Manuel Bouyer <bouyer%antioche.eu.org@localhost> writes: > On Wed, Mar 25, 2015 at 06:55:52PM -0400, Greg Troxel wrote: >> [...] >> It's not useful to remove these to hide the bug, but they do slow down >> the kernel, especially DEBUG. And normal policy is that DIAGNOSTIC is >> enabled in -current and not on release branches, and DEBUG (and >> LOCKDEBUG, etc.) is always off. > > I think this policy was not already in place when netbsd-5 was > branched. > >> This patch moves netbsd-5 back to the normal policy. I've tested >> kernels with this change: amd64 dom0, i386 domU. I intend to submit a >> pullup request to have this applied; speak up if you think I shouldn't. > > I wonder if this could cause new bugs to show up. netbsd-5 has not > been that well tested without DEBUG/DIAGNOSTIC for Xen ... That's true. We have a plan now to have DIAGNOSTIC enabled on netbsd-7 and to turn it off just at the moment of release. This doesn't feel so different. In theory, DIAGNOSTIC and DEBUG merely add extra checks for things that have gone wrong and if so panic. So turning them off should not cause any usage that works now to start failing I'm not proposing to touch netbsd-5-2. The risk of trouble seems low, and this is easy to back out if so. Are you saying "please don't", or just pointing out that we don't have a lot of experience with netbsd-5 w/o DIAGNOSTIC,DEBUG?
Attachment:
pgpyL4DD3IvrA.pgp
Description: PGP signature