Re: Memory Leaks in RPG (was: Is RPG 'DEAD")

[David Gibbs <david@xxxxxxxxxxxx>]

Scott Klement wrote:
> It's theoretically possible to put a pointer in a user space and share 
> it with another job, but doing so is dangerous for this very reason... 
> if the job that allocated the memory ends, the pointer will point to 
> unreserved memory, and produce the same sorts of "unpredictable results" 
> that you'd have with a buffer overflow.

Even if you handed a pointer around from job to job, would different 
jobs even be able to access the same memory?  Seems to me that the OS 
should protect against memory access across job domains (unless the 
memory is allocated in an approved 'shared memory' method).

I'm not at a system right now ... but I'm wondering if, when you 
retrieve a pointer to the same user space in two different jobs, you get 
the same pointer value?

> b) Or, use the shared memory APIs (shmget, shmat, shmdt, etc) to reserve 
> shared memory.

Hmmm ... this makes me wonder if the shared memory api's use user spaces 
under the covers?

Weird as this sounds ... even though I was annoyed at the impetus of 
this thread, it's gotten darn interesting.

david