[Twisted-Python] linux kernel 2.6.11-rc broke twisted process pipes

[Andrea Arcangeli]

On Mon, Feb 28, 2005 at 10:40:00PM -0500, James Y Knight wrote:
> Right. Twisted treats all of POLLERR/HUP/NVAL the same, anyways.

Ok.

> Nope -- Notice that rpipe and wpipe are backwards for the read and 
> write fds. "rpipe" on the write fd won't ever have any data in it. 
> (Except of course that pipes in BSD are bidirectional!). In BSD, 

I didn't know it was bidirectional, I thought the peer pointer pointed
to the other side of the pipe like in linux, and not the other direction
of the same side.

> Yes, it will get called, because linux returns (returned?) 
> POLLERR|POLLIN when the other side is closed. pollreactor doesn't 

It return_ed_ POLLERR|POLLIN, correct.

Now it returns only POLLERR for a wronly fd and only POLLHUP for a
rdonly fd which seems even more correct since it will prevent the
suprious doRead call with pollreactor.

> assume the connection was lost immediately when POLLERR/etc is set, if 
> POLLIN is also set, because POLLIN|POLLHUP means there is more data 
> available to be read from the kernel buffer that was sent before the 
> connection closed. And, since normal transport implementations of 

Yep, except that's a wronly fd so calling doRead can only detect a
disconnect that is already detected thanks to the pollerr.

> doRead actually do try to read, if it was lying about POLLIN, that's 
> okay: the read syscall will fail, and CONNECTION_LOST will be returned 
> from doRead.

Ok. So I'll take the exception path. The exception should be ValueError:
filedescriptor out of range in select(). Anyway with the new patched
kernel I'm running and with latest 2.6.11 this shouldn't happen anymore
because POLLIN isn't set ;).

Thanks a lot for all the help, this clarified many things.