On vacation and at a loss

Uncreative

Calendar

Back May '12
Mon Tue Wed Thu Fri Sat Sun
  1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31      

Quicksearch

Archives

Categories



All categories

Get Jimdo!

Get a free website with professional content management features at Jimdo.com!

On vacation and at a loss

Sunday, August 20. 2006

The good news is: I'm on vacation! Yay :-D Well, actually vacation is almost over... I have until Wednesday to become totally relaxed and yet there's so much left to do... need to toss out lots of old stuff, scratch the rust off an old Singer sawing machine rack to refit it as a table and things like that. At least in some respect my vacation has been a little bit productive - since last Wednesday I am now officially a Zend Certified Engineer (most probably the last one to ever do the PHP4 certification... well, I still had this voucher that would have gone void on August 31st - guess when the PHP5 certification is due to hit the road...). Of course, I also wanted to do lots of coding during my vacation - what better method to relax than a good few thousand lines of code, huh? No, I have no life... or do I? ;-) Anyway, the coding part was kind of put back when I hit a very nasty bug that I couldn't find yet. Worst of all, I have no idea if it's the stuff I wrote (more likely), parts of the Zend Framework (ruled out already), PHP itself (less likely) or Apache (likeliness hits rock bottom). In the unlikely case you should be interested in my pain, read on (this is a lenghty and rather uninteresting one - you have been warned!!).

A while ago, I wrote this ActiveRecord implementation called CWAC_ActiveRecord to satisfy my need for a high-performance and easy-to-use ORM layer directly on top of PDO that I could use in my Zend Framework powered applications. It's the first package I ever wrote using a test driven development approach, and all unit tests are working okay. Thus I thought to have coded a pretty well-working and seemingly stable thing. I also wrote an extension of the Zend_Controller_Action class to satisfy my need for auto-templating. After some small difficulties and finally tossing the "output on destructor call"-approach, this worked extremely well, too. To put this to the test, I wrote a very small controller script for the Zend Framework containing an action that would display the form and, if the form data validated okay, insert a new record (user account) into the database and then redirect to a welcome page. Here's the action method:
  1.     public function createAccountAction()
  2.     {
  3.         error_log('Call to createAccountAction');
  4.         $this->_applicationView->title = 'Create a new account';
  5.         if (!empty($_POST)) {
  6.             $post = new Zend_Filter_Input($_POST);
  7.             $values = array('alias' => $post->getRaw('alias'),
  8.                             'email' => $post->getRaw('email'),
  9.                             'pw'    => $post->getRaw('pw'),
  10.                             'pw2'   => $post->getRaw('pw2'));
  11.             $errors = array();
  12.  
  13.             if (!$post->testAlnum('alias')) {
  14.                 $errors['alias'][] = 'This field may contain only alphanumeric characters (letters and numbers)';
  15.             }
  16.             if (!$post->testRegex('email', '/^[a-zA-Z]([.]?([[:alnum:]<u>-]+)*)?@([[:alnum:]\-</u>]+\.)+[a-zA-Z]{2,4}$/')) {
  17.                 $errors['email'][] = 'This is not a valid email address. Please check the spelling.';
  18.             }
  19.             $pwerror = 'Password must be at least 5 characters long and may contain '.
  20.                        'only letters, numbers and one or more of the following characters: '.
  21.                        '_ ! . $ # @ ? -';
  22.             if (!$post->testRegex('pw', '/^[A-Z0-9a-z_\.\!\$#@\/\?-]{5,}$/')) {
  23.                 $errors['pw'][] = $pwerror;
  24.             }
  25.             if (!$post->testRegex('pw2', '/^[A-Z0-9a-z_\.\!\$#@\/\?-]{5,}$/')) {
  26.                 $errors['pw2'][] = $pwerror;
  27.             }
  28.             if ($values['pw'] != $values['pw2']) {
  29.                 $errors['pw2'][] = 'The passwords do not match. Please try again.';
  30.             }
  31.             if (count($errors) == 0) {
  32.                 error_log('Creating ActiveRecord object');
  33.                 $user = new Usr_user();
  34.                 $user->alias = $values['alias'];
  35.                 $user->email = $values['email'];
  36.                 $user->pw = $values['pw'];
  37.  
  38.                 try {
  39.                     error_log('Saving...');
  40.                     $stmt = $user->save();
  41.                     error_log('Unsetting statement object');
  42.                     unset($stmt);
  43.                     $_SESSION['user'] = $user->toArray();
  44.                     error_log('Going to redirect...');
  45.                     $this->_redirect('http://dev.local/membership/welcome/');
  46.                     error_log('Still there?');
  47.                 } catch (Exception $ex) {
  48.                     echo "<h1>Error</h1><pre>".$ex->getMessage()."</pre>";
  49.                 }
  50.             } else {
  51.                 $this->_actionView->errors = $errors;
  52.                 $this->_actionView->values = $values;
  53.                 $this->render();
  54.             }
  55.         } else {
  56.             $this->render();
  57.         }
  58.     }
When executing the POST request, this is the result as seen in the Apache error log: 
[Sun Aug 20 00:39:19 2006] [notice] child pid 18870 exit signal Segmentation fault (11)
The PHP error log shows nothing wrong:
[20-Aug-2006 00:39:19] Call to createAccountAction
[20-Aug-2006 00:39:19] Creating ActiveRecord object
[20-Aug-2006 00:39:19] Goodbye, and thanks for all the fish!
[20-Aug-2006 00:39:19] Saving...
[20-Aug-2006 00:39:19] Executing statement: INSERT INTO usr_user (alias, pw, email) VALUES (:alias, :pw, :email)
[20-Aug-2006 00:39:19] Preparing new statement
[20-Aug-2006 00:39:19] Executing with bindings: Array
(
    [:alias] => bla
    [:pw] => 12345
    [:email] => hink@honk.de
)

