> On Fri, Apr 17, 2009 at 12:46 PM, Holger Levsen <hol...@layer-acht.org> wrote: > > Hi Thomas, > > > > why dont you want to fix this issue? The provided patch looks sane and > > non-intrusive and is generally helpful, so? > > ACK, +1 > > Thomas, good error checking, and dying early, is part of good software > developpment techniques. >
I'm not sure which position to take here. Yes, of course, error checking is very important and should be done whenever possible. BUT: It is generally very hard with shell snippets. Ok, the case reported by the user will be caught. But what about all the other possible syntax error that are not actually shell syntax errors but just wrong use of those variables? Nevertheless ... > Similar as the "fai-continues-even though-debootstrap-failed" error I > reported yesterday, a program should check what's happening in it's > course as often as possible and useful, and die immediately when it > comes to situations where continuing is not useful. > > Especially when there is even a patch, and a very simple one, that > solves such a problem, I really don't understand your reaction here. > > Please accept it. > ... I agree that this simple patch should have been accepted. But, well, I'd actually propose a somewhat different approach: Why not use set -e . /etc/fai/fai.conf . /etc/fai/make-fai-nfsroot.conf set +e This avoids the problem that sourcing a shell snippet twice may cause: there is no guarantee that it does not do some magic stuff that must not happen twice. Best, Michael
pgpJvgcsuhALs.pgp
Description: PGP signature
