Re: [RFC/base] Minor GSObjCRuntime usage cleanups

[Richard Frith-Macdonald]

On Tuesday, July 29, 2003, at 12:28 PM, David Ayers wrote:
> I take it, your currently still working on the NSSocket issues of 
> gdnc/NSDistributedNotificationCenter.

Yes ... still thinking about it.  Alexander was also looking at doing a 
patch to make distributed notifications use unix domain socketss by 
default ... I need to coordinate with him.
> In GSFFInvocation/GSFFCallInvocation you changed the NSCAssert1 to an 
> explicit [NSExcaption raise:-...].  I take it that's because Asserts 
> are there to be removable by compiler options, but the exception 
> should stay.  But you also used  sel_get_name in favor of 
> GSNameFromSelector.  Was that merely habit?

Yes ... my mistake.

> Or is there some technical reason for it (avoiding the check for 
> nil...)?  If not, I'd like to commit this patch that uses the 
> GSObjCRuntime functions.  People learn from our patches and I think we 
> should try to be consistent in our conventions and use the API we 
> define.

Yes.

> I'd kind of like GSObjCRuntime to turn into a real abstraction layer 
> of the runtimes.
>
> This would also introduce:
> GS_STATIC_INLINE const char *
> GSClassNameFromObject(id obj)
> {
>  if (obj == 0)
>    return 0;
>  return object_get_class_name(obj);
> }
>
> There are cases I haven't dared yet convert:
> object_is_instance(obj)->GSObjCIsInstance(obj)
> or
> object_get_class(obj)->GSObjCClass(obj)
>
> because of potential differing implementations and requirements (e.g. 
> obj being null, but that has been tested before in most cases).  But 
> maybe we should first decide whether the GSObjCRuntime funcktions 
> should in general be nil tolarant.

I prefer tolerance generally.

> There are also minor details of the testing subdirectory.  It 
> currently still fails to build for me in gstcpport-client:
> Making all for test_tool gstcpport-client...
> Compiling file gstcpport-client.m ...
> gstcpport-client.m: In function `main':
> gstcpport-client.m:27: warning: assignment from incompatible pointer 
> type
> gstcpport-client.m:33: warning: `GSTcpPort' does not respond to 
> `sendBeforeDate:components:from:reserved:'
> Linking test_tool gstcpport-client ...
> shared_obj/gstcpport-client.o(.data.rel+0xb0): In function `main':
> /usr/local/src/cvs/gnustep/core/base/Testing/gstcpport-client.m:24: 
> undefined reference to `__objc_class_name_GSTcpPort'
> collect2: ld returned 1 exit status
> make[1]: *** [shared_obj/gstcpport-client] Error 1
> make: *** [gstcpport-client.all.test-tool.variables] Error 2
>
> I'm sure you can quickly solve it if you have the time, otherwise I'll 
> look into it later.  Richt now I'd like to turn my attention to the 
> Header issue.

ok