request.path_info sometimes used incorrectly

[Elrond]

When running mediagoblin in a subpath, like /mediagoblin/, then some things are a little complex.

For /mediagoblin/u/elrond/ request.path_info becomes /u/elrond/.

• This is good, when trying to match routes. Simple, fine, job done.
• This is BAD, when generating URLs like "request.path_info + foo" for the browser, because the browser will only see /u/elrond/foo. Which is wrong. Problem.

I know, that the environ has SCRIPT_NAME ("/mediagoblin" in this case), but it's not entirely clear, on how to procede from here.

Assigning to Chris, as he created most of the base infrastructure and surely has comments.

[Elrond]

If mediagoblin wants to support subdir hosting for 0.1.0, then this bug should be targetted for 0.1.0.

[Elrond]

One possible, simple way to help this issue:

request.long_path = environ["SCRIPT_NAME"] + request.path_info
# or
request.long_path = environ["SCRIPT_URL"]

And then request.long_path can be used in many places to mean the long, useful URL to reach the current page (without query params).

In reality, I think, no places (except routing) need request.path_info. So maybe it should be banned somehow?

[Christopher Allan Webber]

Just to be sure, you changed how this was "mounted" via paste, is that correct? Or was it via fastcgi, or some other thing?

[Elrond]

plain fastcgi: My main webserver serves my main site. And it redirects /mediagoblin/ via fcgi to GMG.

[Elrond]

I volunteered to take care of this. Especially, as I understand the problem quite well and have an idea on how to fix it.

[Elrond]

Branch: misc/i592_use_full_path

Comments welcome!

[Christopher Allan Webber]

Merged!

[Will Kahn-Greene]

The original url for this bug was ​http://bugs.foocorp.net/issues/592 .