On Sat, Mar 11, 2006 at 11:47:25AM +0100, Martin Schulze wrote: > Why would we need "more total CPU time"? Not even leisner is > overloaded at the moment, and it's probably the slowliest machine. > (leisner has a different problem, though). > Hence, please explain why we need "more total CPU time" and when a > downtime from a couple of days maximum is a problem. "Developers accessible machines" are used by human beings which are by nature much less patient and much more subject to real life issues than build daemons. The faster a port machine is, the less painful it is to debug a problem and so developers are more willing to work on it. Fixing the bug sooner given them also more time to work on others bugs and reduce the delay caused by the bug. If there is no port machines available to debug a problem when a developer has a timeslot to work on it, the resolution will have to wait for the next timeslot, which might be more that 3 days latter. A port machine which is not also a buildd performs faster. Personnally, I regularly have to use the port machines to debug gcc incorrect code generation affecting my packages. This is already very painful on a fast machine, but it become a real nightmare on a slow one. > > (db.debian.org do not list gluck as having chroot, and list vore as > > a sparc port machine. However vore seems to be down currently. > > pergolesi has both amd64 chroot and i386 chroot.) > > Hmm, vore should be up. Should be up soon again. This is good to know (though it seems to be still down). > My question stays: Why? > > Of course, we could add all machines that get donated to the Debian > project, but why should we? If we are donated machines that are significantly faster than the developers machines for the same architecture, I think we should provide access to them (unless the machines were affected to another usage, of course). Cheers, -- Bill. <firstname.lastname@example.org> Imagine a large red swirl here.
Description: PGP signature