ignores *init_ignore(ignores *parent, const char *dirname, const size_t dirname_len) {
ignores *ig = ag_malloc(sizeof(ignores));
ig->names = NULL;
ig->names_len = 0;
ig->slash_names = NULL;
ig->slash_names_len = 0;
ig->regexes = NULL;
ig->regexes_len = 0;
ig->slash_regexes = NULL;
ig->slash_regexes_len = 0;
ig->dirname = dirname;
ig->dirname_len = dirname_len;
if (parent && is_empty(parent) && parent->parent) {
ig->parent = parent->parent;
} else {
ig->parent = parent;
}
if (parent && parent->abs_path_len > 0) {
ag_asprintf(&(ig->abs_path), "%s/%s", parent->abs_path, dirname);
ig->abs_path_len = parent->abs_path_len + 1 + dirname_len;
} else if (dirname_len == 1 && dirname[0] == '.') {
ig->abs_path = ag_malloc(sizeof(char));
ig->abs_path[0] = '\0';
ig->abs_path_len = 0;
} else {
ag_asprintf(&(ig->abs_path), "%s", dirname);
ig->abs_path_len = dirname_len;
}
return ig;
}
char *make_lang_regex(const char **extensions) {
int regex_capacity = 100;
char *regex = ag_malloc(regex_capacity);
int regex_length = 3;
int subsequent = 0;
const char **extension;
strcpy(regex, "\\.(");
for (extension = extensions; *extension; ++extension) {
int extension_length = strlen(*extension);
while (regex_length + extension_length + 3 + subsequent > regex_capacity) {
regex_capacity *= 2;
regex = ag_realloc(regex, regex_capacity);
}
if (subsequent) {
regex[regex_length++] = '|';
} else {
subsequent = 1;
}
strcpy(regex + regex_length, *extension);
regex_length += extension_length;
}
regex[regex_length++] = ')';
regex[regex_length++] = '$';
regex[regex_length++] = 0;
return regex;
}
char *make_lang_regex(char *ext_array, size_t num_exts) {
int regex_capacity = 100;
char *regex = ag_malloc(regex_capacity);
int regex_length = 3;
int subsequent = 0;
char *extension;
size_t i;
strcpy(regex, "\\.(");
for (i = 0; i < num_exts; ++i) {
extension = ext_array + i * SINGLE_EXT_LEN;
int extension_length = strlen(extension);
while (regex_length + extension_length + 3 + subsequent > regex_capacity) {
regex_capacity *= 2;
regex = ag_realloc(regex, regex_capacity);
}
if (subsequent) {
regex[regex_length++] = '|';
} else {
subsequent = 1;
}
strcpy(regex + regex_length, extension);
regex_length += extension_length;
}
regex[regex_length++] = ')';
regex[regex_length++] = '$';
regex[regex_length++] = 0;
return regex;
}
size_t combine_file_extensions(size_t *extension_index, size_t len, char **exts) {
/* Keep it fixed as 100 for the reason that if you have more than 100
* file types to search, you'd better search all the files.
* */
size_t ext_capacity = 100;
(*exts) = (char *)ag_malloc(ext_capacity * SINGLE_EXT_LEN);
memset((*exts), 0, ext_capacity * SINGLE_EXT_LEN);
size_t num_of_extensions = 0;
size_t i;
for (i = 0; i < len; ++i) {
size_t j = 0;
const char *ext = langs[extension_index[i]].extensions[j];
do {
if (num_of_extensions == ext_capacity) {
break;
}
char *pos = (*exts) + num_of_extensions * SINGLE_EXT_LEN;
strncpy(pos, ext, strlen(ext));
++num_of_extensions;
ext = langs[extension_index[i]].extensions[++j];
} while (ext);
}
return num_of_extensions;
}
static int check_symloop_enter(const char *path, dirkey_t *outkey) {
#ifdef _WIN32
return SYMLOOP_OK;
#else
struct stat buf;
symdir_t *item_found = NULL;
symdir_t *new_item = NULL;
memset(outkey, 0, sizeof(dirkey_t));
outkey->dev = 0;
outkey->ino = 0;
int res = stat(path, &buf);
if (res != 0) {
log_err("Error stat()ing: %s", path);
return SYMLOOP_ERROR;
}
outkey->dev = buf.st_dev;
outkey->ino = buf.st_ino;
HASH_FIND(hh, symhash, outkey, sizeof(dirkey_t), item_found);
if (item_found) {
return SYMLOOP_LOOP;
}
new_item = (symdir_t*)ag_malloc(sizeof(symdir_t));
memcpy(&new_item->key, outkey, sizeof(dirkey_t));
HASH_ADD(hh, symhash, key, sizeof(dirkey_t), new_item);
return SYMLOOP_OK;
#endif
}
ignores *init_ignore(ignores *parent) {
ignores *ig = ag_malloc(sizeof(ignores));
ig->names = NULL;
ig->names_len = 0;
ig->regexes = NULL;
ig->regexes_len = 0;
ig->parent = parent;
return ig;
}
/* BTW, in case you're wondering, this returns void* so that
* it can be used as a pthread...
