> Some regexps there are extended, contrary to what Karl > stated in his comment. And, by the way, I think that > having: > > # Assume there is a root password if NIS is being used. > > if egrep -q '^+:' /etc/passwd; then > is a plain bug, because intent of that code is to match
Sure, this is a plain bug, definitely. > literal "+:" at the beginning of line (AFAIK), but the > "+" symbol is treated as metacharacter in extended regexp > context. > That's why "egrep -q '^[+]:'" or "grep -q '^+:'" _must_ > be used here instead. > > IMHO we need to examine the scripts affected (and the > patch) more thoroughly... > > BTW, the patch already contains the changes I proposed > to remove-shell/add-shell scripts (namely the > "#!/bin/bash -e" --> "#!/bin/sh -e" move). Yep, when looking at that patch, I remembered this. Maybe merge the two bugs together? Do you take that bug, Alex? If so, don't forget thet patch I sent for #245332, which impacts passwd.config. The passwd.config thing is not easy to test because most of it is only used at install time. You have to hack it down a little to make it testable on a live system. PS�: while I am at it, feel free to object to me nicknaming you guys. I now see that I began writing "Alex" or "Nico" for Alexander and Nicolas respectively. This is just a kind of automatism, which saves me a few hits but I don't want to hurt you in any way. A lot of French "Nicolas" are often nicknamed "Nico" but, well, Nicolas maybe dislikes this....same for you, Alexander, even if I indeed don't know whether Ukrainian Alexanders are usually nicknamed "Alex"...:-) Martin, I have no nickname for you...:)
