Svante Signell, le Thu 18 Oct 2012 14:55:33 +0200, a écrit : > > > POLLNVAL > > > The specified fd value is invalid. This flag is only valid in > > > the revents member; it shall ignored in the events member. > > > > > > Specifically, my interpretation of POLLNVAL is that it should be > > > reported for each invalid fd given as input. > > > > Ah. You should have pointed to that from the start :) > > The changes I made was 15 months ago, and I did not remember by heart > when sending my code, sorry.
That's why taking notes is always a good thing to do. > > > And my interpretation of POLLERR is that it should be reported to > > > _every_ fd (device or stream) given as input. > > > > > > And for these cases the return code should be the number of good fds as > > > well as bad fds, since the revents member is nonzero. (I did not come > > > that far in my changes by then, but it could easily be done with today's > > > knowledge! There are also other case to take care of like: POLLWRBAND > > > etc) > > > > Indeed. That's also what was completely missing in your previous > > description. It's indeed a bug. And it has to be fixed in poll only, not > > in select. > > I can try to make an update to that code if you are interested ;) That can be useful, yes. To have any chance of seeing it applied, you'll however have to take care of (like any kind of patch, actually): - avoid code refactoring which does not actually change the behavior, or submit it separately - avoid affecting the select() behavior: unlike poll, that one *has* to return EBADF as soon as a descriptor is seen as bogus. So that'll probably mean adding a parameter to _hurd_select which selects between both behaviors. Since that function is actually public, that actually means introducing another function with the parameter, and making _hurd_select just call it with choosing the "select" behavior, while poll() will choose the "poll" behavior. Samuel
