Searching for a date, e.g. "14/05/2005", results in a 404 error.

search.module redirects to http://drupal.org/search/node/14%2F05%2F2005 which isn't handled correctly by mod_rewrite.

The attached patch replaces all occurrences of '%2F' in the query with '/'.

A test query for '1/2 1~2' results in a redirect to http://drupalhead.local/search/node/1/2+1%7E2 which is handled correctly by mod_rewrite.

(Solution found via comment on http://ccca.nctu.edu.tw/~hlb/tavi/index.php?page=URL+Rewriting)

CommentFileSizeAuthor
#5 urlencode.diff4.81 KBSteven
search.module.patch740 byteswulff

Comments

dries’s picture

dries commented

Waiting for Steven's feedback on this. He know what the best fix is.

Steven’s picture

Steven commented

Actually it is not clear to me at all why an urlencoded slash would be a problem. It is even weirder that mod_rewrite can deal with unescaped slashes, but 404s on encoded slashes. In fact, if it wasn't URLencoded, then the non-clean URL would be invalid (as far as I know it is not allowed to use slashes in GET query data).

The page you linked to doesn't make me any wiser as to the cause of the problem. Does this problem occur with more than just slash characters?

Even if it is just limited to slashes, we'd need a cleaner fix than this. Someone with more mod_rewrite skills should look into this problem as it has really nothing to do with searching. If it is a mod_rewrite bug that cannot be fixed, then a wrapper around urlencode() sounds like a good idea.

wulff’s picture

wulff commented

I have done some more digging. It seems that this is not really related to mod_rewrite. According to http://137.113.100.11/manual/mod/core.html#allowencodedslashes Apache versions prior to 2.0.46 automatically return a 404 error when encountering '%2F' in the URL.

Apparently, the AllowEncodedSlashes functionality may be backported to 1.3.x (http://www.mail-archive.com/dev@httpd.apache.org/msg25089.html).

drumm’s picture

drumm
he/him
Location NY, US
commented
Status: Needs review » Needs work
Steven’s picture

Steven commented
Assigned: Unassigned » Steven
Status: Needs work » Needs review
StatusFileSize
new urlencode.diff4.81 KB

This patch adds a wrapper around urlencode() which restores the escaped slashes to real slashes, when clean URLs are turned on. This doesn't introduce any problems or breaks standards. Mod rewrite will re-escape it for us when it rewrites the query on the server side.

Before:
site.com/search/node/query+with+a+slash+%2F+in+it

After:
site.com/search/node/query+with+a+slash+/+in+it

So, it simply makes our fake directories look more like real directories in case there's a slash in there. The non-clean URL situation doesn't change:

site.com/?q=search/node/query+with+a+slash+%2F+in+it

drumm’s picture

drumm
he/him
Location NY, US
commented

Tested and works. Code looks good too. +1

Steven’s picture

Steven commented
Status: Needs review » Fixed

Committed to HEAD.

Anonymous’s picture

(not verified) commented
Status: Fixed » Closed (fixed)