*/
static void *default_start_server(void *arg)
{
server_t self = arg;
connection_t con;
int i;
if (!self) return NULL;
/*
* TODO.
* This shouldn't be done here. It doesn't really matter
* because it is basically a no-op right now anyway.
* A single process can have several servers running
* in it, conceivably.
*
* Maybe it should be moved to ag.c and left up to main.
*/
become_daemon(self);
/* configure the wait_queue's max size */
self->wq->max = self->maxcon;
/* this is the connection handler: */
self->handlers = ag_malloc(sizeof(handler_rec) * self->numthreads);
for (i = 0; i < self->numthreads; i++) {
handler_t hd;
hd = self->newhandler();
hd->svr = self;
self->handlers[i].hd = hd;
(void) pthread_create(&self->handlers[i].tid, /* thread id */
NULL, /* attributes */
hd->start, /* function to run */
(void *)hd /* arg to pass to function */
);
}
/*
* must do this before changing uids, because we may want to bind
* to a port < 1024 and then become a non-root user.
*
* TODO
* At this point we should also open up the logfile.
*/
self->listen_sk = setup_connection(self);
/* only change users if self->uid is non-zero (non-root) */
if (self->uid) become_user(self);
for (;;) {
con = accept_connection(self->listen_sk);
if (!self->wq->add_con(self->wq, con)) con->destroy(con);
}
}
queue_t new_queue()
{
queue_t q;
q = ag_malloc(queue_size);
q->push = default_push;
q->pop = default_pop;
q->destroy = default_destroy_q;
return q;
}
char *ag_strndup(const char *s, size_t size) {
char *str = NULL;
#ifdef HAVE_STRNDUP
str = strndup(s, size);
CHECK_AND_RETURN(str)
#else
str = (char *)ag_malloc(size + 1);
strlcpy(str, s, size + 1);
return str;
#endif
}
server_t new_server(server_t svr)
{
if (!svr) svr = ag_malloc(svr_size);
svr->wq = new_wait_queue();
svr->destroy = default_destroy_svr;
svr->summary = default_summary;
svr->start = default_start_server;
svr->newhandler = NULL;
svr->pidfile = svr->logfile = svr->name = "";
return svr;
}
void generate_find_skip(const char *find, const size_t f_len, size_t **skip_lookup, const int case_sensitive) {
size_t i;
size_t s_len;
size_t *sl = (size_t *)ag_malloc(f_len * sizeof(size_t));
*skip_lookup = sl;
size_t last_prefix = f_len;
for (i = last_prefix; i > 0; i--) {
if (is_prefix(find, f_len, i, case_sensitive)) {
last_prefix = i;
}
sl[i - 1] = last_prefix + (f_len - i);
}
for (i = 0; i < f_len; i++) {
s_len = suffix_len(find, f_len, i, case_sensitive);
if (find[i - s_len] != find[f_len - 1 - s_len]) {
sl[f_len - 1 - s_len] = f_len - 1 - i + s_len;
}
}
}
static connection_t accept_connection(int lsock)
{
int sock;
struct sockaddr_in *name;
int len;
connection_t con;
name = ag_malloc(sizeof(struct sockaddr_in));
if ((sock = accept(lsock, name, &len)) <= 0) {
fprintf(stderr, "accept_connection: error\n");
}
if (len > sizeof(struct sockaddr_in)) {
fprintf(stderr, "accept_connection: memory corruption.\n");
exit(1);
}
con = new_connection();
con->sk = sock;
con->name = name;
return con;
}
void search_buf(char *buf, size_t buf_len,
const char *dir_full_path, char *tmp_file_path) {
int binary = -1; /* 1 = yes, 0 = no, -1 = don't know */
size_t buf_offset = 0;
//if (opts.search_stream) {
// binary = 0;
//}
//else if (!opts.search_binary_files) {
// binary = is_binary(buf, buf_len);
// if (binary) {
// log_debug("File %s is binary. Skipping...", dir_full_path);
// if (!convert_to_text(&buf, &buf_len, dir_full_path, tmp_file_path))
// return;
// }
//}
binary = is_binary(buf, buf_len);
if (binary) {
log_debug("File %s is binary. Convert to text...", dir_full_path);
if (!convert_to_text(&buf, &buf_len, dir_full_path, tmp_file_path))
return;
}
size_t matches_len = 0;
match_t *matches;
size_t matches_size;
size_t matches_spare;
//if (opts.invert_match) {
// /* If we are going to invert the set of matches at the end, we will need
// * one extra match struct, even if there are no matches at all. So make
// * sure we have a nonempty array; and make sure we always have spare
// * capacity for one extra.
