static int check_speling(request_rec *r)
{
spconfig *cfg;
char *good, *bad, *postgood, *url;
int filoc, dotloc, urlen, pglen;
DIR *dirp;
struct DIR_TYPE *dir_entry;
array_header *candidates = NULL;
cfg = ap_get_module_config(r->per_dir_config, &speling_module);
if (!cfg->enabled) {
return DECLINED;
}
/* We only want to worry about GETs */
if (r->method_number != M_GET) {
return DECLINED;
}
/* We've already got a file of some kind or another */
if (r->proxyreq != NOT_PROXY || (r->finfo.st_mode != 0)) {
return DECLINED;
}
/* This is a sub request - don't mess with it */
if (r->main) {
return DECLINED;
}
/*
* The request should end up looking like this:
* r->uri: /correct-url/mispelling/more
* r->filename: /correct-file/mispelling r->path_info: /more
*
* So we do this in steps. First break r->filename into two pieces
*/
filoc = ap_rind(r->filename, '/');
/*
* Don't do anything if the request doesn't contain a slash, or
* requests "/"
*/
if (filoc == -1 || strcmp(r->uri, "/") == 0) {
return DECLINED;
}
/* good = /correct-file */
good = ap_pstrndup(r->pool, r->filename, filoc);
/* bad = mispelling */
bad = ap_pstrdup(r->pool, r->filename + filoc + 1);
/* postgood = mispelling/more */
postgood = ap_pstrcat(r->pool, bad, r->path_info, NULL);
urlen = strlen(r->uri);
pglen = strlen(postgood);
/* Check to see if the URL pieces add up */
if (strcmp(postgood, r->uri + (urlen - pglen))) {
return DECLINED;
}
/* url = /correct-url */
url = ap_pstrndup(r->pool, r->uri, (urlen - pglen));
/* Now open the directory and do ourselves a check... */
dirp = ap_popendir(r->pool, good);
if (dirp == NULL) { /* Oops, not a directory... */
return DECLINED;
}
candidates = ap_make_array(r->pool, 2, sizeof(misspelled_file));
dotloc = ap_ind(bad, '.');
if (dotloc == -1) {
dotloc = strlen(bad);
}
while ((dir_entry = readdir(dirp)) != NULL) {
sp_reason q;
/*
* If we end up with a "fixed" URL which is identical to the
* requested one, we must have found a broken symlink or some such.
* Do _not_ try to redirect this, it causes a loop!
*/
if (strcmp(bad, dir_entry->d_name) == 0) {
ap_pclosedir(r->pool, dirp);
return OK;
}
/*
* miscapitalization errors are checked first (like, e.g., lower case
* file, upper case request)
*/
else if (strcasecmp(bad, dir_entry->d_name) == 0) {
misspelled_file *sp_new;
sp_new = (misspelled_file *) ap_push_array(candidates);
sp_new->name = ap_pstrdup(r->pool, dir_entry->d_name);
sp_new->quality = SP_MISCAPITALIZED;
}
/*
* simple typing errors are checked next (like, e.g.,
* missing/extra/transposed char)
*/
else if ((q = spdist(bad, dir_entry->d_name)) != SP_VERYDIFFERENT) {
misspelled_file *sp_new;
sp_new = (misspelled_file *) ap_push_array(candidates);
sp_new->name = ap_pstrdup(r->pool, dir_entry->d_name);
sp_new->quality = q;
}
/*
* The spdist() should have found the majority of the misspelled
* requests. It is of questionable use to continue looking for
* files with the same base name, but potentially of totally wrong
* type (index.html <-> index.db).
* I would propose to not set the WANT_BASENAME_MATCH define.
* 08-Aug-1997 <*****@*****.**>
*
* However, Alexei replied giving some reasons to add it anyway:
* > Oh, by the way, I remembered why having the
* > extension-stripping-and-matching stuff is a good idea:
* >
* > If you're using MultiViews, and have a file named foobar.html,
* > which you refer to as "foobar", and someone tried to access
* > "Foobar", mod_speling won't find it, because it won't find
* > anything matching that spelling. With the extension-munging,
* > it would locate "foobar.html". Not perfect, but I ran into
* > that problem when I first wrote the module.
