packages with portupgrade update on release
in [FreeBSD]
|
Prev: NYC LOCAL: Wednesday 2 June 2010 NYCBUG: Nikolai Fetissov on the Use of GDB for Systems Administration
Next: freebsd support for aws ec2
From: Indi on 3 Jun 2010 07:23 On 2010-06-03, Michel Talon <talon(a)lpthe.jussieu.fr> wrote: > Chronos <me3(a)privacy.net> wrote: >> Michel Talon wrote: >> >> > Indi <indi(a)satcidananda.16x108.merseine.nu> wrote: >> >> You should try runing pkdb -fF, update your ports tree and try >> >> portupgrade -avf >> >> >> >> But you might have to deinstall libpng (and possibly libjpg) and >> >> everything that depends on it then build it fresh. >> >> >> > >> > Fantastic! I had always beleived that the FreeBSD ports system was >> > foolproof and only idiots like me had problems with it. Another >> > newby coming to grips with reality... >> >> portupgrade != the ports system. portupgrade has foibles of its own. > > Yes, this is true. But the ports system itself has problems such that > no automatic upgrade system can work reliably. By work, i mean, not > recompile everything, of course. > Please name the OS that automagically updates everything by building it and does that perfectly every time. -- Caveat utilitor, indi
From: Balwinder S Dheeman on 3 Jun 2010 15:27 On 06/03/2010 04:53 PM, Indi wrote: > On 2010-06-03, Michel Talon <talon(a)lpthe.jussieu.fr> wrote: >> Chronos <me3(a)privacy.net> wrote: >>> Michel Talon wrote: >>> >>>> Indi <indi(a)satcidananda.16x108.merseine.nu> wrote: >>>>> You should try runing pkdb -fF, update your ports tree and try >>>>> portupgrade -avf >>>>> >>>>> But you might have to deinstall libpng (and possibly libjpg) and >>>>> everything that depends on it then build it fresh. >>>>> >>>> >>>> Fantastic! I had always beleived that the FreeBSD ports system was >>>> foolproof and only idiots like me had problems with it. Another >>>> newby coming to grips with reality... >>> >>> portupgrade != the ports system. portupgrade has foibles of its own. >> >> Yes, this is true. But the ports system itself has problems such that >> no automatic upgrade system can work reliably. By work, i mean, not >> recompile everything, of course. >> > > Please name the OS that automagically updates everything by building > it and does that perfectly every time. Nothing is perfect in this world, not even the God ;) Anyway, downloading source dist files, extracting and building source based packages and, or ports *just for an upgrade and, or update* seems a height of stupidity to me; users who do not know and, or are not competent enough to fine-tune a build lose a lot of CPU cycles, disk IO and, or bandwidth. On most of the Linux systems, users not rebuild even a kernel these days, because there not such thing as GENERIC kernel there. That's just an opinion only and opinions of course may differ, but never ever got a satisfactory response on what FreeBSD, Gentoo and, or other source based distributions' users gain by wasting the above said resources. -- Balwinder S "bdheeman" Dheeman Registered Linux User: #229709 Anu'z Linux(a)HOME (Unix Shoppe) Machines: #168573, 170593, 259192 Chandigarh, UT, 160062, India Plan9, T2, Arch/Debian/FreeBSD/XP Home: http://werc.homelinux.net/ Visit: http://counter.li.org/
From: Indi on 3 Jun 2010 15:34 On 2010-06-03, Chronos <me3(a)privacy.net> wrote: > Indi wrote: > >> Anyway, your argument is a lot like: >> "Lamborghinis only average about 13MPG and that makes them terrible >> sports cars. The Prius is much better for my needs, it gets really >> good mileage and has plenty of luggage room. >> Therefore I must make them see the wisdom of usng Priuses in F1, >> it makes perfect sense!" > > May I just try a modification to your analogy? I assume it is BSD > licensed... > > "I've sold my Lamborghini, bought a Prius, but I'm going to hang > around the Lamborghini dealership, whine, moan and turn potential > customers away until Lamborghini make a car that is more like the > Prius, even though most of their existing customers didn't want a > Prius or, for that matter, the inferior copy of a G-Wiz [1] that is > likely to result from the changes I propose." > > That's the bit that seems illogical to me. > > [1] Introducing the G-Wiz, http://www.goingreen.co.uk/ a car so feeble > that you can have motion or lights, but not both. Those of us with a > brain have to ask where the electricity comes from? Zero carbon > electricity fairies? Spot on, thank you. :) -- Caveat utilitor, indi
From: Mark Andrews on 3 Jun 2010 22:35 portupgrade would work much better if there was a simple away to say "rebuild everything that has changed and everything that depends on those changes." This gets rid of lots of subtle version dependancy bugs. "portupgrade -raf" would seem to match that description but it behaves like "portupgrade -af". Mark
From: Michel Talon on 4 Jun 2010 04:06
Mark Andrews <marka(a)drugs.dv.isc.org> wrote: > portupgrade would work much better if there was a simple away to > say "rebuild everything that has changed and everything that depends > on those changes." This gets rid of lots of subtle version dependancy > bugs. The real problem, is that, thanks to human error in some ports, this information is not always correct. In some cases, the information is not here at all, because it is supposed to be collected from the file UPDATING. And in other cases, some ports are massively marked to be rebuilt when it is not necessary, causing huge pain. This is the reason why i am saying the the ports system itself has problems. In my opinion, the existence of the file UPDATING is, in itself, an heresy. It is an admission of weakness, and an obvious obstacle to automatic updating. The first urgent step should be to delete this file, and do whatever necessary to replace it by necessary metadata in the ports system. For the human errors, unfortunately there is no obvious solution, except to have formal testing procedures, like the "unstable, testing, stable" system used by Debian, which has other problems, and the FreeBSD team doesn't have the necessary man power. > > "portupgrade -raf" would seem to match that description but it > behaves like "portupgrade -af". > > Mark Portupgrade is and has always been a very fragile program, very buggy, very slow, a huge disservice to FreeBSD. I think that portmaster is more rustic, but at least more reasonably bugfree. If one accepts building from source, i don't see many objections to portmaster. Still i am more and more convinced that a system based on building from source will never be reliable, because the team doesn't check the result of the the build procedure before releasing, except in an informal way, which opens the door to all sorts of problems. -- Michel TALON |