@matheusbsbs / matheus-bagliano / commits / 19716b21a4

cleanup: fix possible overflow errors in binary search

A common mistake when writing binary search is to allow possible integer overflow by using the simple average: mid = (min + max) / 2; Instead, use the overflow-safe version: mid = min + (max - min) / 2; This translation is safe since the operation occurs inside a loop conditioned on "min < max". The included changes were found using the following git grep: git grep '/ *2;' '*.c' Making this cleanup will prevent future review friction when a new binary search is contructed based on existing code. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Derrick Stolee committed Oct 8, 2017 at 14:29 UTC 19716b21a4255ecc7148b54ab2c78039c59f25bf

12 files changed +15 -15

builtin/index-pack.c

+2 -2

@@ -633,7 +633,7 @@ static int find_ofs_delta(const off_t offset, enum object_type type)

633 int first = 0, last = nr_ofs_deltas;

634

635 while (first < last) {

636 - int next = (first + last) / 2;

636 + int next = first + (last - first) / 2;

637 struct ofs_delta_entry *delta = &ofs_deltas[next];

638 int cmp;

639

@@ -687,7 +687,7 @@ static int find_ref_delta(const unsigned char *sha1, enum object_type type)

687 int first = 0, last = nr_ref_deltas;

688

689 while (first < last) {

690 - int next = (first + last) / 2;

690 + int next = first + (last - first) / 2;

691 struct ref_delta_entry *delta = &ref_deltas[next];

692 int cmp;

693

builtin/pack-objects.c

+1 -1

@@ -1277,7 +1277,7 @@ static int done_pbase_path_pos(unsigned hash)

1277 int lo = 0;

1278 int hi = done_pbase_paths_num;

1279 while (lo < hi) {

1280 - int mi = (hi + lo) / 2;

1280 + int mi = lo + (hi - lo) / 2;

1281 if (done_pbase_paths[mi] == hash)

1282 return mi;

1283 if (done_pbase_paths[mi] < hash)

builtin/unpack-objects.c

+1 -1

@@ -394,7 +394,7 @@ static void unpack_delta_entry(enum object_type type, unsigned long delta_size,

394 lo = 0;

395 hi = nr;

396 while (lo < hi) {

397 - mid = (lo + hi)/2;

397 + mid = lo + (hi - lo) / 2;

398 if (base_offset < obj_list[mid].offset) {

399 hi = mid;

400 } else if (base_offset > obj_list[mid].offset) {

cache-tree.c

+1 -1

@@ -49,7 +49,7 @@ static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)

49 lo = 0;

50 hi = it->subtree_nr;

51 while (lo < hi) {

52 - int mi = (lo + hi) / 2;

52 + int mi = lo + (hi - lo) / 2;

53 struct cache_tree_sub *mdl = down[mi];

54 int cmp = subtree_name_cmp(path, pathlen,

55 mdl->name, mdl->namelen);

compat/regex/regex_internal.c

+2 -2

@@ -613,7 +613,7 @@ re_string_reconstruct (re_string_t *pstr, int idx, int eflags)

613 int low = 0, high = pstr->valid_len, mid;

614 do

615 {

616 - mid = (high + low) / 2;

616 + mid = low + (high - low) / 2;

617 if (pstr->offsets[mid] > offset)

618 high = mid;

619 else if (pstr->offsets[mid] < offset)

@@ -1394,7 +1394,7 @@ re_node_set_contains (const re_node_set *set, int elem)

1394 right = set->nelem - 1;

1395 while (idx < right)

1396 {

1397 - mid = (idx + right) / 2;

1397 + mid = idx + (right - idx) / 2;

1398 if (set->elems[mid] < elem)

1399 idx = mid + 1;

1400 else

compat/regex/regexec.c

+1 -1

@@ -4284,7 +4284,7 @@ search_cur_bkref_entry (const re_match_context_t *mctx, int str_idx)

4284 last = right = mctx->nbkref_ents;

4285 for (left = 0; left < right;)

4286 {

4287 - mid = (left + right) / 2;

4287 + mid = left + (right - left) / 2;

4288 if (mctx->bkref_ents[mid].str_idx < str_idx)

4289 left = mid + 1;

4290 else

packfile.c

+1 -1

@@ -1743,7 +1743,7 @@ off_t find_pack_entry_one(const unsigned char *sha1,

1743 sha1[0], sha1[1], sha1[2], lo, hi, p->num_objects);

1744

1745 while (lo < hi) {

1746 - unsigned mi = (lo + hi) / 2;

1746 + unsigned mi = lo + (hi - lo) / 2;

1747 int cmp = hashcmp(index + mi * stride, sha1);

1748

1749 if (debug_lookup)

sha1-lookup.c

+2 -2

@@ -10,7 +10,7 @@ static uint32_t take2(const unsigned char *sha1)

10 * Conventional binary search loop looks like this:

11 *

12 * do {

13 - * int mi = (lo + hi) / 2;

13 + * int mi = lo + (hi - lo) / 2;

14 * int cmp = "entry pointed at by mi" minus "target";

15 * if (!cmp)

16 * return (mi is the wanted one)

@@ -95,7 +95,7 @@ int sha1_pos(const unsigned char *sha1, void *table, size_t nr,

95 hi = mi;

96 else

97 lo = mi + 1;

98 - mi = (hi + lo) / 2;

98 + mi = lo + (hi - lo) / 2;

99 } while (lo < hi);

100 return -lo-1;

101 }

sha1_name.c

+1 -1

@@ -157,7 +157,7 @@ static void unique_in_pack(struct packed_git *p,

157 num = p->num_objects;

158 last = num;

159 while (first < last) {

160 - uint32_t mid = (first + last) / 2;

160 + uint32_t mid = first + (last - first) / 2;

161 const unsigned char *current;

162 int cmp;

163

string-list.c

+1 -1

@@ -16,7 +16,7 @@ static int get_entry_index(const struct string_list *list, const char *string,

16 compare_strings_fn cmp = list->cmp ? list->cmp : strcmp;

18 while (left + 1 < right) {

19 - int middle = (left + right) / 2;

19 + int middle = left + (right - left) / 2;

20 int compare = cmp(string, list->items[middle].string);

21 if (compare < 0)

22 right = middle;

utf8.c

+1 -1

@@ -32,7 +32,7 @@ static int bisearch(ucs_char_t ucs, const struct interval *table, int max)

32 if (ucs < table[0].first || ucs > table[max].last)

33 return 0;

34 while (max >= min) {

35 - mid = (min + max) / 2;

35 + mid = min + (max - min) / 2;

36 if (ucs > table[mid].last)

37 min = mid + 1;

38 else if (ucs < table[mid].first)

xdiff/xpatience.c

+1 -1

@@ -166,7 +166,7 @@ static int binary_search(struct entry **sequence, int longest,

166 int left = -1, right = longest;

167

168 while (left + 1 < right) {

169 - int middle = (left + right) / 2;

169 + int middle = left + (right - left) / 2;

170 /* by construction, no two entries can be equal */

171 if (sequence[middle]->line2 > entry->line2)

172 right = middle;