Can hunchentoot host CGI application?

We have a CGI application developed using python, which can be hosted in erlang YAWS easily:

>cat ~/yaws.conf
...
<server 192.168.1.2开发者_如何学运维>
    port = 8000
    listen = 0.0.0.0
    docroot = /media/G/www/qachina/
    access_log = false
    appmods = <cgi-bin, yaws_appmod_cgi>
</server>
...

Now we want to host the application in a lisp web server. Maybe hunchentoot can do it ?

Sincerely!

Probably something to do with EOL characters.

From Chunga API docs, you can see the read-line* function expects a CR as EOL marker, which is not the default for *nixes. The following should make it work (it works for me):

(setf chunga:*accept-bogus-eols* t)

EDIT: further reading the description of *accept-bogus-eols* I can see why this is the normal behavior of Chunga: the library itself conforms to RFC2616 (HTTP 1.1) in which the HTTP protocol's default EOL marker is CRLF.

I have installed hunchentoot-cgi and tested it. The following is a simple python script file:

>cat cgi-bin/nav.py
#!/usr/local/bin/python
# -*- coding: utf-8 -*- 

print "Content-type: text/html\n\n"
print """from Python"""

when i visited http://127.0.0.1:8000/cgi-bin/nav.py, hunchentoot reported:

11111111111111111
**End of file, but expected #\Newline.**
[2011-12-28 13:35:16 [ERROR]] error in handle-cgi-script from URL /cgi-bin/nav.py
127.0.0.1 - [2011-12-28 13:35:16] "GET /cgi-bin/nav.py HTTP/1.1" 200 - "-"    
"Opera/9.80 (X11; FreeBSD 8.2-RELEASE i386; U; zh-cn) Presto/2.10.229 Version/11.60"

By hackiing hunchentoot-cgi.lisp, i found the function "handle-cgi-script" reported the error:

 (handler-case
  (with-input-from-program (in path nil env)
  (format t "11111111111111111~%")
  (chunga:with-character-stream-semantics
      (loop for line = (chunga:read-line* in)
       until (equal line "")
       do
         (format t "22222222222222222:~A~%" line)
         (destructuring-bind (key val)
              (ppcre:split ": " line :limit 2)
              (setf (hunchentoot:header-out key) val))
              (format t "22222222222222222~%")))
  (format t "33333333333333333~%")
  (let ((out (flexi-streams:make-flexi-stream
      (tbnl:send-headers)
      :external-format tbnl::+latin-1+)))
      (copy-stream in out 'character))
  (format t "33333333333333333~%"))
  (error (error)
      (format t "~A~%" error)
      (tbnl:log-message* :error "error in handle-cgi-script from URL ~A"
      (tbnl:request-uri*))))

Any suggestion is appreciated!

Strange behavior about "merge-pathnames":

* (merge-pathnames  "nav.py" "/media/E/myapp/cgi-bin/")
#P"/media/E/myapp/cgi-bin/nav.py"

it works normally in SBCL REPL. However, when hacking "create-cgi-dispatcher-and-handler", i add the following lines:

(defun create-cgi-dispatcher-and-handler (uri-prefix base-path &optional content-type)
;...
(format t "SName=~A SPath=~A BPath=~A~% the path is ~A~%" script-name script-path base-path (merge-pathnames script-path base-path))
;...

Calling it as below:

(pushnew (hunchentoot-cgi::create-cgi-dispatcher-and-handler
      "/cgi-bin/"
      (make-pathname :name "cgi-bin/" :type nil :version nil :defaults *this-file*)
      ) *dispatch-table* :test #'equal)

Then visiting http://127.0.0.1:8000/cgi-bin/nav.py, it reports:

SName=/cgi-bin/nav.py SPath=nav.py BPath=/media/E/myapp/cgi-bin/
the path is /media/E/myapp/nav.py

In a short word:

(merge-pathnames "nav.py" "/media/E/myapp/cgi-bin/") returns #P"/media/E/myapp/cgi-bin/nav.py" in REPL. But it returns "/media/E/myapp/nav.py" in hunchentoot-cgi.lisp.

Sincerely!