Enhance wp_debug_backtrace_summary() with the optional ability to include arguments

[DavidAnderson]

The attached patch for wp_debug_backtrace_summary():

• Adds the optional ability to include arguments in the output

• Leaves the default behaviour as-is

• Removes an unnecessary check on whether the PHP version is at least a value which is still lower than the minimum one that WP supports

[johnbillion]

I think this is a good enhancement.

Query Monitor does a similar thing: ​https://github.com/johnbillion/query-monitor/blob/143c31484656495b024144b9e8697c863f8ec159/classes/Util.php#L336-L368 . After a skim over your patch, I think the main difference is that QM adds the object ID for some known object types.

[DavidAnderson]

Updated patch - previous version was not based off current SVN trunk

[DavidAnderson]

Though changing default behaviour is usually undesirable, I wonder if in the case of a debugging function whose output was, and remains, a textual string, if there's not a good case for changing it in this case? The extra info is harmless, and it's hard to envisage anybody parsing the output and relying upon it in a way that would be breaking. Having to pass all 4 parameters to get the info is a bit annoying.

[jdgrimes]

Replying to DavidAnderson:

Though changing default behaviour is usually undesirable, I wonder if in the case of a debugging function whose output was, and remains, a textual string, if there's not a good case for changing it in this case? The extra info is harmless, and it's hard to envisage anybody parsing the output and relying upon it in a way that would be breaking. Having to pass all 4 parameters to get the info is a bit annoying.

The extra info is not necessarily harmless. I can see how this could inadvertently expose passwords or be used for XSS under the correct circumstances.

Also, I don't know about parsing the output, but core itself searches the output for a particular function name in one of the unit tests. I don't know if any plugins do that in production code, but I wouldn't be surprised. If they do, it would potentially be possible for the user to alter behavior if they control the content of any of the parameters in the backtrace stack, by including a function name in the argument value.

[DavidAnderson]

@jdgrimes Good points.