Mailinglist Archive: opensuse (4570 mails)

[Randall R Schulz <rschulz@xxxxxxxxx>]

Steven,

On Saturday 12 November 2005 17:17, Steven T. Hatton wrote:
> ...
>
> Java's answer to memory leaks is garbage collection. C++'s answer is
> RAII. RAII requires you to think about what you are doing at a more
> fundamental level.  That may take more work, but the end result is
> typically much better design.

I love answers of the form: If you were just omnicient, it wouldn't be a 
problem.

GC is not an answer to "leaks" (which are bugs), it's a system of 
automatic storage management. Storage management is ubitquitous in 
programming. It's incumbent upon a language runtime to provide good 
support for it. C++ says "I'll give you a built-in, uniform mechanism 
to dynamically allocate storage, but you're on your own for keeping 
track of every single allocation you make and if you don't give it back 
to me, your program will slowly suffocate or take over the host on 
which it's running." What kind of attitude is that for a language 
designer to take toward the users?

C++ storage leaks are Murphy's Law in operation. The design of the 
system affords a failure mode, so the occurrence of that failure mode 
is predictable. Java designs the failure mode out at the language 
level, hence it does not occur.

And just how does RAII and auto-pointers deal with cyclic structures? 


> You may want to look at _The Design and Evolution of C++_, by Bjarne
> Stroustrup.
>
> Steven


Randall Schulz