#include "quote.h"
#include "sha1-lookup.h"
#include "run-command.h"
+#include "log-tree.h"
#include "bisect.h"
-
-struct sha1_array {
- unsigned char (*sha1)[20];
- int sha1_nr;
- int sha1_alloc;
- int sorted;
-};
+#include "sha1-array.h"
+#include "argv-array.h"
static struct sha1_array good_revs;
static struct sha1_array skipped_revs;
static const unsigned char *current_bad_sha1;
-struct argv_array {
- const char **argv;
- int argv_nr;
- int argv_alloc;
-};
-
-static const char *argv_diff_tree[] = {"diff-tree", "--pretty", NULL, NULL};
static const char *argv_checkout[] = {"checkout", "-q", NULL, "--", NULL};
static const char *argv_show_branch[] = {"show-branch", NULL, NULL};
+static const char *argv_update_ref[] = {"update-ref", "--no-deref", "BISECT_HEAD", NULL, NULL};
/* bits #0-15 in revision.h */
enum object_type type;
unsigned long size;
char *buf = read_sha1_file(commit->object.sha1, &type, &size);
- char *ep, *sp;
+ const char *subject_start;
+ int subject_len;
fprintf(stderr, "%c%c%c ",
(flags & TREESAME) ? ' ' : 'T',
fprintf(stderr, " %.*s", 8,
sha1_to_hex(pp->item->object.sha1));
- sp = strstr(buf, "\n\n");
- if (sp) {
- sp += 2;
- for (ep = sp; *ep && *ep != '\n'; ep++)
- ;
- fprintf(stderr, " %.*s", (int)(ep - sp), sp);
- }
+ subject_len = find_commit_subject(buf, &subject_start);
+ if (subject_len)
+ fprintf(stderr, " %.*s", subject_len, subject_start);
fprintf(stderr, "\n");
}
}
return best;
}
-static void argv_array_push(struct argv_array *array, const char *string)
-{
- ALLOC_GROW(array->argv, array->argv_nr + 1, array->argv_alloc);
- array->argv[array->argv_nr++] = string;
-}
-
-static void argv_array_push_sha1(struct argv_array *array,
- const unsigned char *sha1,
- const char *format)
-{
- struct strbuf buf = STRBUF_INIT;
- strbuf_addf(&buf, format, sha1_to_hex(sha1));
- argv_array_push(array, strbuf_detach(&buf, NULL));
-}
-
-static void sha1_array_push(struct sha1_array *array,
- const unsigned char *sha1)
-{
- ALLOC_GROW(array->sha1, array->sha1_nr + 1, array->sha1_alloc);
- hashcpy(array->sha1[array->sha1_nr++], sha1);
-}
-
static int register_ref(const char *refname, const unsigned char *sha1,
int flags, void *cb_data)
{
if (!strcmp(refname, "bad")) {
current_bad_sha1 = sha1;
} else if (!prefixcmp(refname, "good-")) {
- sha1_array_push(&good_revs, sha1);
+ sha1_array_append(&good_revs, sha1);
} else if (!prefixcmp(refname, "skip-")) {
- sha1_array_push(&skipped_revs, sha1);
+ sha1_array_append(&skipped_revs, sha1);
}
return 0;
return for_each_ref_in("refs/bisect/", register_ref, NULL);
}
-void read_bisect_paths(struct argv_array *array)
+static void read_bisect_paths(struct argv_array *array)
{
struct strbuf str = STRBUF_INIT;
const char *filename = git_path("BISECT_NAMES");
FILE *fp = fopen(filename, "r");
if (!fp)
- die("Could not open file '%s': %s", filename, strerror(errno));
+ die_errno("Could not open file '%s'", filename);
while (strbuf_getline(&str, fp, '\n') != EOF) {
- char *quoted;
- int res;
-
strbuf_trim(&str);
- quoted = strbuf_detach(&str, NULL);
- res = sq_dequote_to_argv(quoted, &array->argv,
- &array->argv_nr, &array->argv_alloc);
- if (res)
+ if (sq_dequote_to_argv_array(str.buf, array))
die("Badly quoted content in file '%s': %s",
- filename, quoted);
+ filename, str.buf);
}
strbuf_release(&str);
fclose(fp);
}
-static int array_cmp(const void *a, const void *b)
-{
- return hashcmp(a, b);
-}
-
-static void sort_sha1_array(struct sha1_array *array)
-{
- qsort(array->sha1, array->sha1_nr, sizeof(*array->sha1), array_cmp);
-
- array->sorted = 1;
-}
-
-static const unsigned char *sha1_access(size_t index, void *table)
-{
- unsigned char (*array)[20] = table;
- return array[index];
-}
-
-static int lookup_sha1_array(struct sha1_array *array,
- const unsigned char *sha1)
-{
- if (!array->sorted)
- sort_sha1_array(array);
-
- return sha1_pos(sha1, array->sha1, array->sha1_nr, sha1_access);
-}
-
static char *join_sha1_array_hex(struct sha1_array *array, char delim)
{
struct strbuf joined_hexs = STRBUF_INIT;
int i;
- for (i = 0; i < array->sha1_nr; i++) {
+ for (i = 0; i < array->nr; i++) {
strbuf_addstr(&joined_hexs, sha1_to_hex(array->sha1[i]));
- if (i + 1 < array->sha1_nr)
+ if (i + 1 < array->nr)
strbuf_addch(&joined_hexs, delim);
}
return strbuf_detach(&joined_hexs, NULL);
}
+/*
+ * In this function, passing a not NULL skipped_first is very special.
