 8) = htonl(len);
 			memcpy(rb, encode, 12);
 			memcpy(rb + 12, bl[lb].l, len);
 			rb += 12 + len;
 		}
 		la = h->a2;
 		lb = h->b2;
 	}
 
 nomem:
 	free(al);
 	free(bl);
 	free(l.base);
 	return result ? result : PyErr_NoMemory();
 }
 
 static char mdiff_doc[] = "Efficient binary diff.";
 
 static PyMethodDef methods[] = {
 	{"bdiff", bdiff, METH_VARARGS, "calculate a binary diff\n"},
 	{"blocks", blocks, METH_VARARGS, "find a list of matching lines\n"},
 	{NULL, NULL}
 };
 
 PyMODINIT_FUNC initbdiff(void)
 {
 	Py_InitModule3("bdiff", methods, mdiff_doc);
 }
 

mercurial/mpatch.c

@@ -1,404 +1,408
 /*
  mpatch.c - efficient binary patching for Mercurial
 
  This implements a patch algorithm that's O(m + nlog n) where m is the
  size of the output and n is the number of patches.
 
  Given a list of binary patches, it unpacks each into a hunk list,
  then combines the hunk lists with a treewise recursion to form a
  single hunk list. This hunk list is then applied to the original
  text.
 
  The text (or binary) fragments are copied directly from their source
  Python objects into a preallocated output string to avoid the
  allocation of intermediate Python objects. Working memory is about 2x
  the total number of hunks.
 
  Copyright 2005 Matt Mackall <[e-mail]>
 
  This software may be used and distributed according to the terms
  of the GNU General Public License, incorporated herein by reference.
 */
 
 #include <Python.h>
 #include <stdlib.h>
 #include <string.h>
+
 #ifdef _WIN32
 #ifdef _MSC_VER
+/* msvc 6.0 has problems */
 #define inline __inline
 typedef unsigned long uint32_t;
 #else
 #include <stdint.h>
 #endif
 static uint32_t ntohl(uint32_t x)
 {
 	return ((x & 0x000000ffUL) << 24) |
 		((x & 0x0000ff00UL) <<  8) |
 		((x & 0x00ff0000UL) >>  8) |
 		((x & 0xff000000UL) >> 24);
 }
 #else
+/* not windows */
 #include <sys/types.h>
 #include <arpa/inet.h>
+# include <stdint.h>
 #endif
 
 static char mpatch_doc[] = "Efficient binary patching.";
 static PyObject *mpatch_Error;
 
 struct frag {
 	int start, end, len;
 	char *data;
 };
 
 struct flist {
 	struct frag *base, *head, *tail;
 };
 
 static struct flist *lalloc(int size)
 {
 	struct flist *a = NULL;
 
 	a = (struct flist *)malloc(sizeof(struct flist));
 	if (a) {
 		a->base = (struct frag *)malloc(sizeof(struct frag) * size);
 		if (a->base) {
 			a->head = a->tail = a->base;
 			return a;
 		}
 		free(a);
 		a = NULL;
 	}
 	if (!PyErr_Occurred())
 		PyErr_NoMemory();
 	return NULL;
 }
 
 static void lfree(struct flist *a)
 {
 	if (a) {
 		free(a->base);
 		free(a);
 	}
 }
 
 static int lsize(struct flist *a)
 {
 	return a->tail - a->head;
 }
 
 /* move hunks in source that are less cut to dest, compensating
    for changes in offset. the last hunk may be split if necessary.
 */
 static int gather(struct flist *dest, struct flist *src, int cut, int offset)
 {
 	struct frag *d = dest->tail, *s = src->head;
 	int postend, c, l;
 
 	while (s != src->tail) {
 		if (s->start + offset >= cut)
 			break; /* we've gone far enough */
 
 		postend = offset + s->start + s->len;
 		if (postend <= cut) {
 			/* save this hunk */
 			offset += s->start + s->len - s->end;
 			*d++ = *s++;
 		}
 		else {
 			/* break up this hunk */
 			c = cut - offset;
 			if (s->end < c)
 				c = s->end;
 			l = cut - offset - s->start;
 			if (s->len < l)
 				l = s->len;
 
