gnustep-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

objc runtime problem


From: Richard Frith-Macdonald
Subject: objc runtime problem
Date: Thu, 26 Jun 2003 14:23:35 +0100

There is a problem with the forwarding mechanism we have at present...

When a message is sent to an object, and there is no method found to handle it, the runtime calls the function registered as _objc_msg_forward to ask for the method implementation to use. We use this function to determine and return the correct method for forwarding to work through a proxy, and to make a note of the information about argument and return types, needed to set up the content of NSInvocation obejcts etc.
Mostly this works fine.

However, there are two cases where it does not work -
1. Where there is more than one type signature for the selector name ...
(eg. -(void)foo; and -(struct abc)foo;) in which case we may not get the correct type information. 2. Where the receiver is an object in a remote process, and there is no type information associated with the message locally.

It is currently impossible for these cases to work, because the _objc_msg_forward() function takes a single argument (the selector to look up), but the type information which should be associated with the selector depends on the receiver, and the compiler may have provided an untyped selector, or one with the wrong type info for the receiver.

If sendmsg.c was modified to pass two arguments to the function, (the second being the receiver), the function would be able to ask the receiver for the type information and return a forwarding function which would work for that receiver.

Is there any reason not to modify the runtime to work this way?





reply via email to

[Prev in Thread] Current Thread [Next in Thread]