// */
// matches_size = 100;
// matches = (match_t *)ag_malloc(matches_size * sizeof(match_t));
// matches_spare = 1;
//}
//else {
// matches_size = 0;
// matches = NULL;
// matches_spare = 0;
//}
matches_size = 0;
matches = NULL;
matches_spare = 0;
if (/*!opts.literal && */opts.query_len == 1 && opts.query[0] == '.') {
matches_size = 1;
matches = (match_t *)ag_malloc(matches_size * sizeof(match_t));
matches[0].start = 0;
matches[0].end = buf_len;
matches_len = 1;
}
else {
const char *match_ptr = buf;
strncmp_fp ag_strnstr_fp = get_strstr(opts.casing);
while (buf_offset < buf_len) {
match_ptr = ag_strnstr_fp(match_ptr, opts.query, buf_len - buf_offset, opts.query_len, alpha_skip_lookup, find_skip_lookup);
if (match_ptr == NULL) {
break;
}
if (opts.word_regexp) {
const char *start = match_ptr;
const char *end = match_ptr + opts.query_len;
/* Check whether both start and end of the match lie on a word
* boundary
*/
if ((start == buf ||
is_wordchar(*(start - 1)) != opts.literal_starts_wordchar) &&
(end == buf + buf_len ||
is_wordchar(*end) != opts.literal_ends_wordchar)) {
/* It's a match */
}
else {
/* It's not a match */
match_ptr += opts.query_len;
buf_offset = end - buf;
continue;
}
}
if (matches_len + matches_spare >= matches_size) {
/* TODO: benchmark initial size of matches. 100 may be too small/big */
matches_size = matches ? matches_size * 2 : 100;
log_debug("Too many matches in %s. Reallocating matches to %zu.", dir_full_path, matches_size);
matches = (match_t *)ag_realloc(matches, matches_size * sizeof(match_t));
}
matches[matches_len].start = match_ptr - buf;
matches[matches_len].end = matches[matches_len].start + opts.query_len;
buf_offset = matches[matches_len].end;
log_debug("Match found. File %s, offset %lu bytes.", dir_full_path, matches[matches_len].start);
matches_len++;
match_ptr += opts.query_len;
if (opts.max_matches_per_file > 0 && matches_len >= opts.max_matches_per_file) {
log_err("Too many matches in %s. Skipping the rest of this file.", dir_full_path);
break;
}
}
}
else {
void search_buf(const char *buf, const size_t buf_len,
const char *dir_full_path) {
int binary = -1; /* 1 = yes, 0 = no, -1 = don't know */
size_t buf_offset = 0;
if (opts.search_stream) {
binary = 0;
} else if (!opts.search_binary_files) {
binary = is_binary((const void *)buf, buf_len);
if (binary) {
log_debug("File %s is binary. Skipping...", dir_full_path);
return;
}
}
int matches_len = 0;
match *matches;
size_t matches_size;
size_t matches_spare;
if (opts.invert_match) {
/* If we are going to invert the set of matches at the end, we will need
* one extra match struct, even if there are no matches at all. So make
* sure we have a nonempty array; and make sure we always have spare
* capacity for one extra.
