_deprecated_file undefined in rss-functions.php

[thib3113]

If you call wp-includes/rss-functions.php directly, you have an error .

Because _deprecated_file is not defined, you can patch with :

<?php
/**
 * Deprecated. Use rss.php instead.
 *
 * @package WordPress
 */

if(defined(ABSPATH)){
    _deprecated_file( basename(__FILE__), '2.1', WPINC . '/rss.php' );
    require_once( ABSPATH . WPINC . '/rss.php' );
}

[dd32]

Production sites shouldn't be running with display_errors turned on, so the undefined constants & undefined functions warnings/notices/errors aren't seen as a bugs within WordPress.

We also cover this in our ​security handbook.

[thib3113]

Yes, but visiting this webpage produce a http 500 error. So yes, it's not visible, but it exist ...

Why don't just add a single ligne for stop this error ? And making the same in other not-standalone page ...

[dd32]

Yes, but visiting this webpage produce a http 500 error.

On a URL which should never be linked to (And you should also have directory indexing disabled ideally).

You'll get a similar error if you access wp-includes/default-filters.php or any number of others too potentially.

[thib3113]

Replying to dd32:

Yes, but visiting this webpage produce a http 500 error.

On a URL which should never be linked to (And you should also have directory indexing disabled ideally).

It's is not linked, but pentest software know this addresses and check for path. You'are right, in production, we will set debug to false, but in my opinion, we can make attack with this error, maybe with reloading lot of time, and log lot of errors ( I don't know, i'm not a professional in pentesting )...
I just think, let an error is not a good thing, mostly if you just add one line for remove this ...

if(!defined(ABSPATH)) exit();

[TheGP]

Replying to dd32:

Production sites shouldn't be running with display_errors turned on, so the undefined constants & undefined functions warnings/notices/errors aren't seen as a bugs within WordPress.

We also cover this in our ​security handbook.

That's stupid. I have error notification alerts, and I had alerts about this fatal error too. So I just blocked all errors from Wordpress. And this not improving security, because I can't see any errors from Wordpress right now. So fix it and dont make excuses.

+ it adds a lot of trash data to log files (i have thousands of these errors)
+ bots use this bugs in wordpress to scan websites with display_errors On

No wonder wordpress code developers considered as the worst in the industry))

[dd32]

Remarking as invalid. Discussion can continue; a committer can re-open if they feel the need to.

[SergeyBiryukov]

#47945 was marked as a duplicate.

[flymike]

Some hacker has discovered many of the WordPress files containing calls to _deprecated_file() and is inundating my server with direct GET requests to them.
Because that function is not defined in WordPress, Apache returns status 500 and - because, as an administrator, I want to be informed of status 500 - my inbox is deluged with alerts.
I would block the originating IPs but they'e all different, so coming from spambots. And the advantage to the hacker eludes me completely - but it is what it is, and I have to deal with it.
Couldn't WordPress handle calls to deprecated files/functions a little more elegantly? Like it does with direct calls to other files which should not be accessed directly - with status 200 and zero bytes?

[SergeyBiryukov]

Related: #47154

[zodiac1978]

I am not sure why this is handled in this way but the PHP updates in a complete other way.

We do not want to annoy the user with technical details about PHP updates, but we try to make them aware that there is something to fix. We should do the same here at least.

I am not sure why the mentioned line is not added to WordPress to all files which are in need of this, but if we are going this route, why not inform the user about the risk of an enabled display_errors setting?

Therefore I have added an issue to the health check plugin about adding this check:
​https://github.com/WordPress/health-check/issues/370

If you are searching for this error message you get roughly 30k hits and these are just those websites which also have enabled public accessible error logging (Google finds the error log file with this error message in it). There will be much more websites with this problem.

[williampatton]

Just noting here that I've also ran into bots hitting this file on a website that's high traffic and it causes errors to be logged resulting in alert notifications being sent when they shouldn't be.

If the solution is as simple as making sure the function is available before accessing it can it be added to all the deprecated files?

[williampatton]

Patch 35835.diff from @whyisjake solves this issue on this file for me.

Is anything else blocking having this merged and released?

[whyisjake]

Yeah, I agree with @williampatton, this is basically a bug that should have been addressed long ago.

[whyisjake]

In 49619:

Feeds: Add an early exit when calling RSS functions directly.

While this file is depracated, it shouldn't throw errors when called directly.

Fixes #35835.

Props thib3113, dd32, TheGP, flymike, zodiac1978, williampatton.

[SergeyBiryukov]

There are ~30 files in core with _deprecated_file(), why only this one is being patched? Is there anything specific calling this file directly, but not the others?

Not arguing for or against the change, just for a consistent policy.

This is also not limited to _deprecated_file(), as noted in comment:3 and comment:1:

You'll get a similar error if you access wp-includes/default-filters.php or any number of others too potentially.

Should we try to address other instances too and edit the ​security handbook accordingly?

[williampatton]

Replying to SergeyBiryukov:

There are ~30 files in core with _deprecated_file(), why only this one is being patched? Is there anything specific calling this file directly, but not the others?

In my case it was a fuzzing bot that was calling the file directly. I presume to get a file inclusion path disclosure of some kind but I can't be sure of the motives.

Should we try to address other instances too and edit the ​security handbook accordingly?

I agree that we should try and address other instances too if they can easily be identified. Today I only had this one file in my logs as being accessed directly but in the past there have been some others (I recall one being from ID3 library but I am certain there have been more over time).

[SergeyBiryukov]

Created #51806 for other instances, re-closing this one as fixed for 5.6.