Opened 8 years ago
Last modified 4 years ago
#40566 new enhancement
add_query_arg() returns only URL fragments in certain circumstances
Reported by: | DavidAnderson | Owned by: | |
---|---|---|---|
Milestone: | Future Release | Priority: | normal |
Severity: | normal | Version: | |
Component: | General | Keywords: | needs-patch |
Focuses: | docs | Cc: |
Description (last modified by )
The documentation for add_query_arg() - https://developer.wordpress.org/reference/functions/add_query_arg/ - and its docblock, both claim that it returns a URL. And this appears to be how it is commonly used. (I started looking into this because of it being used in WooCommerce in this way, a wrong assumption which ulimately causes a reproducible-every-time 404 on my webserver on password reset requests via WooCommerce).
e.g. in the documentation: "Retrieves a modified URL query string." "You can rebuild the URL and append query variables to the URL query by using this function". It is also recommended to use esc_url() on what is returned - a function that states that it acts upon URLs, implying that what is returned is indeed a URL.
The docblock says "Retrieves a modified URL query string" - which isn't a very clear statement (does it return a query string? Or a URL *with* modified query string?). It continues later with "Omitting the URL from either use results in the current URL being used (the value of $_SERVER['REQUEST_URI']
)".
The last clause is where the problem comes in. $_SERVER['REQUEST_URI']
, actually stores a path that is relative to the current host, only (i.e. not "U"niversal). Another beautiful bit of PHP mis-design: http://php.net/manual/en/reserved.variables.server.php - "The URI which was given in order to access this page; for instance, '/index.html'."
As a result, despite the documentation and expectation, add_query_arg() returns relative fragments. And when these are passed to wp_redirect(), as WooCommerce does, the result can be, as in my case, that the webserver gives a 404. (This appears to be related to changes to relative path handling in CGI output in recent versions of lighttpd. The problem is CGI-specific I think - it relates to how a webserver handles the output of a CGI script when that output has a Location: header without a full URL in it).
It appears to me that in the default case of no URL being specified, add_query_arg() should return an actual URL, instead of a relative path.
Perhaps related: #14062
Change History (4)
#2
@
8 years ago
- Description modified (diff)
- Summary changed from Defect: add_query_arg() returns only URL fragments in certain circumstances to add_query_arg() returns only URL fragments in certain circumstances
- Version trunk deleted
This ticket was mentioned in Slack in #core by noisysocks. View the logs.
4 years ago
#4
@
4 years ago
- Focuses docs added
- Keywords needs-patch added
- Milestone changed from Awaiting Review to Future Release
- Type changed from defect (bug) to enhancement
Sounds like all that is required here is to update the docs for add_query_arg()
(and possibly remove_query_arg()
) to warn that if $_SERVER['REQUEST_URI']
is a relative path then the returned value will be as well.
N.B. I believe, on reading the CGI spec, that lighttpd has a bug here. That's just the back-story - I think there's still an issue with add_query_arg() indicating that it returns an absolute location in the default case of no URL being supplied as input, rather than a relative one.