deal with two error cases

Signed-off-by: Patrick McCarty <patrick.mccarty@intel.com>

.gitignore

@@ -1,4 +1,8 @@
+*~
+*.swp
 *.o
+*.i
+*.s
 .libs/
 *.lo
 *.pc
@@ -19,13 +23,15 @@ install-sh
 libbsdiff.la
 libtool
 ltmain.sh
-m4/
+m4/*
+!m4/.gitignore
 missing
 src/.dirstamp
 src/*.gcda
 src/*.gcno
 stamp-h1
-bsdiff-*.tar.xz
+/bsdiff-*.tar.xz
+/bsdiff-*/
 test/*.diff
 test/*.out
 test/*.log

Makefile.am

@@ -97,11 +97,18 @@ libbsdiff_la_LDFLAGS = \
 	-version-info $(LIBBSDIFF_CURRENT):$(LIBBSDIFF_REVISION):$(LIBBSDIFF_AGE) \
 	-Wl,--version-script=$(top_srcdir)/src/bsdiff.sym
 
+mostlyclean-local:
+	-rm -f *.i
+	-rm -f *.s
+
 distclean-local:
-	rm -rf aclocal.m4 ar-lib autom4te.cache config.guess config.h.in config.h.in~ config.sub configure depcomp install-sh ltmain.sh m4 Makefile.in missing compile
+	-rm -f config.guess~
+	-rm -f config.h.in~
+	-rm -f config.sub~
+	-rm -f configure~
 
 install-exec-hook:
-	perl findstatic.pl */*.o | grep -v Checking ||:
+	perl $(top_srcdir)/findstatic.pl $(top_builddir)/src/*.o | grep -v Checking || :
 
 TEST_EXTENSIONS = .sh
 

configure.ac

@@ -1,8 +1,7 @@
 AC_PREREQ([2.66])
-AC_INIT([bsdiff], [1.0.4], [patrick.mccarty@intel.com])
+AC_INIT([bsdiff],[1.0.4],[patrick.mccarty@intel.com])
 AC_CONFIG_MACRO_DIR([m4])
 AC_PROG_CC
-AC_PROG_CC_STDC
 AC_LANG(C)
 AC_CONFIG_HEADERS([config.h])
 AC_PREFIX_DEFAULT(/usr/local)

src/diff.c

@@ -74,12 +74,12 @@ static int bsdiff_fulldl;
 #undef MIN
 #define MIN(x, y) (((x) < (y)) ? (x) : (y))
 
-static int64_t matchlen(u_char *old, int64_t oldsize, u_char *new,
+static int64_t matchlen(u_char *old, int64_t old_size, u_char *new,
-			int64_t newsize)
+			int64_t new_size)
 {
 	int64_t i;
 
-	for (i = 0; (i < oldsize) && (i < newsize); i++) {
+	for (i = 0; (i < old_size) && (i < new_size); i++) {
 		if (old[i] != new[i]) {
 			break;
 		}
@@ -92,34 +92,34 @@ static int64_t matchlen(u_char *old, int64_t oldsize, u_char *new,
  * Finds the longest matching array of bytes between the OLD and NEW file. The
  * old file is suffix-sorted; the suffix-sorted array is stored at I, and
  * indices to search between are indicated by ST (start) and EN (end). The
- * function does not return a value, but once a match is determined, POS is
+ * function does not return a value, but once a match is determined, OLD_POS is
  * updated to the position of the match within OLD, and MAX_LEN is set to the
  * match length.
  */
-static void search(int64_t *I, u_char *old, int64_t oldsize,
+static void search(int64_t *I, u_char *old, int64_t old_size,
-		   u_char *new, int64_t newsize, int64_t st, int64_t en,
+		   u_char *new, int64_t new_size, int64_t st, int64_t en,
-		   int64_t *pos, int64_t *max_len)
+		   int64_t *old_pos, int64_t *max_len)
 {
 	int64_t x, y;
 
 	/* Initialize max_len for the binary search */
-	if (st == 0 && en == oldsize) {
+	if (st == 0 && en == old_size) {
-		*max_len = matchlen(old, oldsize, new, newsize);
+		*max_len = matchlen(old, old_size, new, new_size);
-		*pos = I[st];
+		*old_pos = I[st];
 	}
 
