Re: segfault with iterators

From: Date: Tue, 17 Feb 2004 19:20:07 +0000
Subject: Re: segfault with iterators
References: 1  Groups: php.internals 
Request: Send a blank email to internals+get-7928@lists.php.net to get a copy of this message
Hello, On Tue, 17 Feb 2004 08:17:54 -0500 "Rob Richards" <rrichards@ctindustries.net> wrote: > I ran into this with simplexml but am able to reproduce with userland > classes. > > The segfault only occurs when the property in question does not exist, > no __get and __set methods implemented (if either is implemented then > no segfault), and is not being called off the initial object in the > foreach. i.e. - $obj->valid()->invalid will segfault while > $obj->invalid doesnt. I got the same thing this morning and wondering why as well. I'm not sure it's due to the iterator but more to the (de)references. At some point ZE2 destroys the object (afair or see just before returning the object). Find below 2 scripts to reproduce the problem. In both the public property is not required. Only to show that is not related to the existence or non existence of a property. Works: ----- <?php class Test { public $a = array(1,2,3,4,5); // if removed, the notice works function c() { return new Test(); } } $obj = new Test(); $b = $obj->c(); foreach ($b->a as $value) { // Segafults print "Test\n"; } ?> Segfault: -------- <?php class Test { public $a = array(1,2,3,4,5); // removed, crash too function c() { return new Test(); } } $obj = new Test(); foreach ($obj->c()->a as $value) { // Segafults print "Test\n"; } ?> Sidenote, if someone has some tips to get the ZE2 macros working in gdb (ie to work with all the the TSRMLS_C*, T, and other macros with explicit names ;) hth pierre

« previous php.internals (#7928) next »