[OpenAFS] Re: modload failing, Sol10 SPARC, 1.4.14

Andrew Deason adeason@sinenomine.net
Thu, 9 Jun 2011 00:35:26 -0500

On Wed, 1 Jun 2011 09:20:51 -0400
Derrick Brashear <shadow@gmail.com> wrote:

> >> or, for now, just replace that with
> >> memset(dirHeader->hashTable, 0, NHASHENT*(unsigned short));
> >>
> >> and move along?
> >
> > Yeah but, I'd like to get a better solution for when this happens
> > again.  Prohibiting loops like that would be a bizarre requirement
> > :)
> well, i assume oracle will fix their bug, eventually...

To follow up on this a bit... from what I got from a conversation with
Oracle (incident review ID 2050428), they're treating it as a bug that
-xvector=%none doesn't prevent this optimization, so that should
suppress it in the future. In the meantime there's an undocumented
option -W2,-Rbcopy that can suppress the optimization, or of course we
can remove the loop (gerrit 4829).

I don't really see why they're not going the route of an "hey I'm
building for the kernel" flag and making the compiler deal, but eh.

Andrew Deason