 	/* The binary search terminates here when "en" and "st" are adjacent
 	 * indices in the suffix-sorted array. */
 	if (en - st < 2) {
-		x = matchlen(old + I[st], oldsize - I[st], new, newsize);
+		x = matchlen(old + I[st], old_size - I[st], new, new_size);
 		if (x > *max_len) {
 			*max_len = x;
-			*pos = I[st];
+			*old_pos = I[st];
 		}
-		y = matchlen(old + I[en], oldsize - I[en], new, newsize);
+		y = matchlen(old + I[en], old_size - I[en], new, new_size);
 		if (y > *max_len) {
 			*max_len = y;
-			*pos = I[en];
+			*old_pos = I[en];
 		}
 
 		return;
@@ -127,21 +127,21 @@ static void search(int64_t *I, u_char *old, int64_t oldsize,
 
 	x = st + (en - st) / 2;
 
-	int64_t length = MIN(oldsize - I[x], newsize);
+	int64_t length = MIN(old_size - I[x], new_size);
 	u_char *oldoffset = old + I[x];
 
 	/* This match *could* be the longest one, so check for that here */
 	int64_t tmp = matchlen(oldoffset, length, new, length);
 	if (tmp > *max_len) {
 		*max_len = tmp;
-		*pos = I[x];
+		*old_pos = I[x];
 	}
 
 	/* Determine how to continue the binary search */
 	if (memcmp(oldoffset, new, length) < 0) {
-		return search(I, old, oldsize, new, newsize, x, en, pos, max_len);
+		return search(I, old, old_size, new, new_size, x, en, old_pos, max_len);
 	} else {
-		return search(I, old, oldsize, new, newsize, st, x, pos, max_len);
+		return search(I, old, old_size, new, new_size, st, x, old_pos, max_len);
 	}
 }
 
@@ -386,16 +386,8 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 {
 	int fd, efd;
 	u_char *old_data, *new_data;
-	int64_t oldsize, newsize;
+	int64_t old_size, new_size;
 	int64_t *I, *V;
-	int64_t scan;
-	int64_t pos = 0;
-	int64_t len;
-	int64_t lastscan, lastpos, lastoffset;
-	int64_t oldscore, scsc;
-	int64_t s, Sf, lenf, Sb, lenb;
-	int64_t overlap, Ss, lens;
-	int64_t i;
 	uint64_t cblen, dblen, eblen;
 	u_char *cb, *db, *eb;
 	struct stat new_stat;
@@ -404,9 +396,12 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 	off_t first_block;
 	int c_enc, d_enc, e_enc;
 	enc_flags_t encodings;
+	char delta_filename_unique[2 * PATH_MAX];
 
 	struct header_v20 large_header;
 	struct header_v21 small_header;
+
+	sprintf(delta_filename_unique, "%s.%i", delta_filename, getpid());
 	FILE *pf;
 
 	ret = lstat(old_filename, &old_stat);
@@ -419,6 +414,8 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		return -1;
 	}
 
+	ret = 0;
+
 	if (S_ISDIR(new_stat.st_mode) || S_ISDIR(old_stat.st_mode)) {
 		/* no delta on symlinks ! */
 		return -1;
@@ -439,16 +436,16 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		return -1;
 	}
 
-	oldsize = old_stat.st_size;
+	old_size = old_stat.st_size;
 
 	/* We may start with an empty file, if so, just mark it for full download
 	 * to throw into the pack. In the case that newfile is <200, it will quit
-	 * and ask for fulldownload, so we only need to check oldsize */
+	 * and ask for fulldownload, so we only need to check old_size */
-	if (oldsize == 0) {
+	if (old_size == 0) {
 		memset(&small_header, 0, sizeof(struct header_v21));
 		memcpy(&small_header.magic, BSDIFF_HDR_FULLDL, 8);
 
-		efd = open(delta_filename, O_CREAT | O_EXCL | O_WRONLY, 00644);
+		efd = open(delta_filename_unique, O_CREAT | O_EXCL | O_WRONLY, 00644);
 		if (efd < 0) {
 			close(fd);
 			return -1;
@@ -465,15 +462,16 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		}
 		fclose(pf);
 		close(fd);
+		rename(delta_filename_unique, delta_filename);		
 		return 1;
 	}
 
 	/* TODO: investigate why this needs to be +1 to not overrun; coverity complains
 	 * that we overrun old_data when we calculate differences otherwise. Tenatively,
 	 * since this is used in qsufsort, it may need to be +1 like I and V because of
-	 * a sentinel byte when sorting. However, newsize does not cause any overruns
+	 * a sentinel byte when sorting. However, new_size does not cause any overruns
 	 * when created with the regular file size */
-	old_data = mmap(NULL, oldsize + 1, PROT_READ, MAP_SHARED, fd, 0);
+	old_data = mmap(NULL, old_size + 1, PROT_READ, MAP_SHARED, fd, 0);
 	close(fd);
 
 	if (old_data == MAP_FAILED) {
@@ -483,19 +481,19 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 
 	/* These arrays are size + 1 because suffix sort needs space for the
 	 * data + 1 sentinel element to actually do the sorting. Not because
-	 * oldsize might be 0. */
+	 * old_size might be 0. */
-	if ((I = malloc((oldsize + 1) * sizeof(int64_t))) == NULL) {
+	if ((I = malloc((old_size + 1) * sizeof(int64_t))) == NULL) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		return -1;
 	}
-	if ((V = malloc((oldsize + 1) * sizeof(int64_t))) == NULL) {
+	if ((V = malloc((old_size + 1) * sizeof(int64_t))) == NULL) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(I);
 		return -1;
 	}
 
