Patch to pass file descriptors

[Conrad Scott]

"David Euresti" <davie@MIT.EDU> wrote:
>         There's the option of keeping full BSD semantics if the
cygserver
> is running.  I like that.  I can require users to run the cygserver.
I
> don't mind. [snip]
>         The question is, can you incorporate my code and then you
guys
> deal with the case when the cygserver isn't running? [snip]
>         I'm sure other people will want full BSD semantics also so
why not
> allow them to have it if the cygserver is running.

This seems to be the "standard" approach for this sort of issue (if
one example suffices to set a standard): the fhandler_tty code uses
cygserver to dup handles if available, otherwise falls back to doing
it locally; see the fhandler_tty_slave::open method in
"fhandler_tty.cc". So I don't see a problem with that.

>         Unfortunately the whole net.cc has been refactored so I'll
have to
> regenerate my patch.

Unless anyone has any objection, if you send this to the list, I'll
put it into the cygwin_daemon branch. It's not a complete or final
solution (as you've noted), and perhaps you ought to follow the
example, from the fhandler_tty.cc code, of putting some conditions
around calling cygserver, leaving a blank spot to be filled in with
the non-cygserver code. When/if such non-cygserver code appears on
HEAD I can merge it into the branch by slotting it into those blanks.

// Conrad