+ * It means that we want to know if the first commit in the list is
+ * skipped because we will want to test a commit away from it if it is
+ * indeed skipped.
+ * So if the first commit is skipped, we cannot take the shortcut to
+ * just "return list" when we find the first non skipped commit, we
+ * have to return a fully filtered list.
+ *
+ * We use (*skipped_first == -1) to mean "it has been found that the
+ * first commit is not skipped". In this case *skipped_first is set back
+ * to 0 just before the function returns.
+ */
struct commit_list *filter_skipped(struct commit_list *list,
struct commit_list **tried,
- int show_all)
+ int show_all,
+ int *count,
+ int *skipped_first)
{
struct commit_list *filtered = NULL, **f = &filtered;
*tried = NULL;
- if (!skipped_revs.sha1_nr)
+ if (skipped_first)
+ *skipped_first = 0;
+ if (count)
+ *count = 0;
+
+ if (!skipped_revs.nr)
return list;
while (list) {
struct commit_list *next = list->next;
list->next = NULL;
- if (0 <= lookup_sha1_array(&skipped_revs,
+ if (0 <= sha1_array_lookup(&skipped_revs,
list->item->object.sha1)) {
+ if (skipped_first && !*skipped_first)
+ *skipped_first = 1;
/* Move current to tried list */
*tried = list;
tried = &list->next;
} else {
- if (!show_all)
- return list;
+ if (!show_all) {
+ if (!skipped_first || !*skipped_first)
+ return list;
+ } else if (skipped_first && !*skipped_first) {
+ /* This means we know it's not skipped */
+ *skipped_first = -1;
+ }
/* Move current to filtered list */
*f = list;
f = &list->next;
+ if (count)
+ (*count)++;
}
list = next;
}
+ if (skipped_first && *skipped_first == -1)
+ *skipped_first = 0;
+
return filtered;
}
+#define PRN_MODULO 32768
+
+/*
+ * This is a pseudo random number generator based on "man 3 rand".
+ * It is not used properly because the seed is the argument and it
+ * is increased by one between each call, but that should not matter
+ * for this application.