-	if (qsufsort(I, V, old_data, oldsize) != 0) {
+	if (qsufsort(I, V, old_data, old_size) != 0) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(I);
 		free(V);
 		return -1;
@@ -504,19 +502,19 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 	free(V);
 
 	if ((fd = open(new_filename, O_RDONLY, 0)) < 0) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(I);
 		return -1;
 	}
 
 	if (fstat(fd, &new_stat) != 0) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(I);
 		close(fd);
 		return -1;
 	}
 
-	newsize = new_stat.st_size;
+	new_size = new_stat.st_size;
 
 	/* Note: testing this to see how diffs between small files affect
 	 * updates. Small files seem to cause some problems between certain
@@ -526,76 +524,77 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 	 * the "is bsdiff < 90% of newfile size" check that would otherwise
 	 * be performed later on.
 	 */
-	if (newsize < 200) {
+	if (new_size < 200) {
 		memset(&small_header, 0, sizeof(struct header_v21));
 		memcpy(&small_header.magic, BSDIFF_HDR_FULLDL, 8);
 
-		efd = open(delta_filename, O_CREAT | O_EXCL | O_WRONLY, 00644);
+		efd = open(delta_filename_unique, O_CREAT | O_EXCL | O_WRONLY, 00644);
 		if (efd < 0) {
 			close(fd);
-			munmap(old_data, oldsize);
+			munmap(old_data, old_size);
 			free(I);
 			return -1;
 		}
 		if ((pf = fdopen(efd, "w")) == NULL) {
 			close(efd);
 			close(fd);
-			munmap(old_data, oldsize);
+			munmap(old_data, old_size);
 			free(I);
 			return -1;
 		}
 		if (fwrite(&small_header, 8, 1, pf) != 1) {
 			fclose(pf);
 			close(fd);
-			munmap(old_data, oldsize);
+			munmap(old_data, old_size);
 
 			free(I);
 			return -1;
 		}
 		fclose(pf);
 		close(fd);
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(I);
+		rename(delta_filename_unique, delta_filename);
 		return 1;
 	}
 
-	if ((new_data = malloc(newsize)) == NULL) {
+	if ((new_data = malloc(new_size)) == NULL) {
 		close(fd);
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(I);
 		return -1;
 	}
 
-	if (pread(fd, new_data, newsize, 0) != newsize) {
+	if (pread(fd, new_data, new_size, 0) != new_size) {
 		close(fd);
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(new_data);
 		free(I);
 		return -1;
 	}
 	if (close(fd) == -1) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(new_data);
 		free(I);
 		return -1;
 	}
 
 	/* we can write 3 8 byte tupples extra, so allocate some headroom */
-	if ((cb = malloc(newsize + 25)) == NULL) {
+	if ((cb = malloc(new_size + 25)) == NULL) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(new_data);
 		free(I);
 		return -1;
 	}
-	if ((db = malloc(newsize + 25)) == NULL) {
+	if ((db = malloc(new_size + 25)) == NULL) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(new_data);
 		free(cb);
 		free(I);
 		return -1;
 	}
-	if ((eb = malloc(newsize + 25)) == NULL) {
+	if ((eb = malloc(new_size + 25)) == NULL) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		free(new_data);
 		free(cb);
 		free(db);
@@ -607,108 +606,152 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 	eblen = 0;
 
 	/* Compute the differences */
-	scan = 0;
+	int64_t new_pos = 0;
-	len = 0;
+	int64_t old_pos = 0;
-	lastscan = 0;
+	int64_t match_len = 0;
-	lastpos = 0;
+	int64_t last_new_pos = 0;
-	lastoffset = 0;
+	int64_t last_old_pos = 0;
-	while (scan < newsize) {
+	int64_t last_offset = 0;
-		oldscore = 0;
+	while (new_pos < new_size) {
+		// Find an exact match between old and new files, and require
+		// that more than 8 of the matching bytes "mismatch" from the
+		// previous exact match. A score (old_score) is used to track
+		// how many bytes match starting from new_pos in new, and from
+		// old_pos in the previous iteration.
+		// NOTE: the magic value 8 is a heuristic; further testing is
+		// needed to prove whether this is the best number, or if the
+		// number should vary according to other factors, etc.
+		int64_t old_score = 0;
+		int64_t new_peek;
+		for (new_peek = new_pos += match_len; new_pos < new_size; new_pos++) {
+			search(I, old_data, old_size, new_data + new_pos, new_size - new_pos,
+			       0, old_size, &old_pos, &match_len);
 