*/
matches_size = 100;
matches = ag_malloc(matches_size * sizeof(match));
matches_spare = 1;
} else {
matches_size = 0;
matches = NULL;
matches_spare = 0;
}
if (opts.query_len == 1 && opts.query[0] == '.') {
matches_size = 1;
matches = ag_malloc(matches_size * sizeof(match));
matches[0].start = 0;
matches[0].end = buf_len;
matches_len = 1;
} else if (opts.literal) {
const char *match_ptr = buf;
strncmp_fp ag_strnstr_fp = get_strstr(opts.casing);
while (buf_offset < buf_len) {
match_ptr = ag_strnstr_fp(match_ptr, opts.query, buf_len - buf_offset, opts.query_len, skip_lookup);
if (match_ptr == NULL) {
break;
}
if (opts.word_regexp) {
const char *start = match_ptr;
const char *end = match_ptr + opts.query_len;
/* Check whether both start and end of the match lie on a word
* boundary
*/
if ((start == buf ||
is_wordchar(*(start - 1)) != opts.literal_starts_wordchar) &&
(end == buf + buf_len ||
is_wordchar(*end) != opts.literal_ends_wordchar)) {
/* It's a match */
} else {
/* It's not a match */
match_ptr += opts.query_len;
buf_offset = end - buf;
continue;
}
}
if ((size_t)matches_len + matches_spare >= matches_size) {
/* TODO: benchmark initial size of matches. 100 may be too small/big */
matches_size = matches ? matches_size * 2 : 100;
log_debug("Too many matches in %s. Reallocating matches to %zu.", dir_full_path, matches_size);
matches = ag_realloc(matches, matches_size * sizeof(match));
}
matches[matches_len].start = match_ptr - buf;
matches[matches_len].end = matches[matches_len].start + opts.query_len;
buf_offset = matches[matches_len].end;
log_debug("Match found. File %s, offset %lu bytes.", dir_full_path, matches[matches_len].start);
matches_len++;
match_ptr += opts.query_len;
if (matches_len >= opts.max_matches_per_file) {
log_err("Too many matches in %s. Skipping the rest of this file.", dir_full_path);
break;
}
}
} else {
int offset_vector[3];
while (buf_offset < buf_len &&
(pcre_exec(opts.re, opts.re_extra, buf, buf_len, buf_offset, 0, offset_vector, 3)) >= 0) {
log_debug("Regex match found. File %s, offset %i bytes.", dir_full_path, offset_vector[0]);
buf_offset = offset_vector[1];
/* TODO: copy-pasted from above. FIXME */
if ((size_t)matches_len + matches_spare >= matches_size) {
matches_size = matches ? matches_size * 2 : 100;
log_debug("Too many matches in %s. Reallocating matches to %zu.", dir_full_path, matches_size);
matches = ag_realloc(matches, matches_size * sizeof(match));
}
matches[matches_len].start = offset_vector[0];
matches[matches_len].end = offset_vector[1];
matches_len++;
if (matches_len >= opts.max_matches_per_file) {
log_err("Too many matches in %s. Skipping the rest of this file.", dir_full_path);
break;
}
}
}
if (opts.invert_match) {
matches_len = invert_matches(buf, buf_len, matches, matches_len);
}
if (opts.stats) {
pthread_mutex_lock(&stats_mtx);
stats.total_bytes += buf_len;
stats.total_files++;
stats.total_matches += matches_len;
pthread_mutex_unlock(&stats_mtx);
}
if (matches_len > 0) {
if (binary == -1 && !opts.print_filename_only) {
binary = is_binary((const void *)buf, buf_len);
}
pthread_mutex_lock(&print_mtx);
if (opts.print_filename_only) {
/* If the --files-without-matches or -L option in passed we should
