On Sat, Dec 14, 2013 at 08:27:45PM +0530, V.Krishn wrote: > On Friday, December 13, 2013 08:31:51 PM you wrote: > > Was going through LSB Core 4.1. > > > > I am looking at LSB core section 20.8 ... > > <quote> > > Confirming.... set -e > > </quote> This is a useless quote. You mean: > Conforming scripts shall not specify the "exit on error" option (i.e. > set -e) when sourcing this file, or calling any of the commands thus > made available. From the bugs you cite, the consensus in Debian seems to be that this is an unreasonable requirement which leads to buggy code. However, since it is a requirement, we follow it and try to be very careful to avoid bugs. > > does this mean /etc/init.d/apache2 will not have "set -e" set in it According to LSB that is what it means. According to our own policy, set -e is acceptable, as long as any function returning an error is handled properly. Note that this script does not use set -e in unstable; I don't know if this was a recent change. > Ok, found some references for ongoing efforts: > http://lists.debian.org/debian-devel/2012/04/msg00248.html > and at bugs.debian.org: > #546743 > #562506 > #661002 > #616131 These are all closed bugs, not "ongoing efforts". I'm not sure what your question is. In any case, nobody would complain if lsb-core would be fixed to allow set -e. However, nobody seems to be interested in doing the work either. > It would be nice if in Release Notes(if not too lengthy) or a wiki page lists > short notes pointing to LSB spec, for items having partial or non > implementation. I'm not sure what you mean here. For the next release this is no problem. If you find an init script which uses set -e without handling errors properly, it's a bug which should be reported. Apache2 is not an example of this; it doesn't use set -e. Thanks, Bas
Attachment:
signature.asc
Description: Digital signature