> > moving 'init' doesn't fix the bug, it just makes it harder to trigger.
> > 
> 	But if I trigger it every time I update my -current, doesn't it make
> 	sense to put a fix into the init install & append a note to a pr that
> 	the fix will go away when the real problem is fixed.
> 	(Sorry, I just dont like corrupting my disk :)

or, better yet, #ifdef out the unmounting/clean-marking, so that the
code is always 'safe,' and note that the #ifdef will be removed when
the code is fixed.

if you update your -current system _completely_, wouldn't you also make
new device nodes (which will trigger the problem, too)?




chris