1 Mar
2005
1 Mar
'05
4:01 a.m.
On Mon, Feb 28, 2005 at 10:49:17PM -0500, James Y Knight wrote:
On Feb 28, 2005, at 10:40 PM, James Y Knight wrote:
I'm glad we all agree as to what the new behavior should be, but you're incorrect about the extent of the old brokenness. Simple tests prove that POLLIN was not forced to be true always:
Which makes sense given that the following says "=" not "|=". - mask = POLLIN | POLLRDNORM; - if (PIPE_EMPTY(*inode)) - mask = POLLOUT | POLLWRNORM;
Ok that explains why you didn't get "r and w" set at the same time, so it wasn't working by luck. It was only the 2.6.11-rc changes that broke it and that could return "r and w" set at the same time. And now they're fixed again.