Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:23207
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Message-ID: <44602183.8010309@bfoz.net>
Date: Mon, 08 May 2006 21:58:43 -0700
User-Agent: Thunderbird 1.5.0.2 (X11/20060507)
MIME-Version: 1.0
To: internals <internals@lists.php.net>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Subject: INIT_CLASS_ENTRY and char*
From: bfoz@bfoz.net (Brandon Fosdick)

I think I found a bug in the INIT_OVERLOADED_CLASS_ENTRY_EX macro. At least, I think its a bug, somebody else might think its a feature. :)

If you do something like INIT_CLASS_ENTRY(ce, "MyClass", ...) everything works fine. However, if you have something like

void register_class(char* name, ...)
{
   ...
   INIT_CLASS_ENTRY(ce, name, ...);
   ...
}

things don't work so well. When I run this through gdb and break right after the macro, ce.name = "MyClass" as expected. But ce.name_length = 3, which is not quite right.

It looks to me like the problem is in the 3rd line of the INIT_OVERLOADED_CLASS_ENTRY_EX macro...

---
#define INIT_OVERLOADED_CLASS_ENTRY_EX(class_container, class_name, functions, handle_fcall, handle_propget, handle_propset, handle_propunset, handle_propisset) \
	{															\
		class_container.name = strdup(class_name);				\
		class_container.name_length = sizeof(class_name) - 1;	\
---

...where sizeof() is used instead of strlen(). When class_name is a variable, sizeof() dutifully returns the size of the variable instead of the string length. Obviously this works, and provides a bit of a speed boost, when using literal strings, but it doesn't work so well for char*'s. 

Is this desired behavior or a bug?