* Alexey Sokolov ([email protected]) wrote: > No, the reason I reverted it is > https://github.com/znc/znc/issues/1165#issuecomment-156880251 (point 2.)
I don't really understand this argument. Using datadir (or pkgdatadir) allows the builder to set it to whatever value they wish at build time, how is that not generic? I think hiding the system configuration inside of a .znc directory is surprising and IMHO we should avoid it. You could provide a different automake variable to configure this and default it ~/.znc. Alternatively, you could add a --system flag (or something) to change the defaults if you really want these paths built into the binary. > To repeat myself: If you guys (packagers) all agree on how this file > should look like, I'm fine with it too. > > If you want to maintain a separate unit file for Debian which is > different from other distros, I'm not in a position to stop you from > doing this. I don't have strong feelings about this, and I'm not the maintainer. I would just like this to work out of the box on Debian :) Patrick what do you think? > I'm happy to accept changes upstream, if that makes packagers' life > easier. Cool. Outside of packaging, there are probably some hardening options that would be useful to add. Have you considered adding socket activation support, sd_notify and syslogging support to make this work a little better with systemd? I'd be happy to contribute some patches. > You can join that discussion if you disagree with @seblu and @Philantrop > > As for _znc username, perhaps that could be either a Debian-specific > one-line patch, or e.g. a env variable used by ./configure... Either of those are fine with me, but the more the service file is configurable this way the less it makes sense to me to try to keep an authoritative service file. We also don't need to use _znc, but znc instead. But znc being so short it does seem like it could conflict with an existing user on the system. > 29.12.2015 22:30, Eric Dorland пишет: > > * Alexey Sokolov ([email protected]) wrote: > >> Hi Eric, Does the service file need to be different from the > >> included one? > >> (https://github.com/znc/znc/blob/master/znc.service.in) > > > > Currently it does, because the patch as is creates a _znc user > > rather than znc to avoid namespace collision. It also uses the > > --datadir flag to avoid the .znc directory. It also uses > > ConditionFileNotEmpty to prevent startup if it's not configured. We > > don't have to do these things but they all seem useful to me. > > There's really room to go further with various systemd protection > > settings, but I didn't want to overcomplicate the diff. > > > >> Also please check the recent discussion about it at > >> https://github.com/znc/znc/issues/1165 > > > > Thanks for pointing this out. There's a lot of issues to intermixed > > in this bug report. It looks like the reason they reverted the > > --datadir change was because it broke existing users, which is fair > > but we don't have that problem in Debian. I think having a unified > > unit file would be good but I don't think that should tie hands > > against good improvements. > > > -- Eric Dorland <[email protected]> 43CF 1228 F726 FD5B 474C E962 C256 FBD5 0022 1E93
signature.asc
Description: PGP signature
