hash_equals() does not compare strings in constant time

[nbachiyski]

The phpdoc of hash_equals says: “Compare two strings in constant time.”

The term “constant time” is widely used in computer science and means ​O(1) time complexity, or its running time not depending o the size of the input data. This is not the case with hash_equals, where we only make sure that input data with the same size takes the same amount of time.

The first time I saw the WordPress code it took me few minutes to understand what was going on.

Is it possible that we steal the description from php.net: “Timing attack safe string comparison” and link to the documentation there at: ​http://php.net/hash_equals?

[Otto42]

Is this a suggestion to change the function, or just the description of it?

Does it perform the same as php's hash_equals? More or less, I mean? If so, then equating the description is fine.

[dd32]

Does it perform the same as php's hash_equals? More or less, I mean? If so, then equating the description is fine.

It performs exactly the same as the PHP function - literally a port of that function to pure PHP.

Given there have been several security reports of "hash_equals() is not constant-time when string lengths differ" (which is the intended behaviour - as most string lengths are known inside PHP/WordPress already, so preventing that is mostly pointless and complex) I'd agree that making it clearer in the docs is worthwhile, if it can be done :)

[AramZS]

Clarifying hash_equals in a way consistent with PHP docs.

[AramZS]

Patch attached to clarify it in a way consistent with the documentation listed on PHP.net for the ported function. Should fulfill the clarification needs.

[nbachiyski]

In 35805:

Docs: clarify inline docs for hash_equals

Before the docs implied the complexity of the function was O(1) by using the term "constant time", now we use the more descriptive term "Timing attack safe".

Props AramZS.
Fixes #32778.