On Tue, September 19, 2006 3:54 pm, Andreas Beckmann said: > Hi Mattia, > >> Isn't a little overkill separating the mem= argument from the others? >> Since umlargs is already there I'd simply document that one. > > For now umlargs is sufficient to set mem. > Just in case rootstrap someday needs to increase the default mem limit > because oom kills happen to often - then a separate mem= option might be > helpful, because this could have a default setting and users don't have > to care to preserve/readd a mem= setting when adding their own umlargs.
how do you deal with [global] umlargs=init=/bin/sh mem=256M bla=b mem=128M ? We could just ship a rootstrap.conf with some clear comment on possible issues deriving by too low mem values, but still I'd go with umlargs otherwise we'd need to create separate config directives for each supported UML option to have full control on them. BTW: I'd also say we have one more bug because the --umlargs command line option and umlargs= confg option are not exclusive, I'll modify them to have a well defined precedence not allowing to overlap. -- mattia :wq!