-		for (scsc = scan += len; scan < newsize; scan++) {
+			for (; new_peek < new_pos + match_len; new_peek++) {
-			search(I, old_data, oldsize, new_data + scan, newsize - scan,
+				if ((new_peek + last_offset < old_size) &&
-			       0, oldsize, &pos, &len);
+				    (old_data[new_peek + last_offset] == new_data[new_peek])) {
-
+					old_score++;
-			for (; scsc < scan + len; scsc++) {
-				if ((scsc + lastoffset < oldsize) &&
-				    (old_data[scsc + lastoffset] == new_data[scsc])) {
-					oldscore++;
 				}
 			}
 
-			if (((len == oldscore) && (len != 0)) ||
+			if (((match_len == old_score) && (match_len != 0)) ||
-			    (len > oldscore + 8)) {
+			    (match_len > old_score + 8)) {
 				break;
 			}
 
-			if ((scan + lastoffset < oldsize) &&
+			// Before beginning the next loop iteration, decrement
-			    (old_data[scan + lastoffset] == new_data[scan])) {
+			// old_score if needed, since new_pos will be
-				oldscore--;
+			// incremented.
+			if ((new_pos + last_offset < old_size) &&
+			    (old_data[new_pos + last_offset] == new_data[new_pos])) {
+				old_score--;
 			}
 		}
 
-		if ((len != oldscore) || (scan == newsize)) {
+		if ((match_len != old_score) || (new_pos == new_size)) {
-			s = 0;
+			int64_t bytes = 0, max = 0;
-			Sf = 0;
+			// Compute the length of a fuzzy match starting from
-			lenf = 0;
+			// the beginning of the fuzzy match recorded at the end
-			for (i = 0;
+			// of the previous iteration (i.e. len_fuzzybackward
-			     (lastscan + i < scan) && (lastpos + i < oldsize);) {
+			// less than the previous match positions). At least
-				if (old_data[lastpos + i] == new_data[lastscan + i]) {
+			// half of the bytes match between old and new. This
-					s++;
+			// fuzzy match will be used to construct a diff string
+			// in the diff block.
+			// NOTE: "at least half matching bytes" is a heuristic
+			// for both fuzzy regions being constructed below;
+			// further testing is needed to prove whether this is
+			// the best percentage, or if the percentage should
+			// vary according to other factors, etc.
+			int64_t len_fuzzyforward = 0;
+			for (int64_t i = 0;
+			     (last_new_pos + i < new_pos) && (last_old_pos + i < old_size);) {
+				if (old_data[last_old_pos + i] == new_data[last_new_pos + i]) {
+					bytes++;
 				}
 				i++;
-				if (s * 2 - i > Sf * 2 - lenf) {
+				if (bytes * 2 - i > max * 2 - len_fuzzyforward) {
-					Sf = s;
+					max = bytes;
-					lenf = i;
+					len_fuzzyforward = i;
 				}
 			}
 
-			lenb = 0;
+			// Compute the length of a fuzzy match ending at the
-			if (scan < newsize) {
+			// current positions in old and new files (old_pos and
-				s = 0;
+			// new_pos). At least half of the bytes match between
-				Sb = 0;
+			// old and new. This fuzzy match will be used for the
-				for (i = 1;
+			// next iteration.
-				     (scan >= lastscan + i) && (pos >= i);
+			int64_t len_fuzzybackward = 0;
+			if (new_pos < new_size) {
+				bytes = 0;
+				max = 0;
+				for (int64_t i = 1;
+				     (new_pos >= last_new_pos + i) && (old_pos >= i);
 				     i++) {
-					if (old_data[pos - i] == new_data[scan - i]) {
+					if (old_data[old_pos - i] == new_data[new_pos - i]) {
-						s++;
+						bytes++;
 					}
-					if (s * 2 - i > Sb * 2 - lenb) {
+					if (bytes * 2 - i > max * 2 - len_fuzzybackward) {
-						Sb = s;
+						max = bytes;
-						lenb = i;
+						len_fuzzybackward = i;
 					}
 				}
 			}
 