+ */
+static unsigned get_prn(unsigned count) {
+ count = count * 1103515245 + 12345;
+ return (count/65536) % PRN_MODULO;
+}
+
+/*
+ * Custom integer square root from
+ * http://en.wikipedia.org/wiki/Integer_square_root
+ */
+static int sqrti(int val)
+{
+ float d, x = val;
+
+ if (val == 0)
+ return 0;
+
+ do {
+ float y = (x + (float)val / x) / 2;
+ d = (y > x) ? y - x : x - y;
+ x = y;
+ } while (d >= 0.5);
+
+ return (int)x;
+}
+
+static struct commit_list *skip_away(struct commit_list *list, int count)
+{
+ struct commit_list *cur, *previous;
+ int prn, index, i;
+
+ prn = get_prn(count);
+ index = (count * prn / PRN_MODULO) * sqrti(prn) / sqrti(PRN_MODULO);
+
+ cur = list;
+ previous = NULL;
+
+ for (i = 0; cur; cur = cur->next, i++) {
+ if (i == index) {
+ if (hashcmp(cur->item->object.sha1, current_bad_sha1))
+ return cur;
+ if (previous)
+ return previous;
+ return list;
+ }
+ previous = cur;
+ }
+
+ return list;
+}
+
+static struct commit_list *managed_skipped(struct commit_list *list,
+ struct commit_list **tried)
+{
+ int count, skipped_first;
+
+ *tried = NULL;
+
+ if (!skipped_revs.nr)
+ return list;
+
+ list = filter_skipped(list, tried, 0, &count, &skipped_first);
+
+ if (!skipped_first)
+ return list;
+
+ return skip_away(list, count);
+}
+
static void bisect_rev_setup(struct rev_info *revs, const char *prefix,
const char *bad_format, const char *good_format,
int read_paths)
{
- struct argv_array rev_argv = { NULL, 0, 0 };
+ struct argv_array rev_argv = ARGV_ARRAY_INIT;
int i;
init_revisions(revs, prefix);
revs->commit_format = CMIT_FMT_UNSPECIFIED;
/* rev_argv.argv[0] will be ignored by setup_revisions */
- argv_array_push(&rev_argv, xstrdup("bisect_rev_setup"));
- argv_array_push_sha1(&rev_argv, current_bad_sha1, bad_format);
- for (i = 0; i < good_revs.sha1_nr; i++)
- argv_array_push_sha1(&rev_argv, good_revs.sha1[i],
- good_format);
- argv_array_push(&rev_argv, xstrdup("--"));
+ argv_array_push(&rev_argv, "bisect_rev_setup");
+ argv_array_pushf(&rev_argv, bad_format, sha1_to_hex(current_bad_sha1));
+ for (i = 0; i < good_revs.nr; i++)
+ argv_array_pushf(&rev_argv, good_format,
+ sha1_to_hex(good_revs.sha1[i]));
+ argv_array_push(&rev_argv, "--");
if (read_paths)
read_bisect_paths(&rev_argv);
- argv_array_push(&rev_argv, NULL);
- setup_revisions(rev_argv.argv_nr, rev_argv.argv, revs, NULL);
+ setup_revisions(rev_argv.argc, rev_argv.argv, revs, NULL);
+ /* XXX leak rev_argv, as "revs" may still be pointing to it */
}
static void bisect_common(struct rev_info *revs)
int fd = open(filename, O_CREAT | O_TRUNC | O_WRONLY, 0600);
if (fd < 0)
- die("could not create file '%s': %s",
- filename, strerror(errno));
+ die_errno("could not create file '%s'", filename);
bisect_rev_hex[len] = '\n';
write_or_die(fd, bisect_rev_hex, len + 1);
die("closing file %s: %s", filename, strerror(errno));
}
-static int bisect_checkout(char *bisect_rev_hex)
+static int bisect_checkout(char *bisect_rev_hex, int no_checkout)
{
int res;
mark_expected_rev(bisect_rev_hex);
argv_checkout[2] = bisect_rev_hex;
- res = run_command_v_opt(argv_checkout, RUN_GIT_CMD);
- if (res)
- exit(res);
+ if (no_checkout) {
+ argv_update_ref[3] = bisect_rev_hex;
+ if (run_command_v_opt(argv_update_ref, RUN_GIT_CMD))
+ die("update-ref --no-deref HEAD failed on %s",
+ bisect_rev_hex);
+ } else {
+ res = run_command_v_opt(argv_checkout, RUN_GIT_CMD);
+ if (res)
+ exit(res);
+ }
argv_show_branch[1] = bisect_rev_hex;
return run_command_v_opt(argv_show_branch, RUN_GIT_CMD);
static struct commit **get_bad_and_good_commits(int *rev_nr)
{
- int len = 1 + good_revs.sha1_nr;
+ int len = 1 + good_revs.nr;
struct commit **rev = xmalloc(len * sizeof(*rev));
int i, n = 0;
rev[n++] = get_commit_reference(current_bad_sha1);
- for (i = 0; i < good_revs.sha1_nr; i++)
+ for (i = 0; i < good_revs.nr; i++)
rev[n++] = get_commit_reference(good_revs.sha1[i]);
*rev_nr = n;
exit(1);
}
-void handle_skipped_merge_base(const unsigned char *mb)
+static void handle_skipped_merge_base(const unsigned char *mb)
{
char *mb_hex = sha1_to_hex(mb);
char *bad_hex = sha1_to_hex(current_bad_sha1);
char *good_hex = join_sha1_array_hex(&good_revs, ' ');
- fprintf(stderr, "Warning: the merge base between %s and [%s] "
+ warning("the merge base between %s and [%s] "
"must be skipped.\n"
"So we cannot be sure the first bad commit is "
"between %s and %s.\n"
- "We continue anyway.\n",
+ "We continue anyway.",
bad_hex, good_hex, mb_hex, bad_hex);
free(good_hex);
}
* - If one is "skipped", we can't know but we should warn.