* not print a matching line. This option currently sets
* opts.print_filename_only and opts.invert_match. Unfortunately
* setting the latter has the side effect of making matches.len = 1
* on a file-without-matches which is not desired behaviour. See
* GitHub issue 206 for the consequences if this behaviour is not
* checked. */
if (!opts.invert_match || matches_len < 2) {
print_path(dir_full_path, '\n');
}
} else if (binary) {
print_binary_file_matches(dir_full_path);
} else {
print_file_matches(dir_full_path, buf, buf_len, matches, matches_len);
}
pthread_mutex_unlock(&print_mtx);
} else {
log_debug("No match in %s", dir_full_path);
}
if (matches_size > 0) {
free(matches);
}
}
void load_svn_ignore_patterns(ignores *ig, const char *path) {
FILE *fp = NULL;
char *dir_prop_base;
ag_asprintf(&dir_prop_base, "%s/%s", path, SVN_DIR_PROP_BASE);
fp = fopen(dir_prop_base, "r");
if (fp == NULL) {
log_debug("Skipping svn ignore file %s", dir_prop_base);
free(dir_prop_base);
return;
}
char *entry = NULL;
size_t entry_len = 0;
char *key = ag_malloc(32); /* Sane start for max key length. */
size_t key_len = 0;
size_t bytes_read = 0;
char *entry_line;
size_t line_len;
int matches;
while (fscanf(fp, "K %zu\n", &key_len) == 1) {
key = ag_realloc(key, key_len + 1);
bytes_read = fread(key, 1, key_len, fp);
key[key_len] = '\0';
matches = fscanf(fp, "\nV %zu\n", &entry_len);
if (matches != 1) {
log_debug("Unable to parse svnignore file %s: fscanf() got %i matches, expected 1.", dir_prop_base, matches);
goto cleanup;
}
if (strncmp(SVN_PROP_IGNORE, key, bytes_read) != 0) {
log_debug("key is %s, not %s. skipping %u bytes", key, SVN_PROP_IGNORE, entry_len);
/* Not the key we care about. fseek and repeat */
fseek(fp, entry_len + 1, SEEK_CUR); /* +1 to account for newline. yes I know this is hacky */
continue;
}
/* Aww yeah. Time to ignore stuff */
entry = ag_malloc(entry_len + 1);
bytes_read = fread(entry, 1, entry_len, fp);
entry[bytes_read] = '\0';
log_debug("entry: %s", entry);
break;
}
if (entry == NULL) {
goto cleanup;
}
char *patterns = entry;
size_t patterns_len = strlen(patterns);
while (*patterns != '\0' && patterns < (entry + bytes_read)) {
for (line_len = 0; line_len < patterns_len; line_len++) {
if (patterns[line_len] == '\n') {
break;
}
}
if (line_len > 0) {
entry_line = ag_strndup(patterns, line_len);
add_ignore_pattern(ig, entry_line);
free(entry_line);
}
patterns += line_len + 1;
patterns_len -= line_len + 1;
}
free(entry);
cleanup:
free(dir_prop_base);
free(key);
fclose(fp);
}
void search_buf(const char *buf, const int buf_len,
const char *dir_full_path) {
int binary = -1; /* 1 = yes, 0 = no, -1 = don't know */
int buf_offset = 0;
if (opts.search_stream) {
binary = 0;
} else if (!opts.search_binary_files) {
binary = is_binary((void*) buf, buf_len);
if (binary) {
log_debug("File %s is binary. Skipping...", dir_full_path);
return;
}
}
int matches_len = 0;
match *matches;
size_t matches_size;
size_t matches_spare;
if (opts.invert_match) {
/* If we are going to invert the set of matches at the end, we will need
* one extra match struct, even if there are no matches at all. So make
* sure we have a nonempty array; and make sure we always have spare
* capacity for one extra.
