Open GoogleCodeExporter opened 9 years ago
Charle why don´t reply, it´s not a normal comportament, i ask information for
this deadlock
Original comment by kristofwycz
on 4 Oct 2014 at 9:40
Well, it was a thoughtful question.
There's nothing really wrong about the proposed fix, but it side-steps the
problem by letting the processes complete. That's fine, because the underlying
problem is an ancient one, of how to make code (critical sections in this case)
uninterruptible when they must be for correctness, without making the
containing process impossible to kill, although there are approximations that
would be fine.
Here, the problem is that we hacked a lock in to Attrdb that suits a
shared-memory environment without kills, when Inferno offers alternatives.
Original comment by Charles....@gmail.com
on 8 Oct 2014 at 9:07
So what would the be right way to fix Attrdb? The ways of removing the lock
that I considered would have Attrdb spawn a process that won't necessarily get
cleaned up by existing programs.
Original comment by kristofwycz
on 9 Oct 2014 at 4:58
Original issue reported on code.google.com by
kristofwycz
on 2 Sep 2014 at 11:23