-			if (lastscan + lenf > scan - lenb) {
+			// If there is an overlap between len_fuzzyforward and
-				overlap = (lastscan + lenf) - (scan - lenb);
+			// len_fuzzybackward in the new file, that overlap must
-				s = 0;
+			// be eliminated.
-				Ss = 0;
+			if (last_new_pos + len_fuzzyforward > new_pos - len_fuzzybackward) {
-				lens = 0;
+				bytes = 0;
-				for (i = 0; i < overlap; i++) {
+				max = 0;
-					if (new_data[lastscan + lenf - overlap + i] ==
+				int64_t overlap = (last_new_pos + len_fuzzyforward) - (new_pos - len_fuzzybackward);
-					    old_data[lastpos + lenf - overlap + i]) {
+				int64_t len_fuzzyshift = 0;
-						s++;
+				// Scan the overlap area for differences
+				// between old and new. If any mismatching
+				// bytes are found, extend len_fuzzyforward to
+				// cover those bytes, because we want them
+				// included in the diff block.
+				for (int64_t i = 0; i < overlap; i++) {
+					if (new_data[last_new_pos + len_fuzzyforward - overlap + i] ==
+					    old_data[last_old_pos + len_fuzzyforward - overlap + i]) {
+						bytes++;
 					}
-					if (new_data[scan - lenb + i] ==
+					if (new_data[new_pos - len_fuzzybackward + i] ==
-					    old_data[pos - lenb + i]) {
+					    old_data[old_pos - len_fuzzybackward + i]) {
-						s--;
+						bytes--;
 					}
-					if (s > Ss) {
+					if (bytes > max) {
-						Ss = s;
+						max = bytes;
-						lens = i + 1;
+						len_fuzzyshift = i + 1;
 					}
 				}
 
-				lenf += lens - overlap;
+				len_fuzzyforward += len_fuzzyshift - overlap;
-				lenb -= lens;
+				len_fuzzybackward -= len_fuzzyshift;
 			}
 
-			for (i = 0; i < lenf; i++) {
+			// Set the diff string in the diff block. For each byte
+			// in the fuzzy forward region, the byte from old is
+			// subtracted from new. When applying the delta (with
+			// bspatch) this operation is reversed, by performing
+			// additions.
+			for (int64_t i = 0; i < len_fuzzyforward; i++) {
 				db[dblen + i] =
-				    new_data[lastscan + i] - old_data[lastpos + i];
+				    new_data[last_new_pos + i] - old_data[last_old_pos + i];
 			}
-			for (i = 0; i < (scan - lenb) - (lastscan + lenf); i++) {
+			// Set the extra string in the extra block. The
-				eb[eblen + i] = new_data[lastscan + lenf + i];
+			// contents are the bytes in new file between the fuzzy
+			// forward and fuzzy backward regions.
+			for (int64_t i = 0; i < (new_pos - len_fuzzybackward) - (last_new_pos + len_fuzzyforward); i++) {
+				eb[eblen + i] = new_data[last_new_pos + len_fuzzyforward + i];
 			}
 
-			dblen += lenf;
+			dblen += len_fuzzyforward;
-			eblen += (scan - lenb) - (lastscan + lenf);
+			eblen += (new_pos - len_fuzzybackward) - (last_new_pos + len_fuzzyforward);
 
 			/* checking for control block overflow...
 			 * See regression test #15 for an example */
-			if ((int64_t)(cblen + 24) > (newsize + 25)) {
+			if ((int64_t)(cblen + 24) > (new_size + 25)) {
-				munmap(old_data, oldsize);
+				munmap(old_data, old_size);
 				free(new_data);
 				free(cb);
 				free(db);
@@ -717,18 +760,28 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 				return -1;
 			}
 
-			offtout(lenf, cb + cblen);
+			// Set three values in the control block:
+			//  1. ADD instruction (value: length of the diff
+			//     string). It uses the offset of the third control
+			//     block value from the previous iteration.
+			//  2. INSERT instruction (value: length of the extra
+			//     string)
+			//  3. offset in old file for the next ADD instruction
+			offtout(len_fuzzyforward, cb + cblen);
 			cblen += 8;
 
-			offtout((scan - lenb) - (lastscan + lenf), cb + cblen);
+			offtout((new_pos - len_fuzzybackward) - (last_new_pos + len_fuzzyforward), cb + cblen);
 			cblen += 8;
 
-			offtout((pos - lenb) - (lastpos + lenf), cb + cblen);
+			offtout((old_pos - len_fuzzybackward) - (last_old_pos + len_fuzzyforward), cb + cblen);
 			cblen += 8;
 