 			offset += s->start + l - c;
 
 			d->start = s->start;
 			d->end = c;
 			d->len = l;
 			d->data = s->data;
 			d++;
 			s->start = c;
 			s->len = s->len - l;
 			s->data = s->data + l;
 
 			break;
 		}
 	}
 
 	dest->tail = d;
 	src->head = s;
 	return offset;
 }
 
 /* like gather, but with no output list */
 static int discard(struct flist *src, int cut, int offset)
 {
 	struct frag *s = src->head;
 	int postend, c, l;
 
 	while (s != src->tail) {
 		if (s->start + offset >= cut)
 			break;
 
 		postend = offset + s->start + s->len;
 		if (postend <= cut) {
 			offset += s->start + s->len - s->end;
 			s++;
 		}
 		else {
 			c = cut - offset;
 			if (s->end < c)
 				c = s->end;
 			l = cut - offset - s->start;
 			if (s->len < l)
 				l = s->len;
 
 			offset += s->start + l - c;
 			s->start = c;
 			s->len = s->len - l;
 			s->data = s->data + l;
 
 			break;
 		}
 	}
 
 	src->head = s;
 	return offset;
 }
 
 /* combine hunk lists a and b, while adjusting b for offset changes in a/
    this deletes a and b and returns the resultant list. */
 static struct flist *combine(struct flist *a, struct flist *b)
 {
 	struct flist *c = NULL;
 	struct frag *bh, *ct;
 	int offset = 0, post;
 
 	if (a && b)
 		c = lalloc((lsize(a) + lsize(b)) * 2);
 
 	if (c) {
 
 		for (bh = b->head; bh != b->tail; bh++) {
 			/* save old hunks */
 			offset = gather(c, a, bh->start, offset);
 
 			/* discard replaced hunks */
 			post = discard(a, bh->end, offset);
 
 			/* insert new hunk */
 			ct = c->tail;
 			ct->start = bh->start - offset;
 			ct->end = bh->end - post;
 			ct->len = bh->len;
 			ct->data = bh->data;
 			c->tail++;
 			offset = post;
 		}
 
 		/* hold on to tail from a */
 		memcpy(c->tail, a->head, sizeof(struct frag) * lsize(a));
 		c->tail += lsize(a);
 	}
 
 	lfree(a);
 	lfree(b);
 	return c;
 }
 
 /* decode a binary patch into a hunk list */
 static struct flist *decode(char *bin, int len)
 {
 	struct flist *l;
 	struct frag *lt;
 	char *end = bin + len;
 	char decode[12]; /* for dealing with alignment issues */
 
 	/* assume worst case size, we won't have many of these lists */
 	l = lalloc(len / 12);
 	if (!l)
 		return NULL;
 
 	lt = l->tail;
 
 	while (bin < end) {
 		memcpy(decode, bin, 12);
 		lt->start = ntohl(*(uint32_t *)decode);
 		lt->end = ntohl(*(uint32_t *)(decode + 4));
 		lt->len = ntohl(*(uint32_t *)(decode + 8));
 		lt->data = bin + 12;
 		bin += 12 + lt->len;
 		lt++;
 	}
 
 	if (bin != end) {
 		if (!PyErr_Occurred())
 			PyErr_SetString(mpatch_Error, "patch cannot be decoded");
 		lfree(l);
 		return NULL;
 	}
 
 	l->tail = lt;
 	return l;
 }
 
 /* calculate the size of resultant text */
 static int calcsize(int len, struct flist *l)
 {
 	int outlen = 0, last = 0;
 	struct frag *f = l->head;
 
 	while (f != l->tail) {
 		if (f->start < last || f->end > len) {
 			if (!PyErr_Occurred())
 				PyErr_SetString(mpatch_Error,
 				                "invalid patch");
 			return -1;
 		}
 		outlen += f->start - last;
 		last = f->end;
 		outlen += f->len;
 		f++;
 	}
 
