Re: Non-null pointers in C/C++?

Angel Tsankov wrote:
> James Kuyper wrote:
>> Angel Tsankov wrote:
>> ...
>>> If you cast T const* to T*, pass the result to a function that takes
>>> T*, and that function changes the pointed-to value, do you get
>>> undefined behaviour?
> 
>> That depends upon whether or not the object pointed at was defined as
>> const. The definition of the pointer is irrelevant, only the
>> definition of the pointed-at object matters.
> [...]
> 
> What I meant is passing a pointer to const object to a function that
expects 
> a pointer to non-const one and changing the object via the passed-in 
> pointer. Taking into account this clarification, I would take your
answer as 
> "yes". This leaves one of my previous questions unanswered:
> 
>> If your answer is "yes", then why do you bother so much about NULL 
>> sneaking into an unnullable pointer and causing undefined behaviour,
too 
>> (which should very, very, really very rarely happen as the explicit
cast 
>> from a nullable pointer to an unnullable one would remind you that you 
>> have to check befoure you cast)? 

I left that one unanswered, because the "you" it refers to appears to 
have been Gordon Burditt (it was harder to figure that out than it 
should have been,because you didn't use any attribution lines in that 
message).

I think that if you had wanted to im****t the C++ concept of references, 
that would have had some advantages, possibly enough to make it a good 
idea. A C++ reference can be implemented in a way that makes it seem 
just like syntactic sugar for a const pointer. However, a consequence of 
that syntactic sugar, combined the special rules that apply to 
references, is that code which might otherwise appear capable to making 
that pointer null is either a syntax error or a constraint violation. 
The underlying pointer (or equivalent) could be invalid, if the object 
it originally referred to having ended it's lifetime, but it can't be
null.

However, the rules you've chosen for your unnullable pointers are too 
lax, and make it possible for it to be nulled. I find that concept much 
less useful.
-- 
comp.lang.c.moderated - moderation address: clcm@[EMAIL PROTECTED]
 -- you must
have an appropriate newsgroups line in your header for your mail to be
seen,
or the newsgroup name in square brackets in the subject line.  Sorry.