Description of problem: LSB, since 3.2, has a series of tests that check the functionality of the install_initd, remove_initd scripts. Systems based on Red Hat's chkconfig typically fail 2 of these tests, where a strict dependency on another service is involved. Version-Release number of selected component (if applicable): chkconfig-1.3.38-1 How reproducible: 100% Steps to Reproduce: 1. install/run lsb-test-core: http://ftp.linuxfoundation.org/pub/lsb/repositories/yum/lsb-4.0/repo-ia32/lsb-test-core-3.97.0-1.i486.rpm 2. 3. Actual results: /tset/initd/misc/install/general-tc 3 FAIL /tset/initd/misc/install/general-tc 5 FAIL Expected results: All tests should pass Additional info: Some patches/mods were applied to chkconfig around 2005, when I did some work on this for Mandriva and sent it upstream. For a while Mandriva passed these tests, but they have since backed off some of the changes. A patch against 1.3.20 I used for LSB's SI build can be found here: http://bzr.linux-foundation.org/lsb/devel/si?cmd=content;rev=stewb%40linux-foundation.org-20080428123712-jel9rr75eut4om30;pathrevid=pqm%40linux-foundation.org-20080620112025-scvq05mfzg74weym;path=build/patches/chkconfig-1.3.20-lsb-si.patch Bug at the LSB end (with more journal detail): http://bugs.linuxbase.org/show_bug.cgi?id=2443 Spec reference: http://refspecs.linux-foundation.org/LSB_3.2.0/LSB-Core-generic/LSB-Core-generic/initscrcomconv.html
Fixed in 1.3.39-1. You can follow the commits at http://git.fedorahosted.org/git/?p=chkconfig.git ; some of the code may be cleaned up later. This is enabled *only* if it's called as install_initd/remove_initd; it's really bad behavior to do this in the general case.
Grabbed a git snapshot of 1.3.40 and built it to test on F10/x86_64, now test 3,4 fail: /tset/initd/misc/install/general-tc 3 FAIL /tset/initd/misc/install/general-tc 4 FAIL ref: http://git.fedorahosted.org/git/?p=chkconfig.git;a=snapshot;h=74ab891968c5ba0e5d43ab16c5edba578c425f62;sf=tgz
Current git (as of 4ed064e287e2dd04290c7cc503b38e2f43693118) should solve both failures. FYI - test #2 in the test suite is not really correct - an implementation could do all dependency resolution at runtime and ignore lexical sorting entirely. (Of course, I'm not sure there are any implementations that actually *do* this.)
I'm still seeing issues using http://git.fedorahosted.org/git/?p=chkconfig.git;a=snapshot;h=bdef046f4feb0340e15bd02fbe45d7d1e3050405;sf=tgz Built chkconfig on F10, x86_64 and ran just the init.d tests (I'll attach a patch to re-enable tests 3,5, which we've disabled due to this issue): /tset/initd/misc/install/general-tc 2 FAIL /tset/initd/misc/install/general-tc 5 FAIL [root@fc10-64 test_sets]# /sbin/chkconfig --version chkconfig version 1.3.42 [root@fc10-64 test_sets]# cat /etc/redhat-release Fedora release 10 (Cambridge) [root@fc10-64 test_sets]# pwd /opt/lsb/test/core/tet/test_sets [root@fc10-64 test_sets]# cat scen.exec all "total tests in LSBinitd 15" /tset/initd/misc/utils/utils-tc /tset/initd/misc/init-functions/init-functions-tc /tset/initd/misc/install/general-tc
Created attachment 334515 [details] re-enable initd tests 3,5, apply from /opt/lsb/test/core
Fixed. All the tests (run individually, setting up the test suite was too painful) pass for me. http://git.fedorahosted.org/git/?p=chkconfig.git;a=commitdiff;h=fa16d756ce801b2dfbff25972f43989398ad5317
Thanks, current git build also passed here.
*** Bug 522053 has been marked as a duplicate of this bug. ***