On Sat, 2011-04-30 at 23:22 +0100, Steve McIntyre wrote: > On Wed, Apr 13, 2011 at 02:12:27PM +0100, Ben Hutchings wrote: > >Package: debian-cd > > > >(Not sure which version is actually used for squeeze.) > > > >On i386 machines with RAM above the 4GB physical address, either PAE or > >LM (Long Mode, 64-bit) must be used to access that RAM. The preferred > >kernel flavour is '686-bigmem', with 'amd64' also being an option. > > > >However the amd64 flavour is on DVD 1 but the 686-bigmem flavour is on > >DVD 2. This means a single-DVD install with no network access will get > >the amd64 flavour if the system supports LM and the 686 flavour if it > >does not. The former is somewhat undesirable as third-party installers > >(and users) can get confused by a 64-bit kernel with 32-bit userland. > >The latter is extremely undesirable as some RAM is then inaccessible. > > > >Similarly the amd64 flavour is currently on CD 5 but the 686-bigmem > >flavour is on CD 10. The 686-bigmem flavour should be increased in > >priority, certainly above the amd64 flavour, and should ideally be on CD > >1 or 2 (but I realise that there is tough competition for CD 1). > > I've added linux-image-686-bigmem to > tasks/squeeze/interesting-fromcd23, which means that the -bigmem > kernels will get pulled in before packages from popcon. A test build > looks good, and I've back-ported the fix back to the squeeze branch of > debian-cd so that 6.0.2 should be correct. Doest this have an impact (i.e. perhaps make it obsolete) the VARIANT_xen stuff e.g. in tools/generate_di+k_list? I suppose the variant stuff is orthogonal to this issue, even if the net result is the same and so it's worth having both by way of explicitly asking for what is required in both cases. Ian. -- Ian Campbell I am currently transitioning to a new OpenPGP key, please see: http://www.hellion.org.uk/key-transition-2011-04-27-2F6BCD59-to-79074FA8.txt If you have to think twice about it, you're wrong.
Description: This is a digitally signed message part