[20-Aug-2006 00:39:19] Unsetting statement object
[20-Aug-2006 00:39:19] Going to redirect...
[20-Aug-2006 00:39:19] Arrived at redirect.
[20-Aug-2006 00:39:19] Done header() call... EXIT!
[20-Aug-2006 00:39:19] PHP shutdown
[20-Aug-2006 00:39:19] Destructing ActiveRecord object.
As can be seen, I've added a few extra error_log() calls in the ActiveRecord class as well as in the redirect() method of the Zend_Controller_Action class, to trace the beast. So it appears as if PHP executes the request without any error and then exits cleanly - but then the Apache dies before sending anything back to the browser (including the desired redirect header). While trying to track things down, I've only been successful in preventing the Apache crash when commenting out the call to ActiveRecord's save() method. When nothing is written to the database, the crash does not occur. Huh? Okay, trying to break things down a bit further and taking everything out of the equation beside ActiveRecord:
  1. require_once 'CWAC/ActiveRecord.php';
  2.  
  3. #### Configuration for CWAC_ActiveRecord
  4. $ARConfig = array('connections' => array(
  5.                                         'default' => array(
  6.                                                         'dsn' => 'mysql:host=localhost;dbname=test',
  7.                                                         'user' => 'root',
  8.                                                         'password' => null
  9.                                                         )
  10.                                         ),
  11.                 'model_path' => dirname(<u>_FILE_</u>).'../application/models/'
  12.                 );
  13. CWAC_ActiveRecord::setConfig($ARConfig);
  14.  
  15. class Usr_user extends CWAC_ActiveRecord
  16. {
  17.     public $id;
  18.     public $alias;
  19.     public $pw;
  20.     public $first_name;
  21.     public $last_name;
  22.     public $email;
  23.     public $active;
  24.     public $created_at;
  25.     public $updated_at;
  26. }
  27.  
  28. error_log('Create instance');
  29. $usr = new Usr_user();
  30. $usr->alias = 'bla';
  31. $usr->pw = 'bli';
  32. $usr->email = 'blubb';
  33. $usr->first_name="Klaus";
  34. $usr->last_name="Piepenbrinck";
  35.  
  36. error_log('Save...');
  37. $usr->save();
  38.  
  39. error_log('Redirect...');
  40. header("Location: http://blog.wolff-hamburg.de/");
Points of failure eliminated: CWAC_Controller_Action, all of the Zend Framework, and no data coming in via POST. The script is completely self-contained except for the CWAC_ActiveRecord include. Still, the same behaviour. So it would seem the Apache dislikes my ActiveRecord component for some strange reason. Even more strange: The crash doesn't happen if the request is the first request after starting Apache. The first request always works, no crashes, all is fine. But on all subsequent requests Apache crashes reliably. Worst of all, I haven't been able to reproduce this on my public development server (Gentoo) - at least with the small test script, haven't tested with the full Zend Framework environment yet. The problem seems to occur only on my home development box (Ubuntu Dapper, with the latest stable XAMPP providing both Apache and PHP 5.1.4). So now I have no idea if this is really a problem with ActiveRecord that just happens only in specific environments or if the problem is specific to just my environment that's totally f...ed up. Maybe I should just forget about my local box for now and develop on the Gentoo box to finally get something done instead of trying to hunt down environment-related esoteric bugs that I'll never find anyway... but still, this bugs me quite a bit. To sum all this up, in regard to "What I did on my holidays" it can be concluded that I've lived in "Interesting Times" (which, for the adept geek, should be a clear indicator on what I've been reading during my vacation) :-D
Posted by Markus in PHP, Programming at 00:39 | Comments (7) | Trackback (1)

