Re: [PATCH] ntdll: Fix race condition with fd_cache when duplicating handle.

[Alexandre Julliard]

Daniel Lehman  writes:
> > > do you mean if the current process received a file handle from another
> > > process and wanted to close the handle on the remote end?
> >
> > Yes, that's why there's a source process parameter.
>
> yeah, my fix doesn't cover receiving a file handle (or sending on
> behalf of another process).  there isn't an existing way (that i see)
> to tell the other process to remove the cached fd
>
> i was hoping to at least fix the race in this case where the current
> process is sending

Note that there can be other reasons for the server call to fail, so we
can't just remove the handle from the cache unconditionally. This will
need more thought.