Vincent Rivière, 25.02.2013 23:05:12:
So I'm relieved that I didn't cause a regression there.
No one said that. It's just me thinking it would also have been fixed, which seemed to be the case, but only because your first version was wrong. I'm really annoyed by this bug since ever!
I guess the first thing to do would be to separate CRMOD for input and output. Maybe some of the 0xf000-bits in sg_lflag could be used for output-CRMOD.
But that can wait. -- Helmut Karlowski