*/
else {
#ifdef WANT_BASENAME_MATCH
/*
* Okay... we didn't find anything. Now we take out the hard-core
* power tools. There are several cases here. Someone might have
* entered a wrong extension (.htm instead of .html or vice
* versa) or the document could be negotiated. At any rate, now
* we just compare stuff before the first dot. If it matches, we
* figure we got us a match. This can result in wrong things if
* there are files of different content types but the same prefix
* (e.g. foo.gif and foo.html) This code will pick the first one
* it finds. Better than a Not Found, though.
*/
int entloc = ap_ind(dir_entry->d_name, '.');
if (entloc == -1) {
entloc = strlen(dir_entry->d_name);
}
if ((dotloc == entloc)
&& !strncasecmp(bad, dir_entry->d_name, dotloc)) {
misspelled_file *sp_new;
sp_new = (misspelled_file *) ap_push_array(candidates);
sp_new->name = ap_pstrdup(r->pool, dir_entry->d_name);
sp_new->quality = SP_VERYDIFFERENT;
}
#endif
}
}
ap_pclosedir(r->pool, dirp);
if (candidates->nelts != 0) {
/* Wow... we found us a mispelling. Construct a fixed url */
char *nuri;
const char *ref;
misspelled_file *variant = (misspelled_file *) candidates->elts;
int i;
ref = ap_table_get(r->headers_in, "Referer");
qsort((void *) candidates->elts, candidates->nelts,
sizeof(misspelled_file), sort_by_quality);
/*
* Conditions for immediate redirection:
* a) the first candidate was not found by stripping the suffix
* AND b) there exists only one candidate OR the best match is not
* ambiguous
* then return a redirection right away.
*/
if (variant[0].quality != SP_VERYDIFFERENT
&& (candidates->nelts == 1
|| variant[0].quality != variant[1].quality)) {
nuri = ap_escape_uri(r->pool, ap_pstrcat(r->pool, url,
variant[0].name,
r->path_info, NULL));
if (r->parsed_uri.query)
nuri = ap_pstrcat(r->pool, nuri, "?", r->parsed_uri.query, NULL);
ap_table_setn(r->headers_out, "Location",
ap_construct_url(r->pool, nuri, r));
ap_log_rerror(APLOG_MARK, APLOG_NOERRNO | APLOG_INFO, r,
ref ? "Fixed spelling: %s to %s from %s"
: "Fixed spelling: %s to %s",
r->uri, nuri, ref);
return HTTP_MOVED_PERMANENTLY;
}
/*
* Otherwise, a "[300] Multiple Choices" list with the variants is
* returned.
*/
else {
pool *p;
table *notes;
pool *sub_pool;
array_header *t;
array_header *v;
if (r->main == NULL) {
p = r->pool;
notes = r->notes;
}
else {
p = r->main->pool;
notes = r->main->notes;
}
sub_pool = ap_make_sub_pool(p);
t = ap_make_array(sub_pool, candidates->nelts * 8 + 8,
sizeof(char *));
v = ap_make_array(sub_pool, candidates->nelts * 5,
sizeof(char *));
/* Generate the response text. */
*(const char **)ap_push_array(t) =
"The document name you requested (<code>";
*(const char **)ap_push_array(t) = ap_escape_html(sub_pool, r->uri);
*(const char **)ap_push_array(t) =
"</code>) could not be found on this server.\n"
"However, we found documents with names similar "
"to the one you requested.<p>"
"Available documents:\n<ul>\n";
for (i = 0; i < candidates->nelts; ++i) {
char *vuri;
const char *reason;
reason = sp_reason_str[(int) (variant[i].quality)];
/* The format isn't very neat... */
vuri = ap_pstrcat(sub_pool, url, variant[i].name, r->path_info,
(r->parsed_uri.query != NULL) ? "?" : "",
(r->parsed_uri.query != NULL)
? r->parsed_uri.query : "",
NULL);
*(const char **)ap_push_array(v) = "\"";
*(const char **)ap_push_array(v) = ap_escape_uri(sub_pool, vuri);
*(const char **)ap_push_array(v) = "\";\"";
*(const char **)ap_push_array(v) = reason;
*(const char **)ap_push_array(v) = "\"";
*(const char **)ap_push_array(t) = "<li><a href=\"";
*(const char **)ap_push_array(t) = ap_escape_uri(sub_pool, vuri);
*(const char **)ap_push_array(t) = "\">";
*(const char **)ap_push_array(t) = ap_escape_html(sub_pool, vuri);
*(const char **)ap_push_array(t) = "</a> (";
*(const char **)ap_push_array(t) = reason;
*(const char **)ap_push_array(t) = ")\n";
/*
* when we have printed the "close matches" and there are
* more "distant matches" (matched by stripping the suffix),
* then we insert an additional separator text to suggest
* that the user LOOK CLOSELY whether these are really the
* files she wanted.