* - If we don't know, we should check it out and ask the user to test.
*/
-static void check_merge_bases(void)
+static void check_merge_bases(int no_checkout)
{
struct commit_list *result;
int rev_nr;
const unsigned char *mb = result->item->object.sha1;
if (!hashcmp(mb, current_bad_sha1)) {
handle_bad_merge_base();
- } else if (0 <= lookup_sha1_array(&good_revs, mb)) {
+ } else if (0 <= sha1_array_lookup(&good_revs, mb)) {
continue;
- } else if (0 <= lookup_sha1_array(&skipped_revs, mb)) {
+ } else if (0 <= sha1_array_lookup(&skipped_revs, mb)) {
handle_skipped_merge_base(mb);
} else {
printf("Bisecting: a merge base must be tested\n");
- exit(bisect_checkout(sha1_to_hex(mb)));
+ exit(bisect_checkout(sha1_to_hex(mb), no_checkout));
}
}
free_commit_list(result);
}
-/*
- * This function runs the command "git rev-list $_good ^$_bad"
- * and returns 1 if it produces some output, 0 otherwise.
- */
-static int check_ancestors(void)
+static int check_ancestors(const char *prefix)
{
- struct argv_array rev_argv = { NULL, 0, 0 };
- struct strbuf str = STRBUF_INIT;
- int i, result = 0;
- struct child_process rls;
- FILE *rls_fout;
-
- argv_array_push(&rev_argv, xstrdup("rev-list"));
- argv_array_push_sha1(&rev_argv, current_bad_sha1, "^%s");
- for (i = 0; i < good_revs.sha1_nr; i++)
- argv_array_push_sha1(&rev_argv, good_revs.sha1[i], "%s");
- argv_array_push(&rev_argv, NULL);
-
- memset(&rls, 0, sizeof(rls));
- rls.argv = rev_argv.argv;
- rls.out = -1;
- rls.git_cmd = 1;
- if (start_command(&rls))
- die("Could not launch 'git rev-list' command.");
- rls_fout = fdopen(rls.out, "r");
- while (strbuf_getline(&str, rls_fout, '\n') != EOF) {
- strbuf_trim(&str);
- if (*str.buf) {
- result = 1;
- break;
- }
- }
- fclose(rls_fout);
- finish_command(&rls);
+ struct rev_info revs;
+ struct object_array pending_copy;
+ int res;
- return result;
+ bisect_rev_setup(&revs, prefix, "^%s", "%s", 0);
+
+ /* Save pending objects, so they can be cleaned up later. */
+ pending_copy = revs.pending;
+ revs.leak_pending = 1;
+
+ /*
+ * bisect_common calls prepare_revision_walk right away, which
+ * (together with .leak_pending = 1) makes us the sole owner of
+ * the list of pending objects.
+ */
+ bisect_common(&revs);
+ res = (revs.commits != NULL);
+
+ /* Clean up objects used, as they will be reused. */
+ clear_commit_marks_for_object_array(&pending_copy, ALL_REV_FLAGS);
+ free(pending_copy.objects);
+
+ return res;
}
/*
* If a merge base must be tested by the user, its source code will be
* checked out to be tested by the user and we will exit.
*/
-static void check_good_are_ancestors_of_bad(const char *prefix)
+static void check_good_are_ancestors_of_bad(const char *prefix, int no_checkout)
{
- const char *filename = git_path("BISECT_ANCESTORS_OK");
+ char *filename = git_pathdup("BISECT_ANCESTORS_OK");
struct stat st;
int fd;
/* Check if file BISECT_ANCESTORS_OK exists. */
if (!stat(filename, &st) && S_ISREG(st.st_mode))
- return;
+ goto done;
/* Bisecting with no good rev is ok. */
- if (good_revs.sha1_nr == 0)
- return;
+ if (good_revs.nr == 0)
+ goto done;
- if (check_ancestors())
- check_merge_bases();
+ /* Check if all good revs are ancestor of the bad rev. */
+ if (check_ancestors(prefix))
+ check_merge_bases(no_checkout);
/* Create file BISECT_ANCESTORS_OK. */
fd = open(filename, O_CREAT | O_TRUNC | O_WRONLY, 0600);
filename, strerror(errno));
else
close(fd);
+ done:
+ free(filename);
+}
+
+/*
+ * This does "git diff-tree --pretty COMMIT" without one fork+exec.
