I did report it to bugs. Ted took care of it in -current apparently. I still haven't looked into getopt really well. I need root to write into /etc/pkg.conf it privilege separates and drops privileges before it gets to network ftp calls. I'm not claiming it's the best it can be right now. But I think it is useful and possibly useful in edge cases in the base. I have been rewriting it a lot and it outputs more than it needs to because I have been debugging new features. I think it would maybe in the future, be an attractive feature for a new install that doesn't have a mirror set, and the user doesn't know which mirror is the best, who may just choose the main install mirror for the perceived stability over minimized latency; overloading the main mirrors. I've seen pkg_add just sit there doing nothing when it had a bad mirror and at least there ought to be some sanity checking, if not at least some manual interface to choosing a new mirror. Especially since web browsers aren't in the base and looking up /etc/examples/pkg.conf, which has cymru, which is dead and others which don't work, isn't the most obvious for people that aren't aware of its existence. I think it only exists, because a guy made it as an answer to my previous program that got it in an insecure manner from openbsd.org/ftp.html I'm not sure why there is resistance to automating mirror selection, even if it could be a manual choice to do so.
On Thu, Aug 25, 2016, 09:26 Jeremie Courreges-Anglas <j...@wxcvbn.org> wrote: > Luke Small <lukensm...@gmail.com> writes: > > > It calls ftp multiple times to find the relative speed and latency of a > > PKG_PATH mirrors found in /etc/examples/pkg.conf and if it is run as > root, > > provides automatic, fine-grained excision of installpath options and > > attributed comments. > > My opinion is that this is still a toy project. This will not be > included in base/, and I would object to an inclusion in the ports tree. > > > It is pledged in many places, provides privilege > > separation, but I run 5.9 and a glitch that is fixed in current forbids > > using NOTE_EXIT in kqueue while using setuid executables. > > What you're saying is scary. In a perfect world no program runs > setuid-root. > > Anyway, if you noticed a behavior change that breaks userland, please > report on bugs@. > > > I need NOTE_EXIT > > to time how long it takes for ftp to run and to kill it if it takes too > > long. It is a fun little utility! > > No you do not *need* NOTE_EXIT. You do not need kqueue(2), you do not > need C, you do not need to rewrite getopt(3) from scratch, etc etc. > > Sorry if it hurts, but please do not waste people's time about this toy > project. > > -- > jca | PGP : 0x1524E7EE / 5135 92C1 AD36 5293 2BDF DDCC 0DFA 74AE 1524 E7EE >
