🚀 go-pugleaf

maxtal@extro.ucc.su.OZ.AU (John MAX Skaller) writes:

>fjh@munta.cs.mu.OZ.AU (Fergus James HENDERSON) writes:
>>maxtal@extro.ucc.su.OZ.AU (John MAX Skaller) writes:
>>
>>>pat@frumious.uucp (Patrick Smith) writes:
>>>>   p == q   =>   ptrcmp(p,q) == 0
>>>
>>>	I'm NOT agreeing with this at present. I reject any relationship
>>>between p ? q and ptrcmp. On some machines any such relationship
>>>might make ptrcmp unimplementable, and thus defeat the proposal completely.
>>
>>I remain unconvinced that such a relationship is unimplementable on
>>some machines.
>
>	On the 486 it is not in general possible to implement
>the comparison p==q for arbitrary logical addresses. If p and q
>are aliases only the operating system kernel could implement 
>the comparison, and if the OS happened not to support this,
>p==q could not be implemented.

*In general* this is true, but we are not talking about the general case.
All that is required is that it be true for conformant C++ programs.
This does not require checking whether segment aliases exist, it just requires
ensuring that they won't occur for conformant C++ programs.

>>>It is only Jims suggestion that ptrcmp might always return 0 that
>>>makes it possible to make its existence mandatory.
>>
>>A ptrcmp that always returned 0 would be useless, IMHO.
>
>	Not at all. My sort will still order the pointers
>into the total order, which in this case is anything.
>
>	This is entirely different from the current situation
>where a sort based on the existing implementation defined <
>on pointers might not terminate.

Yes, you are correct. I retract that statement.

-- 
Fergus Henderson             fjh@munta.cs.mu.OZ.AU      
This .signature virus is a self-referential statement that is true - but 
you will only be able to consistently believe it if you copy it to your own
.signature file!