Clearly-defined "undefined property"

[spud]

Within a certain class file, I have the following lines:

echo '<form id="' . $this->id_prefix . 'form" class="' . $t_form_class . '" action="' . $this->action . '" method="'  . $this->method . '" accept-charset="utf-8">';

// Give the form a title if provided
if (!empty($this->title)) {
   echo '<h2 class="f_title" id="' . $this->id_prefix . '_title">' . $this->title . '</h2>';
}

Note that the 1st and 5th lines both refer to $this->id_prefix. When I run PHPStan, it generates an error: Access to an undefined property object::$id_prefix only for the second instance of the property reference (line 5 here, within the conditional). If I remove the conditional "if (!empty($this->title))" wrapping, it passes without error.

What's going on that it acknowledges the legitimacy of the first reference, but not the second?

[ondrejmirtes]

Please reproduce your issue at phpstan.org and share the link, thanks.

[spud]

Interestingly, I cannot reproduce it on the site. I'm running PHPStan (0.12.64) against several directories of files within my codebase, and bootstrapping a bunch of stuff, so the errors between my execution and those on phpstan.org are quite different. The particular error that concerns me does not appear on phpstan.org.

I have two bootstrap files and 7 directories in my config path (phpstan.neon). If I remove those, I get 70+ errors, similar to phpstan.org, and I do not get the undefined property error.

Somehow, the error is the result of parsing more than just the single file.

I can make the error disappear in one of two ways:

1. Remove the conditional if wrapper around the line containing the "undefined" reference
2. A few lines earlier in my file, there is a middleware-like block of code that includes $listener->call_function($this); (And the call_function method takes its parameter by reference) If I change the $this to $foo (which is undefined), the file then passes PHPStan without error.

So I'm not sure what's going on, nor why one reference to $this->id_prefix is recognized as valid, but the other is not.

And I don't want to waste your support time on a very specific anomaly, but if you think of anything, let me know!

In the meantime, I'll continue working to remove all the errors (370 to go!) to see if there is some cumulative effect responsible. I have noticed that PHPStan returns results in a slightly strange manner if it encounters a single parse error in any of the many files in my scans.

But...thanks for making this, too! More work for me, but better software!