+ */
+static void show_diff_tree(const char *prefix, struct commit *commit)
+{
+ struct rev_info opt;
+
+ /* diff-tree init */
+ init_revisions(&opt, prefix);
+ git_config(git_diff_basic_config, NULL); /* no "diff" UI options */
+ opt.abbrev = 0;
+ opt.diff = 1;
+
+ /* This is what "--pretty" does */
+ opt.verbose_header = 1;
+ opt.use_terminator = 0;
+ opt.commit_format = CMIT_FMT_DEFAULT;
+
+ /* diff-tree init */
+ if (!opt.diffopt.output_format)
+ opt.diffopt.output_format = DIFF_FORMAT_RAW;
+
+ log_tree_commit(&opt, commit);
}
/*
* We use the convention that exiting with an exit code 10 means that
* the bisection process finished successfully.
* In this case the calling shell script should exit 0.
+ *
+ * If no_checkout is non-zero, the bisection process does not
+ * checkout the trial commit but instead simply updates BISECT_HEAD.
*/
-int bisect_next_all(const char *prefix)
+int bisect_next_all(const char *prefix, int no_checkout)
{
struct rev_info revs;
struct commit_list *tried;
- int reaches = 0, all = 0, nr;
+ int reaches = 0, all = 0, nr, steps;
const unsigned char *bisect_rev;
char bisect_rev_hex[41];
if (read_bisect_refs())
die("reading bisect refs failed");
- check_good_are_ancestors_of_bad(prefix);
+ check_good_are_ancestors_of_bad(prefix, no_checkout);
bisect_rev_setup(&revs, prefix, "%s", "^%s", 1);
revs.limited = 1;
bisect_common(&revs);
revs.commits = find_bisection(revs.commits, &reaches, &all,
- !!skipped_revs.sha1_nr);
- revs.commits = filter_skipped(revs.commits, &tried, 0);
+ !!skipped_revs.nr);
+ revs.commits = managed_skipped(revs.commits, &tried);
if (!revs.commits) {
/*
exit(1);
}
+ if (!all) {
+ fprintf(stderr, "No testable commit found.\n"
+ "Maybe you started with bad path parameters?\n");
+ exit(4);
+ }
+
bisect_rev = revs.commits->item->object.sha1;
memcpy(bisect_rev_hex, sha1_to_hex(bisect_rev), 41);
if (!hashcmp(bisect_rev, current_bad_sha1)) {
exit_if_skipped_commits(tried, current_bad_sha1);
- printf("%s is first bad commit\n", bisect_rev_hex);
- argv_diff_tree[2] = bisect_rev_hex;
- run_command_v_opt(argv_diff_tree, RUN_GIT_CMD);
+ printf("%s is the first bad commit\n", bisect_rev_hex);
+ show_diff_tree(prefix, revs.commits->item);
/* This means the bisection process succeeded. */
exit(10);
}
nr = all - reaches - 1;
- printf("Bisecting: %d revisions left to test after this "
- "(roughly %d steps)\n", nr, estimate_bisect_steps(all));
+ steps = estimate_bisect_steps(all);
+ printf("Bisecting: %d revision%s left to test after this "
+ "(roughly %d step%s)\n", nr, (nr == 1 ? "" : "s"),
+ steps, (steps == 1 ? "" : "s"));
+
+ return bisect_checkout(bisect_rev_hex, no_checkout);
+}
+
+static inline int log2i(int n)
+{
+ int log2 = 0;
- return bisect_checkout(bisect_rev_hex);
+ for (; n > 1; n >>= 1)
+ log2++;
+
+ return log2;
+}
+
+static inline int exp2i(int n)
+{
+ return 1 << n;
}
+/*
+ * Estimate the number of bisect steps left (after the current step)
+ *
+ * For any x between 0 included and 2^n excluded, the probability for
+ * n - 1 steps left looks like:
+ *
+ * P(2^n + x) == (2^n - x) / (2^n + x)
+ *
+ * and P(2^n + x) < 0.5 means 2^n < 3x
+ */
+int estimate_bisect_steps(int all)
+{
+ int n, x, e;
+
+ if (all < 3)
+ return 0;
+
+ n = log2i(all);
+ e = exp2i(n);
+ x = all - e;
+
+ return (e < 3 * x) ? n : n - 1;
+}