*/
matches_size = 100;
matches = ag_malloc(matches_size * sizeof(match));
matches_spare = 1;
} else {
matches_size = 0;
matches = NULL;
matches_spare = 0;
}
if (opts.literal) {
const char *match_ptr = buf;
strncmp_fp ag_strnstr_fp = get_strstr(opts);
while (buf_offset < buf_len) {
match_ptr = ag_strnstr_fp(match_ptr, opts.query, buf_len - buf_offset, opts.query_len, skip_lookup);
if (match_ptr == NULL) {
break;
}
if (opts.word_regexp) {
const char *start = match_ptr;
const char *end = match_ptr + opts.query_len;
/* Check whether both start and end of the match lie on a word
* boundary
*/
if ((start == buf ||
is_wordchar(*(start - 1)) != opts.literal_starts_wordchar)
&&
(end == buf + buf_len ||
is_wordchar(*end) != opts.literal_ends_wordchar))
{
/* It's a match */
} else {
/* It's not a match */
match_ptr += opts.query_len;
buf_offset = end - buf;
continue;
}
}
if ((size_t)matches_len + matches_spare >= matches_size) {
matches_size = matches ? matches_size * 2 : 100;
log_debug("Too many matches in %s. Reallocating matches to %zu.", dir_full_path, matches_size);
matches = ag_realloc(matches, matches_size * sizeof(match));
}
matches[matches_len].start = match_ptr - buf;
matches[matches_len].end = matches[matches_len].start + opts.query_len;
buf_offset = matches[matches_len].end;
log_debug("Match found. File %s, offset %i bytes.", dir_full_path, matches[matches_len].start);
matches_len++;
match_ptr += opts.query_len;
if (matches_len >= opts.max_matches_per_file) {
log_err("Too many matches in %s. Skipping the rest of this file.", dir_full_path);
break;
}
}
} else {
int rc;
int offset_vector[3];
while (buf_offset < buf_len &&
(rc = pcre_exec(opts.re, opts.re_extra, buf, buf_len, buf_offset, 0, offset_vector, 3)) >= 0) {
log_debug("Regex match found. File %s, offset %i bytes.", dir_full_path, offset_vector[0]);
buf_offset = offset_vector[1];
/* TODO: copy-pasted from above. FIXME */
if ((size_t)matches_len + matches_spare >= matches_size) {
matches_size = matches ? matches_size * 2 : 100;
log_debug("Too many matches in %s. Reallocating matches to %zu.", dir_full_path, matches_size);
matches = ag_realloc(matches, matches_size * sizeof(match));
}
matches[matches_len].start = offset_vector[0];
matches[matches_len].end = offset_vector[1];
matches_len++;
if (matches_len >= opts.max_matches_per_file) {
log_err("Too many matches in %s. Skipping the rest of this file.", dir_full_path);
break;
}
}
}
if (opts.invert_match) {
matches_len = invert_matches(matches, matches_len, buf_len);
}
if (opts.stats) {
pthread_mutex_lock(&stats_mtx);
stats.total_bytes += buf_len;
stats.total_files++;
stats.total_matches += matches_len;
pthread_mutex_unlock(&stats_mtx);
}
if (matches_len > 0) {
if (binary == -1 && !opts.print_filename_only) {
binary = is_binary((void*) buf, buf_len);
}
pthread_mutex_lock(&print_mtx);
if (opts.print_filename_only) {
print_path(dir_full_path, '\n');
} else if (binary) {
print_binary_file_matches(dir_full_path);
} else {
print_file_matches(dir_full_path, buf, buf_len, matches, matches_len);
}
pthread_mutex_unlock(&print_mtx);
} else {
log_debug("No match in %s", dir_full_path);
}
if (matches_size > 0) {
free(matches);
}
}
/* TODO: Append matches to some data structure instead of just printing them out.
* Then ag can have sweet summaries of matches/files scanned/time/etc.
*/
void search_dir(ignores *ig, const char *base_path, const char *path, const int depth) {
struct dirent **dir_list = NULL;
struct dirent *dir = NULL;
scandir_baton_t scandir_baton;
int results = 0;
char *dir_full_path = NULL;
const char *ignore_file = NULL;
int i;
/* find agignore/gitignore/hgignore/etc files to load ignore patterns from */