-			lastscan = scan - lenb;
+			// Save old/new file positions to the beginning of the
-			lastpos = pos - lenb;
+			// fuzzy backward region, since the next fuzzy forward
-			lastoffset = pos - scan;
+			// region will be calculated from that point.
+			last_new_pos = new_pos - len_fuzzybackward;
+			last_old_pos = old_pos - len_fuzzybackward;
+			last_offset = old_pos - new_pos;
 		}
 	}
 	free(I);
@@ -744,7 +797,7 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 
 	/* Create the patch file */
 
-	efd = open(delta_filename, O_CREAT | O_EXCL | O_WRONLY, 00644);
+	efd = open(delta_filename_unique, O_CREAT | O_EXCL | O_WRONLY, 00644);
 	if (efd < 0) {
 		ret = -1;
 		goto fulldl_free;
@@ -766,8 +819,8 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		small_header.control_length = cblen;
 		small_header.diff_length = dblen;
 		small_header.extra_length = eblen;
-		small_header.old_file_length = oldsize;
+		small_header.old_file_length = old_size;
-		small_header.new_file_length = newsize;
+		small_header.new_file_length = new_size;
 		small_header.file_mode = new_stat.st_mode;
 		small_header.file_owner = new_stat.st_uid;
 		small_header.file_group = new_stat.st_gid;
@@ -777,7 +830,7 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		eblock_set_enc(&small_header.encoding, e_enc);
 		encodings = small_header.encoding;
 
-		if ((first_block + cblen + dblen + eblen > 0.90 * newsize) && (enc != BSDIFF_ENC_NONE)) { /* tune */
+		if ((first_block + cblen + dblen + eblen > 0.90 * new_size) && (enc != BSDIFF_ENC_NONE)) { /* tune */
 			memcpy(&small_header.magic, BSDIFF_HDR_FULLDL, 8);
 			ret = 1;
 			if (fwrite(&small_header, 8, 1, pf) != 1) {
@@ -805,8 +858,8 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		large_header.control_length = cblen;
 		large_header.diff_length = dblen;
 		large_header.extra_length = eblen;
-		large_header.old_file_length = oldsize;
+		large_header.old_file_length = old_size;
-		large_header.new_file_length = newsize;
+		large_header.new_file_length = new_size;
 		large_header.file_mode = new_stat.st_mode;
 		large_header.file_owner = new_stat.st_uid;
 		large_header.file_group = new_stat.st_gid;
@@ -816,7 +869,7 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 		eblock_set_enc(&large_header.encoding, e_enc);
 		encodings = large_header.encoding;
 
-		if ((first_block + cblen + dblen + eblen > 0.90 * newsize) && (enc != BSDIFF_ENC_NONE)) { /* tune */
+		if ((first_block + cblen + dblen + eblen > 0.90 * new_size) && (enc != BSDIFF_ENC_NONE)) { /* tune */
 			memcpy(&large_header.magic, BSDIFF_HDR_FULLDL, 8);
 			ret = 1;
 			if (fwrite(&large_header, 8, 1, pf) != 1) {
@@ -847,7 +900,7 @@ int make_bsdiff_delta(char *old_filename, char *new_filename, char *delta_filena
 	}
 
 	bsdiff_files++;
-	bsdiff_newbytes += newsize;
+	bsdiff_newbytes += new_size;
 	bsdiff_outputbytes += first_block + cblen + dblen + eblen;
 
 	if (cblock_get_enc(encodings) == BSDIFF_ENC_NONE) {
@@ -899,9 +952,10 @@ fulldl_close_free:
 	if (fclose(pf)) {
 		ret = -1;
 	}
+	rename(delta_filename_unique, delta_filename);
 fulldl_free:
 	/* Free the memory we used */
-	munmap(old_data, oldsize);
+	munmap(old_data, old_size);
 	free(new_data);
 	free(cb);
 	free(db);

src/patch.c

@@ -254,7 +254,7 @@ static size_t xzread(xzfile *xzf, u_char *buf, size_t len, lzma_ret *err)
 
