Edit report at https://bugs.php.net/bug.php?id=66773&edit=1
ID: 66773
Comment by: kernins at gmail dot com
Reported by: carl dot vuorinen at w3 dot fi
Summary: Autoload with Opcache allows importing conflicting
class name to namespace
Status: Assigned
Type: Bug
Package: opcache
Operating System: Linux (Ubuntu)
PHP Version: 5.5.9
Assigned To: dmitry
Block user comment: N
Private report: N
New Comment:
>Isn't the "expected" error behavior here directly in conflict with the
>documentation?
Absolutely! The actual bug here is the "expected" behaviour itself. The code in
some_foo.php from test example is perfectly valid as imports are _per_file_, not per process or
thread. Imports have precedence and allows to "shadow" classes from current NS, and they
shouldn't care what is already loaded or imported, except from current file. Otherwise, with
"expected" behaviour, they not only would be alogical and almost useless, but also a hell
to maintain in a big projects.
I have a project, that worked perfectly on some older 5.4.x and all previous versions I've used
since NS was introduced in 5.3. But after update to 5.5.9 I see exactly the same fatal errors here
and there on previously flawlessly working code when runing in CLI. Opcache is and definitely was
turned off for CLI SAPI. With FPM, wich has opcache enabled, it worked fine.
It seems, that bug was also backported to 5.4 branch - 5.4.34 I have on hand is also affected, but
I'm absolutely confident some previous 5.4.x was not.
Please, fix this nonsense asap! This bug is pretty annoying
Previous Comments:
------------------------------------------------------------------------
[2015-06-02 13:53:53] Erutan409 at Hotmail dot com
Has this been resolved, yet and/or is this still considered a bug?
------------------------------------------------------------------------
[2014-11-07 17:15:44] john at zerocrates dot org
Isn't the "expected" error behavior here directly in conflict with the documentation?
http://php.net/manual/en/language.namespaces.faq.php#language.namespaces.faq.conflict
That documentation section gives the following example:
file1.php
---------
<?php
namespace my\stuff;
class MyClass {}
?>
another.php
-----------
<?php
namespace another;
class thing {}
?>
file2.php
---------
<?php
namespace my\stuff;
include 'file1.php';
include 'another.php';
use another\thing as MyClass;
$a = new MyClass; // instantiates class "thing" from namespace another
?>
So, in file2.php, once the includes are done, the class name "my\stuff\MyClass" is already
"in use." But, the documentation says that the "conflicting" use statement
"use another\thing as MyClass" should be fine, so long as the other MyClass is declared in
another file (which it is). On this subject the documentation says:
"There is no name conflict, even though the class MyClass exists within the my\stuff namespace,
because the MyClass definition is in a separate file."
Other than the use of autoloading, I can't see any distinction between that example and the
test script here, and it doesn't seem like the autoloading should make any difference.
------------------------------------------------------------------------
[2014-10-23 15:40:45] nikic@php.net
I think all we need to do is drop the if (+ 10 lines) at http://lxr.php.net/xref/PHP_TRUNK/Zend/zend_compile.c#4723,
which checks whether a class with the imported name already exists.
------------------------------------------------------------------------
[2014-10-23 11:39:09] dmitry@php.net
Nikta, it would be great if you can take care about this.
I completely forgot namespaces implementation details.
I'm even not sure, if it can be done without hacks.
------------------------------------------------------------------------
[2014-10-23 10:07:33] nikic@php.net
I agree with Dmitry, it would be best to allow shadowing class definitions with imports in PHP 7.
The current behavior is pretty weird, especially as it depends on the order in which classes are
loaded.
------------------------------------------------------------------------
The remainder of the comments for this report are too long. To view
the rest of the comments, please view the bug report online at
https://bugs.php.net/bug.php?id=66773
--
Edit this bug report at https://bugs.php.net/bug.php?id=66773&edit=1