The changes made for #465956 <https://bugs.gentoo.org/show_bug.cgi?id=465956> result in sssd not finding the memberof module, causing sssd to fail to start. Reproducible: Always Steps to Reproduce: 1. Install sssd and ldb-1.1.15-r1 2. Try to start sssd service (preferably in interactive mode with debug messages cranked up) Actual Results: sssd fails to find its memberof module, resulting in the service not started. Expected Results: sssd starts Masking the -r1 revision makes everything work again.
Alternatively, sssd ebuild needs to be changed to the new path.
Created attachment 350800 [details, diff] sssd-1.9.5.ebuild.diff Possible fix which I suggest to do through a revbump.
Hello, I just tried out the patch and it seems to work just fine on one of my machines. Just a note, though: I believe this same bug affects all sssd versions, not just the latest one. Thanks for your help.
(In reply to Lars Wendler (Polynomial-C) from comment #2) > Created attachment 350800 [details, diff] [details, diff] > sssd-1.9.5.ebuild.diff > > Possible fix which I suggest to do through a revbump. Should we also hard depend on the specific ldb version?
Bumping severity since this breaks the authentication of boxes out of the blue Another solution is to modify the init script like this start(){ ebegin "Starting sssd" export LDB_MODULES_PATH=/usr/lib64/ldb/modules/ldb start-stop-daemon --start --exec /usr/sbin/sssd -- -Df eend ${?} }
Well, IMHO (as I noted in the mentioned bug (or rather in bug 447022)), the real problem is the lack of synchronization between ldb, samba and other ldb consumers in Gentoo. Initially, ldb was a part of samba 3 (still is in 3.5.21 ebuild, AFAICT). When samba 4 arrived (as rc and under mask), at the same time it lost most (all ?) active maintainers. As ldb was one of the samba libs that at the time was released as separate tarballs, it (for one reason or another) got a new maintainer, independent of samba herd. Also, the *draft* ebuld for samba 4 has never been properly reviewed (check samba and gvfs/wine - there's *still* useflag mismatch, even tough you can build both against samba 4, if you 'emerge -O' to ignore the downgrade). In the meanwhile, ldb lost its maintainer (at least that's how I see what happened a couple months back). In short, all of the people/herds involved must come up with a general policy on ldb and all of its consumers to solve this longterm. While there is a chance, I might have been wrong in my first point in bug 44702 comment 2, as I see it nobody really tested it one way or another (after all, there is a chance there's a bug in samba build system too) and I don't really use samba.
Fixed. Thanks for the report @ldb maintainers. Please be more careful next time :). I am willing to test the new ldb versions against sssd so feel free to ask me +*sssd-1.9.5-r1 (03 Jul 2013) +*sssd-1.9.4-r3 (03 Jul 2013) + + 03 Jul 2013; Markos Chandras <hwoarang@gentoo.org> +sssd-1.9.4-r3.ebuild, + +sssd-1.9.5-r1.ebuild, -sssd-1.9.4-r1.ebuild, -sssd-1.9.4-r2.ebuild, + -sssd-1.9.4.ebuild, -sssd-1.9.5.ebuild, metadata.xml, sssd-1.8.6-r1.ebuild, + sssd-1.8.6.ebuild: + Revbump to fix sys-libs/ldb dependencies and runtime problems against sys- + libs/ldb-1.1.15-r1. Remove old versions +