for (i = 0; opts.skip_vcs_ignores ? (i == 0) : (ignore_pattern_files[i] != NULL); i++) {
ignore_file = ignore_pattern_files[i];
ag_asprintf(&dir_full_path, "%s/%s", path, ignore_file);
if (strcmp(SVN_DIR, ignore_file) == 0) {
load_svn_ignore_patterns(ig, dir_full_path);
} else {
load_ignore_patterns(ig, dir_full_path);
}
free(dir_full_path);
dir_full_path = NULL;
}
if (opts.path_to_agignore) {
load_ignore_patterns(ig, opts.path_to_agignore);
}
scandir_baton.ig = ig;
scandir_baton.base_path = base_path;
results = ag_scandir(path, &dir_list, &filename_filter, &scandir_baton);
if (results == 0) {
log_debug("No results found in directory %s", path);
goto search_dir_cleanup;
} else if (results == -1) {
if (errno == ENOTDIR) {
/* Not a directory. Probably a file. */
/* If we're only searching one file, don't print the filename header at the top. */
if (depth == 0 && opts.paths_len == 1) {
opts.print_heading = -1;
}
search_file(path);
} else {
log_err("Error opening directory %s: %s", path, strerror(errno));
}
goto search_dir_cleanup;
}
int offset_vector[3];
int rc = 0;
work_queue_t *queue_item;
for (i = 0; i < results; i++) {
queue_item = NULL;
dir = dir_list[i];
ag_asprintf(&dir_full_path, "%s/%s", path, dir->d_name);
/* If a link points to a directory then we need to treat it as a directory. */
if (!opts.follow_symlinks && is_symlink(path, dir)) {
log_debug("File %s ignored becaused it's a symlink", dir->d_name);
goto cleanup;
}
if (!is_directory(path, dir)) {
if (opts.file_search_regex) {
rc = pcre_exec(opts.file_search_regex, NULL, dir_full_path, strlen(dir_full_path),
0, 0, offset_vector, 3);
if (rc < 0) { /* no match */
log_debug("Skipping %s due to file_search_regex.", dir_full_path);
goto cleanup;
} else if (opts.match_files) {
log_debug("match_files: file_search_regex matched for %s.", dir_full_path);
pthread_mutex_lock(&print_mtx);
print_path(dir_full_path, '\n');
pthread_mutex_unlock(&print_mtx);
goto cleanup;
}
}
queue_item = ag_malloc(sizeof(work_queue_t));
queue_item->path = dir_full_path;
queue_item->next = NULL;
pthread_mutex_lock(&work_queue_mtx);
if (work_queue_tail == NULL) {
work_queue = queue_item;
} else {
work_queue_tail->next = queue_item;
}
work_queue_tail = queue_item;
pthread_mutex_unlock(&work_queue_mtx);
pthread_cond_signal(&files_ready);
log_debug("%s added to work queue", dir_full_path);
} else if (opts.recurse_dirs) {
if (depth < opts.max_search_depth) {
log_debug("Searching dir %s", dir_full_path);
ignores *child_ig = init_ignore(ig);
search_dir(child_ig, base_path, dir_full_path, depth + 1);
cleanup_ignore(child_ig);
} else {
log_err("Skipping %s. Use the --depth option to search deeper.", dir_full_path);
}
}
cleanup:
;
free(dir);
dir = NULL;
if (queue_item == NULL) {
free(dir_full_path);
dir_full_path = NULL;
}
}
search_dir_cleanup:
;
free(dir_list);
dir_list = NULL;
}
void search_file(const char *file_full_path, int search_zip_files) {
int fd;
off_t f_len = 0;
char *buf = NULL;
struct stat statbuf;
int rv = 0;
FILE *pipe = NULL;
char* tmp_file_path = NULL;
fd = open(file_full_path, O_RDONLY);
if (fd < 0) {
/* XXXX: strerror is not thread-safe */
log_err("Skipping %s: Error opening file: %s", file_full_path, strerror(errno));
goto cleanup;
}
rv = fstat(fd, &statbuf);
if (rv != 0) {
log_err("Skipping %s: Error fstat()ing file.", file_full_path);
goto cleanup;
}
//if (opts.stdout_inode != 0 && opts.stdout_inode == statbuf.st_ino) {
// log_debug("Skipping %s: stdout is redirected to it", file_full_path);
// goto cleanup;
//}
if ((statbuf.st_mode & S_IFMT) == 0) {
log_err("Skipping %s: Mode %u is not a file.", file_full_path, statbuf.st_mode);
goto cleanup;
}
if (statbuf.st_mode & S_IFIFO) {
log_debug("%s is a named pipe. stream searching", file_full_path);
