== link:index.html[Index] -> link:modules.html[Modules] -> link:modules_handlers.html[Handlers]
Handler: CGI
------------
The cgi handler executes CGI programs.
[[parameters]]
Parameters
~~~~~~~~~~
This module accepts three parameters:
[cols="30%,15%,55%",options="header"]
|====================================================================
|Parameters |Type |Description
|Script Alias |String |It defines a kind of execution
interpreter. The CGI will be run as a
parameter of this script.
|Change to UID |String |Executes each CGI under its owner's
user ID.
|Error handler |Boolean |Use CGI output as error message.
|Check file |Boolean |Check file existence to build
PATH_INFO and SCRIPT_NAME.
|Pass request |Boolean |Pass request headers.
|Allow X-Sendfile |Boolean |Use the non-standard X-Sendfile
header.
|Pass Request Headers |Boolean |Forward all the headers to the CGI as
HTTP_*
|====================================================================
Note that the link:other_goodies.html#x-sendfile[X-Sendfile] feature
also supports `X-Accel-Redirect` headers transparently. This is to
allow the migration of applications supporting it whithout having to
make major code rewrites.
[[check_file_warning]]
Warning:
~~~~~~~~
The `Check file` flag affects how PATH_INFO and SCRIPT_NAME are
generated. You should disable this when you have an application that
is running the whole virtual server tree, as it is a common source of
problems. You can look at a practical example about PATH_INFO in the
link:modules_handlers_common.html[List & Send] handler documentation.
`Check file` must be disabled for cases such as Rails, Django, etc,
where everything that arrives to Cherokee must be re-sent to the
framework as request.
Enabling `Check file` is the way to go if you are using any kind of
file-based system, such as PHP.
If you are unsure of the way this is being taken into account, try
both settings and see how your application behaves.
[[examples]]
Examples
~~~~~~~~
This is a pretty common example. It defines a directory called `cgi`
that points to a path out of the Sever root directory where the
CGI programs are located.
.Handler CGI configuration
[cols="50%,50%"]
|====================================================================
|Handler | CGI
|Document Root | /usr/lib/cgi-bin
|Script Alias | None
|Change UID | Disabled
|Error handler | Disabled
|Check file | Enabled
|Pass Request Headers | Enabled
|Allow X-Sendfile | Disabled
|====================================================================
.Handler CGI intreface
image::media/images/admin_handler_cgi.png[CGI]
.Script Alias
Another example could be the case where you wanted your default
requests to be managed by a specific script.
A real life example could be running
link:http://trac.edgewall.org[Trac] as a CGI script. In that case you
would want to map an URL to the trac.cgi script.
For this you only need to set up the Script Alias option. Assuming you
are using a `Directory`-type rule for the path `/trac` managed by the
`CGI` handler, you could set `/usr/share/trac/cgi-bin/trac.cgi` as
Script Alias.
After this, every request that matched the `/trac` path would be
managed by the `trac.cgi` script.
//////
Environment variables
~~~~~~~~~~~~~~~~~~~~~
When working with CGI and redirections, sometimes it is convenient to
have some persistent information between the CGI invocations. This is
the case when an error has to be redirected to a error-handling page.
To ease up accessing such values, some server environment variables are
set by Cherokee. The complete list of environment variables added
follows:
This info will be added upon completion of issue 40
* REDIRECT_URL: the URL-path that was not found. For example, if a
missing page -`http://example.net/nonexistent.htm`- was requested,
this variable would be set to `/nonexistent.html`.
* REDIRECT_STATUS: the HTTP response obtained from the original
request. This is `404`.
* REDIRECT_REQUEST_METHOD: the method of the original request, such as
GET or POST.
* REDIRECT_ERROR_NOTES: a brief description of what went wrong, for
example, "File does not exist: /usr/local/docroot/nonexistent.html".
* SERVER_ADDR: This will typically be `::1` which is a unicast
localhost address. This means that the packets sent to this address
will be looped back on the same virtual interface.
* SERVER_PORT: Port to which the request was performed originally.
* REMOTE_PORT: Port to which the request was served.
//////