Date: Wed, 24 Jul 2002 20:48:32 +0100 From: Matthew Whelan <muttley@gotadsl.co.uk> To: "Brian T. Schellenberger" <bts@babbleon.org>, Jamie Bowden <ragnar@sysabend.org> Cc: freebsd-stable@FreeBSD.ORG Subject: Re: make world considered harmful Message-ID: <YXVUXSA91TYSYSNLG971USP5FD04VR.3d3f0490@VicNBob> In-Reply-To: <Pine.BSF.4.10.10207240646200.47292-100000@moo.sysabend.org>
next in thread | previous in thread | raw e-mail | index | archive | help
24/07/2002 14:59:42, Jamie Bowden <ragnar@sysabend.org> wrote: >On Wed, 24 Jul 2002, Brian T. Schellenberger wrote: > >:On Wednesday 24 July 2002 07:01 am, Jamie Bowden wrote: > >[make world, good, bad, or ugly?] > >:make buildworld >:make kernel >:reboot >:[see if the kernel is sane before . . .] >:make installworld >:mergemaster > >:That way if the new kernel is bad, you still have time to repair it with your >:old, known-to-be-working world. If you do the whole world at once and it's >:bad, you are looking at re-installing from scratch. > >Except when you have to run mergemaster before you can even buildworld >(see the recent commit of the new and improved sendmail). You're always >facing the possibility of running into some chicken and egg problem. If >you want total safety, run RELENG_4_6. But the documented process is the least-risk method. Even if you successfully upgrade the dangerous way 1,000 times for every breakage, the time you saved in typing it the shorter way is completely dwarfed by the time it takes to restore your system. Further, in the case you mentioned, running mergemaster first didn't require a change in the order of the remaining steps - and wasn't in itself fully necessary (you could add the new users by hand and then do the full mergemaster as per the documented procedure) Also, I always back up my /etc/ prior to running mergemaster, so backing out of that is pretty trivial, again, much easier than backing out of a broken world. I've a hunch I got the idea from reading the handbook. >Anyone running -stable or -current should well be prepared to endanger >themselves regularly. I endanger my system in this manner several times a >week. I've been using this method (make world)since early 2.0 builds, and >if I break my system beyond repair (which I have done), I did so knowing >it was a distinct possiblity. This doesn't mean that everyone should take undue risk. make world survived in large part because it was already there, I imagine. People like you are able to make the value judgement, but others with less experience aren't as well placed. They are the people we're concerned about here. Whilst in general I'm a strong believer that breakage encountered by people deviating from clear instructions should be very low on the priority list, in this the benefit to the shortcut when its safe doesn't justify the damage when it's not. The inconvenience to users in losing the shortcut is low, so the best case for keeping it is probably developer time. It seems to me the change to remove the rule itself would be very quick - updating docs and the make release process to match a little less so but still not exactly a mountain of effort. [re-insert from Brian's post] >:Personally, I think it would be better to remove it; for those who dislike >:typing and don't mind endangering their system, it would be better to have >:instead a >: >:make universe Maybe 'make bravenewworld'? Matthew To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?YXVUXSA91TYSYSNLG971USP5FD04VR.3d3f0490>