weak_vendors mode and phars

[greg0ire]

Q	A
Bug report?	yes
Feature request?	no
BC Break report?	no
RFC?	no
Symfony version	3.3.*

When introducing this mode, I overlooked the fact that file paths could look like this:

phar:///usr/local/bin/phpunit-5/phpunit/Util/Fileloader.php

In that kind of case, there is not always a way to tell if the deprecation was triggered inside or outside a vendor, so IMO I should add a special case just for phar paths in that method:

symfony/src/Symfony/Bridge/PhpUnit/DeprecationErrorHandler.php

Lines 64 to 86 in ba3b177

	$inVendors = function ($path) {
	/** @var string[] absolute paths to vendor directories */
	static $vendors;
	if (null === $vendors) {
	foreach (get_declared_classes() as $class) {
	if ('C' === $class[0] && 0 === strpos($class, 'ComposerAutoloaderInit')) {
	$r = new \ReflectionClass($class);
	$v = dirname(dirname($r->getFileName()));
	if (file_exists($v.'/composer/installed.json')) {
	$vendors[] = $v;
	}
	}
	}
	}
	$path = realpath($path) ?: $path;
	foreach ($vendors as $vendor) {
	if (0 === strpos($path, $vendor) && false !== strpbrk(substr($path, strlen($vendor), 1), '/'.DIRECTORY_SEPARATOR)) {
	return true;
	}
	}
	return false;
	};

Not sure how realpath behave with such paths, BTW

I wonder how / if I'm going to be able to test that though... any ideas?

[stof]

realpath will return false

I think we should consider that any code being inside a phar is vendor code in this mode. When running a testsuite and having code inside a phar, it is much more likely to be the test runner than the code under test (do you actually build a phar before each test execution, taking into account that this cannot be your executable phar if you are doing unit tests ?)

[greg0ire]

@stof I experienced this bug once with a phar, but non-vendor code: it was eval'd code produced by getMockForAbstractClass, and the class was a non-vendor class.

But I agree with you that in doubt, we should consider it vendor code.

realpath will return false

good to know :/

[greg0ire]

A simple fix might be to return true if realpath returns false then?

[greg0ire]

I think I will be able to test this by using eval