Seems that somethings have been changed, and this piece of code: case ${CHOST} in *-freebsd*) group=wheel ;; *) group=root ;; esac is killing us, since we are with FreeBSD (there's no root group, but wheel).
And the same for dosbin.
And same goes for DragonFly, *-dragonfly* for chost check.
The same will be for OSX... Probably this time a linux or gnu check would be appropriate.
> Seems that somethings have been changed, and this piece of code: correct, the code before would check to see if USERLAND was GNU and then default to using 'wheel' otherwise ... this goes against the general behavior in portage, that is, we check for non-GNU systems first and then assume everything else is GNU > case ${CHOST} in > *-freebsd*) group=wheel ;; > *) group=root ;; > esac > > is killing us, since we are with FreeBSD (there's no root group, but wheel). how is this killing freebsd ? does your CHOST not match *-freebsd* ?
I'm moving these vars to make.conf in the prefix branch: rootuser rootuid wheelgroup wheelgid portageuser portagegroup the last 2 have no merit for stable portage, but perhaps the root/wheel stuff would be better suited as a configuration option to avoid having to hardcode even more crap.
(In reply to comment #4) > how is this killing freebsd ? does your CHOST not match *-freebsd* ? > It's not killing FreeBSD. It's killing NetBSD, OpenBSD and DragonFly, afaik. We "follow" FreeBSD, in this case.
In similar situtations we've used 0 as the UID or GID in order to sidestep the root/wheel difference. Can that approach apply here or not?
And what about putting "0" as the group for everyone, as Diego has recommended in documentations?
(In reply to comment #7) > In similar situtations we've used 0 as the UID or GID in order to sidestep the > root/wheel difference. Can that approach apply here or not? > Sounds perfect setting this for everyone.
> And what about putting "0" as the group for everyone, as Diego has recommended > in documentations? i dont mind such hacks in backend portage, but i think it's a shitty solution for ebuilds
Created attachment 81077 [details, diff] patch dobin and dosbin to use uid and gid 0 Please test this patch and if it works for everyone then I'll roll it into 2.1_pre5-r3.
(In reply to comment #11) > Created an attachment (id=81077) [edit] > patch dobin and dosbin to use uid and gid 0 > > Please test this patch and if it works for everyone then I'll roll it into > 2.1_pre5-r3. > It's fine.
(In reply to comment #3) > The same will be for OSX... > Probably this time a linux or gnu check would be appropriate. Are there any arguments against the assumption that both the uid and gid should be 0? (In reply to comment #5) > perhaps the root/wheel stuff would be better suited as a configuration option > to avoid having to hardcode even more crap. I agree that the hardcoded aspect sucks, but if it works I can live with it until we decide on a more unified approach like you have in the prefix branch. (In reply to comment #10) > i dont mind such hacks in backend portage, but i think it's a shitty solution > for ebuilds If you can find a unix system that this doesn't work on, then I'll agree that it's a bad solution.
(In reply to comment #13) > Are there any arguments against the assumption that both the uid and gid should > be 0? > Fine for darwin/os x. > If you can find a unix system that this doesn't work on, then I'll agree that > it's a bad solution. I don't see what is so hacky about using 0 (other than the aforementioned hardcoding), it works fine on every system I have tested on(linux,solaris, haiku, macos, and bsd)
Released in 2.1_pre5-r2