On 2000-Feb-23 23:46:35 +1100, Brad Knowles <[EMAIL PROTECTED]> wrote:
> However, this has brought up an interesting question. It is my
>understanding that X/Open requires that this union be defined within
>the client source code
Both Solaris and the Single UNIX Specification, Version 2 explicitly
state this.
> Is this changed in -CURRENT?
Not at present.
> If not, are there any plans to
>change it in -CURRENT? Or has the X/Open standard changed, and
>FreeBSD is following the new standard in this area?
I can't comment on this. There doesn't seem to be a high level of
interest in the SysV semaphore handling.
> I'm also curious to know why there is a difference in how this
>object is defined under Linux and FreeBSD.
There are two differences:
- The type of "array": SUSv2 and Linux specify "unsigned short int *",
Solaris uses "ushort *" and FreeBSD uses "u_short *". The underlying
type should be the same. (Though it does bring up another `bug' in
our implementation - SUSv2 states <sys/sem.h> should be idempotent,
but we require <sys/types.h> to be explicitly included first).
- Linux includes an "__buf" field. This field (and the associated
IPC_INFO) don't appear in either SUSv2 or Solaris - I suspect it is
a Linux extension to make ipcs(1) cleaner.
>However, I'm also interested as to who is "right" on this issue, and
>more importantly, why they are "right".
It looks very much like FreeBSD is in the wrong here.
Peter
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