 typedef struct {
 	FILE *f; /* method = NONE, BZIP2, ZEROS */
-	int fd;  /* method = BZIP2 */
+	int fd;	 /* method = BZIP2 */
 	union {
 #ifdef BSDIFF_WITH_BZIP2
 		BZFILE *bz2; /* method = BZIP2 */
@@ -539,12 +539,12 @@ static int apply_delta_v2(int subver, FILE *f,
 	cfile cf, df, ef;
 	unsigned char *old_data = NULL, *new_data;
 	unsigned char buf[8];
-	off_t oldpos, newpos;
+	off_t old_pos, new_pos;
 	int64_t ctrl[3];
 	int i, ret, fd;
 	off_t data_offset;
 	off_t ctrllen, difflen, extralen;
-	off_t oldsize, newsize;
+	off_t old_size, new_size;
 	mode_t mode;
 	uid_t uid;
 	gid_t gid;
@@ -561,8 +561,8 @@ static int apply_delta_v2(int subver, FILE *f,
 		ctrllen = header.control_length;
 		difflen = header.diff_length;
 		extralen = header.extra_length;
-		oldsize = header.old_file_length;
+		old_size = header.old_file_length;
-		newsize = header.new_file_length;
+		new_size = header.new_file_length;
 		mode = header.file_mode;
 		uid = header.file_owner;
 		gid = header.file_group;
@@ -576,8 +576,8 @@ static int apply_delta_v2(int subver, FILE *f,
 		ctrllen = header.control_length;
 		difflen = header.diff_length;
 		extralen = header.extra_length;
-		oldsize = header.old_file_length;
+		old_size = header.old_file_length;
-		newsize = header.new_file_length;
+		new_size = header.new_file_length;
 		mode = header.file_mode;
 		uid = header.file_owner;
 		gid = header.file_group;
@@ -588,7 +588,7 @@ static int apply_delta_v2(int subver, FILE *f,
 
 	if ((ret = check_header(f, encoding,
 				ctrllen, difflen, extralen,
-				oldsize, newsize, data_offset)) < 0) {
+				old_size, new_size, data_offset)) < 0) {
 		return ret;
 	}
 
@@ -597,29 +597,29 @@ static int apply_delta_v2(int subver, FILE *f,
 		return ret;
 	}
 
-	ret = read_file(old_filename, &old_data, oldsize);
+	ret = read_file(old_filename, &old_data, old_size);
 	if (ret < 0) {
 		goto preperror;
 	}
 
-	if (newsize > BSDIFF_MAX_FILESZ) {
+	if (new_size > BSDIFF_MAX_FILESZ) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		ret = -1;
 		goto preperror;
 	}
 
-	/* Allocate newsize+1 bytes instead of newsize bytes to ensure
+	/* Allocate new_size+1 bytes instead of new_size bytes to ensure
 	   that we never try to malloc(0) and get a NULL pointer */
-	if ((new_data = malloc(newsize + 1)) == NULL) {
+	if ((new_data = malloc(new_size + 1)) == NULL) {
-		munmap(old_data, oldsize);
+		munmap(old_data, old_size);
 		ret = -1;
 		goto preperror;
 	}
-	memset(new_data, 0, newsize + 1);
+	memset(new_data, 0, new_size + 1);
 
-	oldpos = 0;
+	old_pos = 0;
-	newpos = 0;
+	new_pos = 0;
-	while (newpos < newsize) {
+	while (new_pos < new_size) {
 		/* Read control data:
 		 *   ctrl[0] == offset into diff block
 		 *   ctrl[1] == offset into extra block
@@ -628,7 +628,7 @@ static int apply_delta_v2(int subver, FILE *f,
 		 * The three control block words manage reads of the diff,
 		 * extra and old_data so that those three sources can be
 		 * combined into new_data.  ctrl[2] in particular may cause
-		 * oldpos to jump forward AND backward in order to allow
+		 * old_pos to jump forward AND backward in order to allow
 		 * copies of the original file content rather than using
 		 * diff or extra content.
 		 */
@@ -641,47 +641,47 @@ static int apply_delta_v2(int subver, FILE *f,
 		}
 
 		/* Sanity-check */
-		if (newpos + ctrl[0] > newsize || ctrl[0] < 0 || newpos + ctrl[0] < 0) {
+		if (new_pos + ctrl[0] > new_size || ctrl[0] < 0 || new_pos + ctrl[0] < 0) {
 			ret = -1;
 			goto readerror;
 		}
 
 		/* Read diff string */
-		ret = cfread(&df, new_data + newpos, ctrl[0], BSDIFF_BLOCK_DIFF, &d_zeros);
+		ret = cfread(&df, new_data + new_pos, ctrl[0], BSDIFF_BLOCK_DIFF, &d_zeros);
 		if (ret < 0) {
 			goto readerror;
 		}
 
 		/* Add old data to diff string */
 		for (i = 0; i < ctrl[0]; i++) {
-			if ((oldpos + i >= 0) && (oldpos + i < oldsize)) {
+			if ((old_pos + i >= 0) && (old_pos + i < old_size)) {
-				new_data[newpos + i] += old_data[oldpos + i];
+				new_data[new_pos + i] += old_data[old_pos + i];
 			}
 		}
 
 		/* Adjust pointers */
-		newpos += ctrl[0];
+		new_pos += ctrl[0];
-		oldpos += ctrl[0];
+		old_pos += ctrl[0];
 
