On Oct 18, 10:27am, jmc@NetBSD.org (James Chacon) wrote:
-- Subject: Re: uninitialized vars

| On Tue, Oct 18, 2005 at 11:14:54AM -0400, Christos Zoulas wrote:
| > On Oct 18,  9:15am, jmc@NetBSD.org (James Chacon) wrote:
| > -- Subject: Re: uninitialized vars
| > 
| > | On Tue, Oct 18, 2005 at 03:30:59PM +0200, Jaka Jejcic wrote:
| > | > > We default to -O2 not to -O. Maybe that is your problem.
| > | > 
| > | > yep... that's it.
| > | > Thank you all.
| > | 
| > | Arbitrarily setting the -O level will possibly bite you anyways depending on
| > | places that may tickle compiler bugs at different settings.
| > | 
| > | It's generally safer to let the system -O options remain in place and only
| > | add -m options as needed.
| > 
| > Indeed; when I compiled my sparc64 system with -O6, I had to patch binutils
| > for things to work.
| 
| -O6? Doesn't gcc top out at -O3?

It does, but before it was O2, so by me making it O6, I actually used -O3
which triggered the bug.

christos