 	outlen += len - last;
 	return outlen;
 }
 
 static int apply(char *buf, char *orig, int len, struct flist *l)
 {
 	struct frag *f = l->head;
 	int last = 0;
 	char *p = buf;
 
 	while (f != l->tail) {
 		if (f->start < last || f->end > len) {
 			if (!PyErr_Occurred())
 				PyErr_SetString(mpatch_Error,
 				                "invalid patch");
 			return 0;
 		}
 		memcpy(p, orig + last, f->start - last);
 		p += f->start - last;
 		memcpy(p, f->data, f->len);
 		last = f->end;
 		p += f->len;
 		f++;
 	}
 	memcpy(p, orig + last, len - last);
 	return 1;
 }
 
 /* recursively generate a patch of all bins between start and end */
 static struct flist *fold(PyObject *bins, int start, int end)
 {
 	int len;
 
 	if (start + 1 == end) {
 		/* trivial case, output a decoded list */
 		PyObject *tmp = PyList_GetItem(bins, start);
 		if (!tmp)
 			return NULL;
 		return decode(PyString_AsString(tmp), PyString_Size(tmp));
 	}
 
 	/* divide and conquer, memory management is elsewhere */
 	len = (end - start) / 2;
 	return combine(fold(bins, start, start + len),
 		       fold(bins, start + len, end));
 }
 
 static PyObject *
 patches(PyObject *self, PyObject *args)
 {
 	PyObject *text, *bins, *result;
 	struct flist *patch;
 	char *in, *out;
 	int len, outlen;
 
 	if (!PyArg_ParseTuple(args, "SO:mpatch", &text, &bins))
 		return NULL;
 
 	len = PyList_Size(bins);
 	if (!len) {
 		/* nothing to do */
 		Py_INCREF(text);
 		return text;
 	}
 
 	patch = fold(bins, 0, len);
 	if (!patch)
 		return NULL;
 
 	outlen = calcsize(PyString_Size(text), patch);
 	if (outlen < 0) {
 		result = NULL;
 		goto cleanup;
 	}
 	result = PyString_FromStringAndSize(NULL, outlen);
 	if (!result) {
 		result = NULL;
 		goto cleanup;
 	}
 	in = PyString_AsString(text);
 	out = PyString_AsString(result);
 	if (!apply(out, in, PyString_Size(text), patch)) {
 		Py_DECREF(result);
 		result = NULL;
 	}
 cleanup:
 	lfree(patch);
 	return result;
 }
 
 /* calculate size of a patched file directly */
 static PyObject *
 patchedsize(PyObject *self, PyObject *args)
 {
 	long orig, start, end, len, outlen = 0, last = 0;
 	int patchlen;
 	char *bin, *binend;
 	char decode[12]; /* for dealing with alignment issues */
 
 	if (!PyArg_ParseTuple(args, "ls#", &orig, &bin, &patchlen))
 		return NULL;
 
 	binend = bin + patchlen;
 
 	while (bin < binend) {
 		memcpy(decode, bin, 12);
 		start = ntohl(*(uint32_t *)decode);
 		end = ntohl(*(uint32_t *)(decode + 4));
 		len = ntohl(*(uint32_t *)(decode + 8));
 		bin += 12 + len;
 		outlen += start - last;
 		last = end;
 		outlen += len;
 	}
 
 	if (bin != binend) {
 		if (!PyErr_Occurred())
 			PyErr_SetString(mpatch_Error, "patch cannot be decoded");
 		return NULL;
 	}
 
 	outlen += orig - last;
 	return Py_BuildValue("l", outlen);
 }
 
 static PyMethodDef methods[] = {
 	{"patches", patches, METH_VARARGS, "apply a series of patches\n"},
 	{"patchedsize", patchedsize, METH_VARARGS, "calculed patched size\n"},
 	{NULL, NULL}
 };
 
 PyMODINIT_FUNC
 initmpatch(void)
 {
 	Py_InitModule3("mpatch", methods, mpatch_doc);
 	mpatch_Error = PyErr_NewException("mpatch.mpatchError", NULL, NULL);
 }