Trackbacks
Trackback specific URI for this entry

Bug found (kinda) - bye bye prepared statements
Last night I blogged about my frustration with a bug in my ActiveRecord implementation that caused Apache to segfault after calling the save() method (which, strangely enough, never occured on the first request after a fresh Apache restart, but on all sub
Weblog: Uncreative
Tracked: Aug 21, 02:10

Comments
Display comments as (Linear | Threaded)

I have seen crashes like that in my Apache log when I started experimenting with PDO. I tried converting an existing application from mysql_* functions to PDO. This on PHP 5.1.4 on Mac OS X. After converting only a few lines, and doing nothing that isn't in the manual, I immediately starting getting segfaults. So I decided to not pursue this any further and use PEAR::MDB2 (with mysqli extension) instead. Haven't had any more segfault problems. My impression was that, at least on Mac OS X, PDO is quite buggy.
#1 Jan Pieter Kunst on 2006-08-20 09:12 (Reply)
Hi Markus, I just stumbled on this message from your error log: [20-Aug-2006 00:39:19] Goodbye, and thanks for all the fish! From your code I can only guess that this message is logged in the constructor of your Usr_user class or in a method that is called by the constructor. A Goodbye message at this state of the process sounds a bit weird, so maybe it is worth a check. HTH, Ralf
#2 Ralf (Homepage) on 2006-08-20 09:26 (Reply)
Thanks, but this message is just a remainder of my first attempts to track this down... it's directly in the action method, not the constructor, at the top of the try block, and it's meant to indicate that from here on, it's all going downhill :-D
#2.1 Markus (Homepage) on 2006-08-20 15:27 (Reply)
Things you do in __destruct sometimes make Apache crash. Especially if you are dealing with database-stuff there. Try commenting out your destructors function-bodies.
#3 Jan on 2006-08-20 10:03 (Reply)
Tried. No effect. The code that would theoretically do anything with a remaining statement object isn't even triggered in this case anyway.
#3.1 Markus (Homepage) on 2006-08-20 15:25 (Reply)
Looks like you would benefit from refactoring the action into a set of multiple smaller methods. This would make your code both easier to test as well as improve the readability. First though, a test driven approach to debugging writing tests to hunt down the bug(s) could be worthwhile.
#4 Mattis on 2006-08-20 14:35 (Reply)
Having tested on my Gentoo dev box together with the full Zend Framework environment now, I can happily (?) say that the bug now appears on my dev box as well - or at least I'm getting the same behaviour, the segfault message doesn't even make it into the Apache log.... FRUSTRATION
#5 Markus (Homepage) on 2006-08-20 16:44 (Reply)

Add Comment

Enclosing asterisks marks text as bold (*word*), underscore are made via _word_.
Standard emoticons like :-) and ;-) are converted to images.
E-Mail addresses will not be displayed and will only be used for E-Mail notifications.

To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly.
CAPTCHA
 
 

Competition entry by David Cummins powered by Serendipity v1.0

Recent Comments

Dave Hall about Finding a date for the next Unconference
Thu, 12.02.2009 13:12
I am not sure I will be able t o make it on either weekend. Go for either weekend and I wi ll see if I am in the ar [...]
Ralfs Zend Framework und PHP Blog about A pragmatic approach to rights management
Tue, 20.01.2009 18:32
In den letzten Wochen sind in diversen englischsprachigen Bl ogs sehr interessante und teil weise recht ausfhrliche [...]
Lukas about A pragmatic approach to rights management
Tue, 13.01.2009 13:40
Well I am mostly at fault for the "complexity". I actually u sed pretty much all the advanc ed features in real worl [...]
Daniel O'Connor about A pragmatic approach to rights management
Mon, 05.01.2009 10:24
I agree completely: KISS. S ince rights management can be complicated, I find it works b est when you push it dow [...]
Hodicska Gergely about See you at the International PHP Conference 2008
Mon, 03.11.2008 23:58
Hi! Are you planing to share t he slides of your talk? TIA!
Ralfs Zend Framework und PHP Blog about See you at the International PHP Conference 2008
Sat, 25.10.2008 10:13
Am nchsten Montag (27.10.) beg innt die 14. Internationale PH P Konferenz in Mainz und luft bis zum Freitag (31.10.) [...]

Entry's Links

Referring links

Syndicate This Blog

Content license

Creative Commons License - Some Rights Reserved
Original content in this work is licensed under a Creative Commons License

Powered by

Serendipity PHP Weblog