##// END OF EJS Templates
merge: don't try to merge subrepos twice (issue4988)...
merge: don't try to merge subrepos twice (issue4988) In my patch series ending with rev 25e4b2f000c5 I switched most change/delete conflicts to be handled at the resolve layer. .hgsubstate was the one file that we weren't able to handle, so we kept the old code path around for it. The old code path added .hgsubstate to one of the other lists as the user specifies, including possibly the 'g' list. Now since we did this check after converting the actions from being keyed by file to being keyed by action type, there was nothing that actually removed .hgsubstate from the 'cd' or 'dc' lists. This meant that the file would eventually make its way into the 'mergeactions' list, now freshly augmented with 'cd' and 'dc' actions. We call subrepo.submerge for both 'g' actions and merge actions. This means that if the resolution to an .hgsubstate change/delete conflict was to add it to the 'g' list, subrepo.submerge would be called twice. It turns out that this doesn't cause any adverse effects on Linux due to caching, but apparently breaks on other operating systems including Windows. The fix here moves this to before we convert the actions over. This ensures that it .hgsubstate doesn't make its way into multiple lists. The real fix here is going to be: (1) move .hgsubstate conflict resolution into the resolve layer, and (2) use a real data structure for the actions rather than shuffling data around between lists and dictionaries: we need a hash (or prefix-based) index by file and a list index by action type. There's a very tiny behavior change here: collision detection on case-insensitive systems will happen after this is resolved, not before. I think this is the right change -- .hgsubstate could theoretically collide with other files -- but in any case it makes no practical difference. Thanks to Yuya Nishihara for investigating this.

File last commit:

r20167:09e41ac6 stable
r27951:6bce6d92 stable
Show More
mpatch.c
420 lines | 8.5 KiB | text/x-c | CLexer
/*
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, 2006 Matt Mackall <mpm@selenic.com>
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 <stdlib.h>
#include <string.h>
#include "util.h"
static char mpatch_doc[] = "Efficient binary patching.";
static PyObject *mpatch_Error;
struct frag {
int start, end, len;
const char *data;
};
struct flist {
struct frag *base, *head, *tail;
};
static struct flist *lalloc(Py_ssize_t size)
{
struct flist *a = NULL;
if (size < 1)
size = 1;
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 Py_ssize_t 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(const char *bin, Py_ssize_t len)
{
struct flist *l;
struct frag *lt;
int pos = 0;
/* assume worst case size, we won't have many of these lists */
l = lalloc(len / 12);
if (!l)
return NULL;
lt = l->tail;
while (pos >= 0 && pos < len) {
lt->start = getbe32(bin + pos);
lt->end = getbe32(bin + pos + 4);
lt->len = getbe32(bin + pos + 8);
if (lt->start > lt->end)
break; /* sanity check */
lt->data = bin + pos + 12;
pos += 12 + lt->len;
lt++;
}
if (pos != len) {
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 Py_ssize_t calcsize(Py_ssize_t len, struct flist *l)
{
Py_ssize_t 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, const char *orig, Py_ssize_t 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, Py_ssize_t start, Py_ssize_t end)
{
Py_ssize_t len, blen;
const char *buffer;
if (start + 1 == end) {
/* trivial case, output a decoded list */
PyObject *tmp = PyList_GetItem(bins, start);
if (!tmp)
return NULL;
if (PyObject_AsCharBuffer(tmp, &buffer, &blen))
return NULL;
return decode(buffer, blen);
}
/* 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;
const char *in;
char *out;
Py_ssize_t len, outlen, inlen;
if (!PyArg_ParseTuple(args, "OO:mpatch", &text, &bins))
return NULL;
len = PyList_Size(bins);
if (!len) {
/* nothing to do */
Py_INCREF(text);
return text;
}
if (PyObject_AsCharBuffer(text, &in, &inlen))
return NULL;
patch = fold(bins, 0, len);
if (!patch)
return NULL;
outlen = calcsize(inlen, patch);
if (outlen < 0) {
result = NULL;
goto cleanup;
}
result = PyBytes_FromStringAndSize(NULL, outlen);
if (!result) {
result = NULL;
goto cleanup;
}
out = PyBytes_AsString(result);
if (!apply(out, in, inlen, 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, pos = 0;
Py_ssize_t patchlen;
char *bin;
if (!PyArg_ParseTuple(args, "ls#", &orig, &bin, &patchlen))
return NULL;
while (pos >= 0 && pos < patchlen) {
start = getbe32(bin + pos);
end = getbe32(bin + pos + 4);
len = getbe32(bin + pos + 8);
if (start > end)
break; /* sanity check */
pos += 12 + len;
outlen += start - last;
last = end;
outlen += len;
}
if (pos != patchlen) {
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}
};
#ifdef IS_PY3K
static struct PyModuleDef mpatch_module = {
PyModuleDef_HEAD_INIT,
"mpatch",
mpatch_doc,
-1,
methods
};
PyMODINIT_FUNC PyInit_mpatch(void)
{
PyObject *m;
m = PyModule_Create(&mpatch_module);
if (m == NULL)
return NULL;
mpatch_Error = PyErr_NewException("mpatch.mpatchError", NULL, NULL);
Py_INCREF(mpatch_Error);
PyModule_AddObject(m, "mpatchError", mpatch_Error);
return m;
}
#else
PyMODINIT_FUNC
initmpatch(void)
{
Py_InitModule3("mpatch", methods, mpatch_doc);
mpatch_Error = PyErr_NewException("mpatch.mpatchError", NULL, NULL);
}
#endif