*/
if (i > 0 && i < candidates->nelts - 1
&& variant[i].quality != SP_VERYDIFFERENT
&& variant[i + 1].quality == SP_VERYDIFFERENT) {
*(const char **)ap_push_array(t) =
"</ul>\nFurthermore, the following related "
"documents were found:\n<ul>\n";
}
}
*(const char **)ap_push_array(t) = "</ul>\n";
/* If we know there was a referring page, add a note: */
if (ref != NULL) {
*(const char **)ap_push_array(t) =
"Please consider informing the owner of the "
"<a href=\"";
*(const char **)ap_push_array(t) = ap_escape_uri(sub_pool, ref);
*(const char **)ap_push_array(t) = "\">referring page</a> "
"about the broken link.\n";
}
/* Pass our table to http_protocol.c (see mod_negotiation): */
ap_table_setn(notes, "variant-list", ap_array_pstrcat(p, t, 0));
ap_table_mergen(r->subprocess_env, "VARIANTS",
ap_array_pstrcat(p, v, ','));
ap_destroy_pool(sub_pool);
ap_log_rerror(APLOG_MARK, APLOG_NOERRNO | APLOG_INFO, r,
ref ? "Spelling fix: %s: %d candidates from %s"
: "Spelling fix: %s: %d candidates",
r->uri, candidates->nelts, ref);
return HTTP_MULTIPLE_CHOICES;
}
}
return OK;
}
const char *fcgi_config_make_dynamic_dir(pool *p, const int wax)
{
const char *err;
pool *tp;
fcgi_dynamic_dir = ap_pstrcat(p, fcgi_socket_dir, "/dynamic", NULL);
if ((err = fcgi_config_make_dir(p, fcgi_dynamic_dir)))
return ap_psprintf(p, "can't create dynamic directory \"%s\": %s", fcgi_dynamic_dir, err);
/* Don't step on a running server unless its OK. */
if (!wax)
return NULL;
#ifdef APACHE2
{
apr_dir_t * dir;
apr_finfo_t finfo;
if (apr_pool_create(&tp, p))
return "apr_pool_create() failed";
if (apr_dir_open(&dir, fcgi_dynamic_dir, tp))
return "apr_dir_open() failed";
/* delete the contents */
while (apr_dir_read(&finfo, APR_FINFO_NAME, dir) == APR_SUCCESS)
{
if (strcmp(finfo.name, ".") == 0 || strcmp(finfo.name, "..") == 0)
continue;
apr_file_remove(finfo.name, tp);
}
}
#else /* !APACHE2 */
{
DIR *dp;
struct dirent *dirp = NULL;
tp = ap_make_sub_pool(p);
dp = ap_popendir(tp, fcgi_dynamic_dir);
if (dp == NULL) {
ap_destroy_pool(tp);
return ap_psprintf(p, "can't open dynamic directory \"%s\": %s",
fcgi_dynamic_dir, strerror(errno));
}
/* delete the contents */
while ((dirp = readdir(dp)) != NULL)
{
if (strcmp(dirp->d_name, ".") == 0 || strcmp(dirp->d_name, "..") == 0)
continue;
unlink(ap_pstrcat(tp, fcgi_dynamic_dir, "/", dirp->d_name, NULL));
}
}
#endif /* !APACHE2 */
ap_destroy_pool(tp);
return NULL;
}