pipe = fdopen(fd, "r");
//search_stream(pipe, file_full_path);
fclose(pipe);
goto cleanup;
}
f_len = statbuf.st_size;
if (f_len == 0) {
log_debug("Skipping %s: file is empty.", file_full_path);
goto cleanup;
}
if (/*!opts.literal && */f_len > INT_MAX) {
log_err("Skipping %s: pcre_exec() can't handle files larger than %i bytes.", file_full_path, INT_MAX);
goto cleanup;
}
#ifdef _WIN32
{
HANDLE hmmap = CreateFileMapping(
(HANDLE)_get_osfhandle(fd), 0, PAGE_READONLY, 0, f_len, NULL);
buf = (char *)MapViewOfFile(hmmap, FILE_SHARE_READ, 0, 0, f_len);
if (hmmap != NULL)
CloseHandle(hmmap);
}
if (buf == NULL) {
FormatMessageA(
FORMAT_MESSAGE_ALLOCATE_BUFFER |
FORMAT_MESSAGE_FROM_SYSTEM |
FORMAT_MESSAGE_IGNORE_INSERTS,
NULL, GetLastError(), 0, (LPSTR)&buf, 0, NULL);
log_err("File %s failed to load: %s.", file_full_path, buf);
LocalFree((void *)buf);
goto cleanup;
}
#else
buf = mmap(0, f_len, PROT_READ, MAP_SHARED, fd, 0);
if (buf == MAP_FAILED) {
log_err("File %s failed to load: %s.", file_full_path, strerror(errno));
goto cleanup;
}
#if HAVE_MADVISE
madvise(buf, f_len, MADV_SEQUENTIAL);
#elif HAVE_POSIX_FADVISE
posix_fadvise(fd, 0, f_len, POSIX_MADV_SEQUENTIAL);
#endif
#endif
tmp_file_path = (char *)ag_malloc(MAX_PATH);
if (search_zip_files) {
ag_compression_type zip_type = is_zipped(buf, f_len);
if (zip_type != AG_NO_COMPRESSION) {
int _buf_len = (int)f_len;
char *_buf = (char*)decompress(zip_type, buf, f_len, file_full_path, &_buf_len);
if (_buf == NULL || _buf_len == 0) {
log_err("Cannot decompress zipped file %s", file_full_path);
goto cleanup;
}
search_buf(_buf, _buf_len, file_full_path, tmp_file_path);
free(_buf);
goto cleanup;
}
}
search_buf(buf, f_len, file_full_path, tmp_file_path);
cleanup:
if (buf != NULL) {
#ifdef _WIN32
UnmapViewOfFile(buf);
#else
munmap(buf, f_len);
#endif
}
if (fd != -1) {
close(fd);
}
if (tmp_file_path != NULL)
{
DeleteFileA(tmp_file_path);
free(tmp_file_path);
}
}
/* This function is REALLY HOT. It gets called for every file */
int filename_filter(const char *path, const struct dirent *dir, void *baton) {
const char *filename = dir->d_name;
size_t filename_len = strlen(filename);
size_t i;
scandir_baton_t *scandir_baton = (scandir_baton_t*) baton;
const ignores *ig = scandir_baton->ig;
const char *base_path = scandir_baton->base_path;
size_t base_path_len = strlen(base_path);
const char *path_start = path;
char *temp;
if (!opts.follow_symlinks && is_symlink(path, dir)) {
log_debug("File %s ignored becaused it's a symlink", dir->d_name);
return 0;
}
if (is_named_pipe(path, dir)) {
log_debug("%s ignored because it's a named pipe", path);
return 0;
}
for (i = 0; evil_hardcoded_ignore_files[i] != NULL; i++) {
if (strcmp(filename, evil_hardcoded_ignore_files[i]) == 0) {
return 0;
}
}
if (!opts.search_hidden_files && filename[0] == '.') {
return 0;
}
if (opts.search_all_files && !opts.path_to_agignore) {
return 1;
}
for (i = 0; base_path[i] == path[i] && i < base_path_len; i++) {
/* base_path always ends with "/\0" while path doesn't, so this is safe */
path_start = path + i + 2;
}
log_debug("path_start is %s", path_start);
if (path_ignore_search(ig, path_start, filename, scandir_baton->level)) {
return 0;
}
if (is_directory(path, dir) && filename[filename_len - 1] != '/') {
ag_asprintf(&temp, "%s/", filename);
int rv = path_ignore_search(ig, path_start, temp, scandir_baton->level);
free(temp);
if (rv) {
return 0;
}
}
if (ig->parent != NULL) {
scandir_baton_t *ignore_relay = ag_malloc(sizeof(scandir_baton_t));
memcpy(ignore_relay, scandir_baton, sizeof(scandir_baton_t));
ignore_relay->ig = ig->parent;
ignore_relay->level++;
int rv = filename_filter(path, dir, (void *)ignore_relay);
free(ignore_relay);
return rv;
}
return 1;
}