Re: io_close proposal

[Marcus Brinkmann]

On Thu, May 16, 2002 at 08:09:04AM -0400, Marcus Brinkmann wrote:
> To me that seems like we can just attempt io_close() once, and when we
> are interrupted, destroy the reply port and return to the user.

Mmh.  I was probably too optimistic.  The state of the fd is undefined,
but that doesn't mean that it isn't well defined, either still open and
valid, or fully closed.  If it has to be well defined, I agree with
your analysis.  It also seems as if SA_RESTART also affects close()...

well, it was a try, wasn't it? :)

This is a bit annoying, but I don't see an alternative to moving the
send right in a message, and I tried a few.  Well, the alternative I
have not completely elided is the "check exact number of extinct send
rights", which matches nice with the current interruptible RPC handling,
but is otherwise note very elegant.

Need to ponder this more, this message is just to step back from my
silly interpretation of close() in POSIX...

Thanks,
Marcus