 		/* Sanity-check */
-		if (newpos + ctrl[1] > newsize || ctrl[1] < 0 || newpos + ctrl[1] < 0) {
+		if (new_pos + ctrl[1] > new_size || ctrl[1] < 0 || new_pos + ctrl[1] < 0) {
 			ret = -1;
 			goto readerror;
 		}
-		if (oldpos + ctrl[2] > oldsize || oldpos + ctrl[2] < 0) {
+		if (old_pos + ctrl[2] > old_size || old_pos + ctrl[2] < 0) {
 			ret = -1;
 			goto readerror;
 		}
 
 		/* Read extra string */
-		ret = cfread(&ef, new_data + newpos, ctrl[1], BSDIFF_BLOCK_EXTRA, &e_zeros);
+		ret = cfread(&ef, new_data + new_pos, ctrl[1], BSDIFF_BLOCK_EXTRA, &e_zeros);
 		if (ret < 0) {
 			goto readerror;
 		}
 
 		/* Adjust pointers */
-		newpos += ctrl[1];
+		new_pos += ctrl[1];
-		oldpos += ctrl[2];
+		old_pos += ctrl[2];
 	}
 
 	/* Clean up the readers */
@@ -696,7 +696,7 @@ static int apply_delta_v2(int subver, FILE *f,
 		goto writeerror;
 	}
 
-	if (write(fd, new_data, newsize) != newsize) {
+	if (write(fd, new_data, new_size) != new_size) {
 		unlink(new_filename);
 		close(fd);
 		ret = -1;
@@ -719,12 +719,12 @@ static int apply_delta_v2(int subver, FILE *f,
 
 writeerror:
 	free(new_data);
-	munmap(old_data, oldsize);
+	munmap(old_data, old_size);
 	return ret;
 
 readerror:
 	free(new_data);
-	munmap(old_data, oldsize);
+	munmap(old_data, old_size);
 preperror:
 	cfclose(&cf);
 	cfclose(&df);

src/sufsort.c

@@ -126,8 +126,8 @@ static void split(int64_t *I, int64_t *V, int64_t arraylen, int64_t start, int64
 }
 
 /* The old_data (previous file data) is passed into this suffix sort and sorted
- * accordingly using the I and V arrays, which are both of length oldsize +1. */
+ * accordingly using the I and V arrays, which are both of length old_size +1. */
-int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t oldsize)
+int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t old_size)
 {
 	int64_t buckets[QSUF_BUCKET_SIZE];
 	int64_t i, h, len;
@@ -135,7 +135,7 @@ int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t oldsize)
 	for (i = 0; i < QSUF_BUCKET_SIZE; i++) {
 		buckets[i] = 0;
 	}
-	for (i = 0; i < oldsize; i++) {
+	for (i = 0; i < old_size; i++) {
 		buckets[old[i]]++;
 	}
 	for (i = 1; i < QSUF_BUCKET_SIZE; i++) {
@@ -146,17 +146,17 @@ int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t oldsize)
 	}
 	buckets[0] = 0;
 
-	for (i = 0; i < oldsize; i++) {
+	for (i = 0; i < old_size; i++) {
-		if (buckets[old[i]] > oldsize + 1) {
+		if (buckets[old[i]] > old_size + 1) {
 			return -1;
 		}
 		I[++buckets[old[i]]] = i;
 	}
 
-	for (i = 0; i < oldsize; i++) {
+	for (i = 0; i < old_size; i++) {
 		V[i] = buckets[old[i]];
 	}
-	V[oldsize] = 0;
+	V[old_size] = 0;
 	for (i = 1; i < QSUF_BUCKET_SIZE; i++) {
 		if (buckets[i] == buckets[i - 1] + 1) {
 			I[buckets[i]] = -1;
@@ -164,9 +164,9 @@ int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t oldsize)
 	}
 	I[0] = -1;
 
-	for (h = 1; I[0] != -(oldsize + 1); h += h) {
+	for (h = 1; I[0] != -(old_size + 1); h += h) {
 		len = 0;
-		for (i = 0; i < oldsize + 1;) {
+		for (i = 0; i < old_size + 1;) {
 			if (I[i] < 0) {
 				len -= I[i];
 				i -= I[i];
@@ -175,7 +175,7 @@ int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t oldsize)
 					I[i - len] = -len;
 				}
 				len = V[I[i]] + 1 - i;
-				split(I, V, oldsize, i, len, h);
+				split(I, V, old_size, i, len, h);
 				i += len;
 				len = 0;
 			}
@@ -185,7 +185,7 @@ int qsufsort(int64_t *I, int64_t *V, u_char *old, int64_t oldsize)
 		}
 	}
 
-	for (i = 0; i < oldsize + 1; i++) {
+	for (i = 0; i < old_size + 1; i++) {
 		I[V[i]] = i;
 	}