Re: Re: com php-src: Fixed bug #67118 DateTime constructor crash with invalid data: NEWS ext/date/php_date.c ext/date/tests/bug67118.phpt

Hi Remi, On Wed, June 4, 2014 20:41, Remi Collet wrote: > Le 04/06/2014 20:25, Remi Collet a écrit : > >> Le 25/04/2014 17:23, Anatol Belski a écrit : >> >>> Commit: c1aa9baf29d1a20fa60c1cef3979a80014f6677b >>> Author: Anatol Belski <ab@php.net> Fri, 25 Apr 2014 >>> 17:23:26 >>> +0200 >>> Parents: 03c703b8bd55679edf30fe17529fab0c2281b01f >>> Branches: PHP-5.4 PHP-5.5 PHP-5.6 master >>> >>> >>> Link: >>> >>> http://git.php.net/?p=php-src.git;a=commitdiff;h=c1aa9baf29d1a20fa60c1 >>> cef3979a80014f6677b >> >> >> It seems this fix breaks Horde (Horde_Imap_Client). >> >> >> Here is the code: >> >> >> >> https://github.com/horde/horde/blob/master/framework/Imap_Client/lib/Ho >> rde/Imap/Client/DateTime.php >> >> If first __construct call fails, this=NULL and other call have no >> effect. >> >> Probably the code seems a bit ugly... but... >> >> >> Any idea ? >> - realloc "this" before php_date_initialize ? >> >> >> Remi. >> >> >> > > Simple reproducer: > > > <?php > > > class Foo extends DateTime { public function __construct($time = null) { > $tz = new DateTimeZone('UTC'); > try { echo "First try\n"; parent::__construct($time, $tz); > return; } catch (Exception $e) { > echo "Second try\n"; parent::__construct($time.'C', $tz); > } > } > } > $date = '12 Sep 2007 15:49:12 UT'; > var_dump(new Foo($date)); > Oh god, why in the world do they have a return there, if they intend to retry with a corrected string :) Yes, the string '12 Sep 2007 15:49:12 UT' is invalid and timelib would complain about the nonexistent timezone, then the script will jump out on that first return. Prior to this fix, no check for php_date_initialize() was done, means the object was created in anyway. And the strange thing is, that timelib returns 0 (means error) but indeed could parse that invalid string. So looking how to unite those two contrasts now. Thanks for the info. Regards Anatol