On Thu, Aug 2, 2012 at 8:40 PM, Jan Steffens <jan.steffens@gmail.com> wrote:
On Fri, Aug 3, 2012 at 2:01 AM, Ray Kohler <ataraxia937@gmail.com> wrote:
As it's been marked out-of-date for a while, and I'm interested in it, I've done the work to upgrade the polkit package to 0.107. I've tested this package locally. The main value of this upgrade (for me at least) is systemd integration, including proper systemd-style bus activation.
Unfortunately, enabling systemd support removes consolekit support, screwing over everyone not running systemd.
Ah, I didn't know about that. Are you sure that's still the model used here with this upstream support? How would I test? I am running systemd, but I also use consolekit.
This now wants a dedicated user and group, by default 'polkitd'. I re-used 102, which was listed as 'policykit' in the UID/GID database. Hopefully it really was unused. The build depends on this user and group already being present (otherwise 'make install' fails to chown some directories, and polkitd fails to read them and dies).
Ah, that sucks. Any suggestions on how to handle this when building in clean chroots?
I did my builds by building a broken package first (the 'make install' failure doesn't actually cause the build to fail), installing that (which created the user and group), and then building it again (getting a working package this time). Not very satisfactory, and not workable in a build chroot. Possibly it would work to allow the directories to get the wrong owner, and then chown them after make install, using the intended numeric UID/GID directly. Not very safe, as we might have polkitd:polkitd being something other than 102:102 on some systems for some reason.