ObjC fixes for --enable-mapped-location

[Zack Weinberg]

Ziemowit Laski <zlaski@apple.com> writes:

>> Here is a revised patch which includes all of the revisions you asked
>> for.  I also kluged a test, proto-lossage-4.m, which was getting
>> excess errors on 64-bit targets -- it is not clear to me what that
>> test is playing at, sending messages to int, but the most expedient
>> fix was to make it long int instead (which is the same width as a
>> pointer on all supported targets).
>
> The test is checking for graceful handling of messages being sent to
> nonsensical types.  By changing the test the way you did, I think you
> are just masking the underlying problem.

No, I don't think it is.  The "invalid receiver type" diagnostic is
still being issued.  All this change does is suppress the diagnostic

 warning: cast to pointer from integer of different size

which would appear on 64- but not 32-bit systems.

> It might even be related to the other mysterious LP64 failures
> people have reported...

I wouldn't be surprised to find places where the objc runtime isn't
64-bit clean, but I don't think this is one of them.

zw