Invalid offsetof from non-POD type

#1 This warning was introduced with gcc3.

There is no way to turn them off, short of changing the gcc3 source code or using the -w flag to turn of _all_ warnings.

The problem is the way torque sets up bindings with the scripting language. It uses the c offsetof macro on the fields of console objects, to compute a relative offset that i think is later used when the field needs to be read/write by the scripting language.

However, a c++-class is non-POD ("Plain Old Data"), and the c++ standard apparently says that you should not use offsetof on such structures.

Despite the hundreds of lines of warings, the resulting executable still seems to work fine. I think the warning may be overly broad. I emailed the person who added this code to gcc but I received no response.

I don't think changing this in torque will be a simple task. The compiler suggests "use pointer to member instead". To do this we would need to build a table of all console objects and fetch the appropriate one when a field is accessed. The beautity of the current system is that the offsets are computed once in a static method, and can be used for any object.

Soon, I'm going to email the gcc mailing list about it. What I want to find is a case where the warning fires, and the resulting code is invalid. If this isn't happening in torque, then I don't see why we should change the code. Instead, maybe gcc should have a flag to disable the warning.

Edit: just posted to the mailing list about this...we'll see what they say.

#2 I checked in a workaround for this that squelches the warning. Let me know if it causes problems.

#3 Wow ! This will be really appreciated ! Thanks much for this John