IRC, freenode, #hurd, 2011-08-11

< youpi> in which error cases a reply port will actually have been consumed
  by mach_msg ?
< youpi> it seems at least MACH_SEND_NOTIFY_IN_PROGRESS do?
< braunr>
< braunr> "These return codes imply that the message was returned to the
  caller with a pseudo-receive operation: "
< braunr> isn't it what you're looking for ?
< youpi> well, it's hard to tell from the name
< youpi> I don't know what "pseudo-receiv operation" means
< braunr> it's described below
< youpi> ew
< braunr> it looks close enough to a normal receive to assume it consumes
  the reply port
< youpi> so it's even more complex than what I thought
< youpi> well, no, it returns the right
< youpi> actually the error I'm getting is MACH_RCV_INVALID_NAME
< youpi> which I guess means the sending part succeeded
< youpi> the case at stake is proc/mgt.c: S_proc_exception_raise()
< youpi> when the proc_exception_raise() forward fails
< youpi> currently we always return 0, but if proc_exception_raise()
  actually managed to send the message, the reply port was consumed and
  MIG_NO_REPLY should be returned instead