[LV] Partial success and a Linux kernel bug

[Maciej W. Rozycki]

On Wed, 29 Aug 2007, Peter Lund wrote:

> I pushed a revised version of my own patch instead (and cc'ed you).  I
> feel it is worth it to get rid of the spurious 'tmp - 1' and '>> 1'
> gymnastics and to make it crystal clear that it is the shift count that
> causes the trouble.

 Either way is fine with me.  Note that this problem is easy to miss and 
there may be places like this hidden elsewhere.  I would expect most 
people to assume "undefined behaviour" to mean possibly a nonsensical 
result, but not an exception.  Under such an assumption it is safe to 
calculate a useless result and discard it based on the condition that 
makes it useless.

> I see that you ran into this back in January!  Have you got other good
> stuff in your bag?

 Probably. ;-)  Too much stuff around.  Whatever I discover and fix gets 
"queued" in some sense, but the latency is frequently beyond acceptation 
by any means.  Time constraints are the problem and less urgent bits 
suffer.  I tried to rectify it by calling clone() on myself once, but all 
I got was ENOSYS.

 Chances are I will get more up to date with VAX/Linux bits once I start 
the glibc effort I plan.

  Maciej
_______________________________________________
Linux-Vax mailing list
Linux-Vax at pergamentum.com
http://www.pergamentum.com/mailman/listinfo/linux-vax