Re: Re: cvs: php4 /pear/Crypt CBC.php

> Colin, in general it's not a good idea to pass mode in raiseError, > because it takes control away from the user. If I have a callback > function that I want to use for all error handling, it will never be > called when you override my default mode in the raiseError call. Fair enough. :) > If you call the PEAR constructor with 'Crypt_CBC_Error' as the first > parameter (or no parameter, you don't really need to override > PEAR_Error unless you really want a different prefix), you only need > to specify the error message in your raiseError call: > > function Crypt_CBC(...) > { > $this->PEAR(); > } > ... > return $this->raiseError('no initialization vector'); Okay I changed my code to something like this: function Crypt_CBC($key, $cipher) { $this->PEAR(); if (!$key) { return $this->raiseError('no key'); } ... } But when, I run it: $c = new Crypt_CBC('', 'des'); $c is not a PEAR_Error object. That's what I'm trying to do. :) > > > I don't need to do something like this, do I? > > > > > > if (something bad) { > > > $this = PEAR::raiseError(...) > > > return $this; > > > } > > Overwriting the $this variable is not a good idea. :-) Agreed. :) However, this seems to be the only way to get the Crypt_CBC constructor to not create a Crypt_CBC object. Sorry if I'm being difficult. :) - Colin P.S. When will someone fix the "list-post:" header for the DEV and CVS lists? Or is that something generated by the news.php.net NNTP server? I'm still trying to filter Tomas' emails properly. :)