Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:19782
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Mime-Version: 1.0 (Apple Message framework v734)
In-Reply-To: <436229A2.1000202@iamjochem.com>
References: <B81FCFC9-D920-4CF4-84F2-1DAF46EE287F@synchromedia.co.uk> <20051028140325.BFE8.MIKE.HALL@twistdigital.co.uk> <436229A2.1000202@iamjochem.com>
Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed
Message-ID: <A2DC5EDF-9A1A-4EF1-A0DC-E1CC48912AF7@synchromedia.co.uk>
Content-Transfer-Encoding: 7bit
Date: Fri, 28 Oct 2005 15:41:18 +0100
To: PHP-DEV <internals@lists.php.net>
Subject: Re: [PHP-DEV] No error output?
From: marcus@synchromedia.co.uk (Marcus Bointon)

On 28 Oct 2005, at 14:37, Jochem Maas wrote:

> I have problems like this on a regular basis in big/complex (well  
> by my standards anyway)
> codebases that have heavy php5 oo in them. no errors _anywhere_ if  
> I have a parse error
> or a require_once statement fails, I havbe had the problem since  
> php5beta3 (when I
> started using php5) and have seen it happen on every version of  
> php5 upto 5.0.4 (the highest version
> I have used properly) on windows, debian, redhat and gentoo using  
> apache and apache2.
>
> basically AFAIC there is something wrong _deep_ in the engine but  
> given that I have
> no idea how to prove this or provide any kind of short reproduce  
> script I'm left to live
> with it ... it may of course be a problem with my code, I'm  
> thinking a possible
> reference related segfault issue (and/or maybe __autoload()  
> related). Actually IIRC
> Derick Rethans has posted before about issues he has had where  
> problems were not easily
> (or at all) reproducable in a short script. (excuses to Derick if I  
> misquote/misremember
> what he wrote in the past)

Glad to know I'm not alone! Well, this seems to be a constant state  
for me - even the very simplest script (as in my first post) fails to  
generate errors. It's not specific to my scripts - if I turn E_STRICT  
on and run anything that uses PEAR, there's not a glimmer of an  
error, when I know full well it should look like the tide coming in  
in warning-land.

I've been using 5.1 for several months for the new SOAP stuff. On the  
whole it's been great - and I've been very pleased with the rapid  
attention my bug reports have received (thanks Ilia!) - but this  
error problem is a new thing.

I often resort to using php -l for checking syntax, but at the moment  
I'm having runtime troubles that syntax checks won't help with.

The only way I've been able to deliberately cause this kind of  
behaviour in the past is to store an object (pretty much any class)  
in a session - on restoring the session it will crash apache with no  
errors logged. The only way to recover is to restart apache and  
delete the cookie that relates to that session. However, I'm not  
doing that at present.

FWIW, xdebug's HTML conversion of var_dump and stack traces works  
very nicely and is an indispensable error tracking tool - for  
example, errors in __autoload (which I keep liking and disliking) are  
a pain to fix without it as you don't know where the call came from  
otherwise.

Overall this seems (unfortunately) very likely to be a config  
dependent thing and I don't have a good idea of the cause (hence my  
config file rebuild), so I'm not keen on submitting a bug report as  
yet. Should I anyway?

Marcus
-- 
Marcus Bointon
Synchromedia Limited: Putting you in the picture
marcus@synchromedia.co.uk | http://www.synchromedia.co.uk